neurological disease research papers

Articles will undergo the journal’s standard peer-review process  and are subject to all of the journal’s standard  policies . Articles will be added to the Collection as they are published.

The Guest Editors have no competing interests with the submissions which they handle through the peer review process. The peer review of any submissions for which the Guest Editors have competing interests is handled by another Editorial Board Member who has no competing interests.

Advertisement

Relationship of Sleep Disorder with Neurodegenerative and Psychiatric Diseases: An Updated Review

• Published: 18 December 2023
• Volume 49 , pages 568–582, ( 2024 )

Cite this article

• Xiao Xiao 1 , 2   na1 ,
• Yimin Rui 1 , 2   na1 ,
• Yu Jin 2 &
• Ming Chen 2  

633 Accesses

2 Citations

10 Altmetric

Explore all metrics

Sleep disorders affect many people worldwide and can accompany neurodegenerative and psychiatric diseases. Sleep may be altered before the clinical manifestations of some of these diseases appear. Moreover, some sleep disorders affect the physiological organization and function of the brain by influencing gene expression, accelerating the accumulation of abnormal proteins, interfering with the clearance of abnormal proteins, or altering the levels of related hormones and neurotransmitters, which can cause or may be associated with the development of neurodegenerative and psychiatric diseases. However, the detailed mechanisms of these effects are unclear. This review mainly focuses on the relationship between and mechanisms of action of sleep in Alzheimer’s disease, depression, and anxiety, as well as the relationships between sleep and Parkinson’s disease, Huntington’s disease, and amyotrophic lateral sclerosis. This summary of current research hotspots may provide researchers with better clues and ideas to develop treatment solutions for neurodegenerative and psychiatric diseases associated with sleep disorders.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

The missing link between sleep disorders and age-related dementia: recent evidence and plausible mechanisms.

Feng Zhang, Rujia Zhong, … Weidong Le

Sleep as a Therapeutic Target in the Aging Brain

Thierno M. Bah, James Goodman & Jeffrey J. Iliff

Sleep in Neurodegenerative Disorders

Roneil Malkani & Hrayr Attarian

Abbreviations

Amyloid β-protein

Alzheimer’s disease

Amyotrophic lateral sclerosis

Amyloid precursor protein

Aquaporin 4

Acute sleep deprivation

Β-site APP cleaving enzyme 1

Brain-derived neurotrophic factor

Brain and muscle arnt-like protein-1

Chronic sleep deprivation

Cerebrospinal fluid

Chronic sleep restriction

Huntington’s disease

Hypothalamic-pituitary-adrenal

Interleukin

Interstitial fluid

Kelch-like ECH-associated protein 1

Low-density lipoprotein receptor-related protein 1

Mild cognitive impairment

Normal cognition

Non-rapid eye movement

Nuclear factor erythroid 2-related factor

Obstructive sleep apnea

Parkinson’s disease

Receptor of advanced glycation end products

Rapid eye movement sleep-behavior disorder

Rapid eye movement

Sleep deprivation

Sleep-disordered breathing

Slow-wave activity

Slow wave sleep

43-kDa TAR DNA-binding protein

Total sleep deprivation

Grandner MA, Fernandez FX (2021) The translational neuroscience of sleep: a contextual framework. Science 374:568–573

Article   CAS   PubMed   PubMed Central   ADS   Google Scholar  

Voysey Z, Fazal SV, Lazar AS, Barker RA (2021) The sleep and circadian problems of Huntington’s Disease: when, why and their importance. J Neurol 268:2275–2283

Article   CAS   PubMed   Google Scholar  

Joiner WJ (2018) The neurobiological basis of Sleep and Sleep disorders. Physiol (Bethesda) 33:317–327

CAS   Google Scholar  

Léger D, Bayon V (2010) Societal costs of insomnia. Sleep Med Rev 14:379–389

Article   PubMed   Google Scholar  

Hillman DR, Lack LC (2013) Public health implications of sleep loss: the community burden. Med J Aust 199:S7–10

Latreille MKP V (2019) Sleep disorders. Am J Med 132:292–299

Liew SC, Aung T (2021) Sleep deprivation and its association with diseases- a review. Sleep Med 77:192–204

Banks S, Dinges DF (2007) Behavioral and physiological consequences of sleep restriction. J Clin Sleep Med 3:519–528

Article   PubMed   PubMed Central   Google Scholar  

Umemura GS, Furtado F, Dos Santos FC, Gonçalves B, Forner-Cordero A (2022) Is Balance Control affected by Sleep Deprivation? A systematic review of the impact of Sleep on the control of balance. Front Neurosci 16:779086

Pires GN, Bezerra AG, Tufik S, Andersen ML (2016) Effects of acute sleep deprivation on state anxiety levels: a systematic review and meta-analysis. Sleep Med 24:109–118

Colwell CS (2021) Defining circadian disruption in neurodegenerative disorders. J Clin Invest 131

Fang H, Tu S, Sheng J, Shao A (2019) Depression in sleep disturbance: a review on a bidirectional relationship, mechanisms and treatment. J Cell Mol Med 23:2324–2332

Wang YQ, Li R, Zhang MQ, Zhang Z, Qu WM, Huang ZL (2015) The neurobiological mechanisms and treatments of REM sleep disturbances in Depression. Curr Neuropharmacol 13:543–553

Article   CAS   PubMed   PubMed Central   Google Scholar  

Guo B, Zhang M, Hao W, Wang Y, Zhang T, Liu C (2023) Neuroinflammation mechanisms of neuromodulation therapies for anxiety and depression. Transl Psychiatry 13:5

Gauthier S, Webster C, Servaes S, Morais JA, Rosa-Neto P (2022) World Alzheimer Report 2022: Life after diagnosis: Navigating treatment, care and support. In. Alzheimer’s Disease International, London, England

Okuda S, Tetsuka J, Takahashi K, Toda Y, Kubo T, Tokita S (2019) Association between sleep disturbance in Alzheimer’s Disease patients and burden on and health status of their caregivers. J Neurol 266:1490–1500

Kastanenka KV, Hou SS, Shakerdge N, Logan R, Feng D, Wegmann S, Chopra V, Hawkes JM, Chen X, Bacskai BJ (2017) Optogenetic restoration of disrupted slow oscillations halts amyloid deposition and restores Calcium Homeostasis in an animal model of Alzheimer’s Disease. PLoS ONE 12:e0170275

Suh SW, Han JW, Lee JR, Byun S, Kwon SJ, Oh SH, Lee KH, Han G, Hong JW, Kwak KP, Kim BJ, Kim SG, Kim JL, Kim TH, Ryu SH, Moon SW, Park JH, Seo J, Youn JC, Lee DY, Lee DW, Lee SB, Lee JJ, Jhoo JH, Kim KW (2018) Sleep and cognitive decline: a prospective nondemented elderly cohort study. Ann Neurol 83:472–482

Lucey BP, Wisch J, Boerwinkle AH, Landsness EC, Toedebusch CD, McLeland JS, Butt OH, Hassenstab J, Morris JC, Ances BM, Holtzman DM (2021) Sleep and longitudinal cognitive performance in preclinical and early symptomatic Alzheimer’s Disease. Brain 144:2852–2862

Maestri M, Carnicelli L, Tognoni G, Di Coscio E, Giorgi FS, Volpi L, Economou NT, Ktonas P, Ferri R, Bonuccelli U, Bonanni E (2015) Non-rapid eye movement sleep instability in mild cognitive impairment: a pilot study. Sleep Med 16:1139–1145

Pase MP, Himali JJ, Grima NA, Beiser AS, Satizabal CL, Aparicio HJ, Thomas RJ, Gottlieb DJ, Auerbach SH, Seshadri S (2017) Sleep architecture and the risk of incident Dementia in the community. NEUROLOGY 89:1244–1250

Lucey BP, McCullough A, Landsness EC, Toedebusch CD, McLeland JS, Zaza AM, Fagan AM, McCue L, Xiong C, Morris JC, Benzinger TLS, Holtzman DM (2019) Reduced non-rapid eye movement sleep is associated with tau pathology in early Alzheimer’s disease. Sci Transl Med 11

Liu S, Meng Y, Wang N, Lei Q, He L, Hong X, Li Z (2023) Disturbance of REM sleep exacerbates microglial activation in APP/PS1 mice. Neurobiol Learn Mem 200:107737

Chylinski D, Van Egroo M, Narbutas J, Muto V, Bahri MA, Berthomier C, Salmon E, Bastin C, Phillips C, Collette F, Maquet P, Carrier J, Lina JM, Vandewalle G (2022) Timely coupling of sleep spindles and slow waves linked to early amyloid-β burden and predicts memory decline. eLife 11

Chylinski DO, Van Egroo M, Narbutas J, Grignard M, Koshmanova E, Berthomier C, Berthomier P, Brandewinder M, Salmon E, Bahri MA, Bastin C, Collette F, Phillips C, Maquet P, Muto V, Vandewalle G (2021) Heterogeneity in the links between sleep arousals, amyloid-β, and cognition. JCI Insight 6

Winer JR, Deters KD, Kennedy G, Jin M, Goldstein-Piekarski A, Poston KL, Mormino EC (2021) Association of short and long sleep duration with Amyloid-β Burden and Cognition in Aging. JAMA Neurol 78:1187–1196

Fan L, Xu W, Cai Y, Hu Y, Wu C (2019) Sleep duration and the risk of Dementia: a systematic review and Meta-analysis of prospective cohort studies. J Am Med Dir Assoc 20:1480–1487e1485

Leng Y, Ackley SF, Glymour MM, Yaffe K, Brenowitz WD (2021) Genetic risk of Alzheimer’s Disease and Sleep Duration in Non-demented elders. Ann Neurol 89:177–181

Xu W, Tan L, Su BJ, Yu H, Bi YL, Yue XF, Dong Q, Yu JT (2020) Sleep characteristics and cerebrospinal fluid biomarkers of Alzheimer’s Disease pathology in cognitively intact older adults: the CABLE study. Alzheimers Dement 16:1146–1152

Cavaillès C, Berr C, Helmer C, Gabelle A, Jaussent I, Dauvilliers Y (2022) Complaints of daytime sleepiness, insomnia, hypnotic use, and risk of Dementia: a prospective cohort study in the elderly. Alzheimers Res Ther 14:12

Anderson EL, Richmond RC, Jones SE, Hemani G, Wade KH, Dashti HS, Lane JM, Wang H, Saxena R, Brumpton B, Korologou-Linden R, Nielsen JB, Åsvold BO, Abecasis G, Coulthard E, Kyle SD, Beaumont RN, Tyrrell J, Frayling TM, Munafò MR, Wood AR, Ben-Shlomo Y, Howe LD, Lawlor DA, Weedon MN, Davey Smith G (2021) Is disrupted sleep a risk factor for Alzheimer’s Disease? Evidence from a two-sample mendelian randomization analysis. Int J Epidemiol 50:817–828

André C, Rehel S, Kuhn E, Landeau B, Moulinet I, Touron E, Ourry V, Le Du G, Mézenge F, Tomadesso C, de Flores R, Bejanin A, Sherif S, Delcroix N, Manrique A, Abbas A, Marchant NL, Lutz A, Klimecki OM, Collette F, Arenaza-Urquijo EM, Poisnel G, Vivien D, Bertran F, de la Sayette V, Chételat G, Rauchs G (2020) Association of Sleep-Disordered Breathing with Alzheimer Disease biomarkers in Community-Dwelling older adults: a secondary analysis of a Randomized Clinical Trial. JAMA Neurol 77:716–724

Lee JE, Yang SW, Ju YJ, Ki SK, Chun KH (2019) Sleep-disordered breathing and Alzheimer’s Disease: a nationwide cohort study. Psychiatry Res 273:624–630

Liguori C, Maestri M, Spanetta M, Placidi F, Bonanni E, Mercuri NB, Guarnieri B (2021) Sleep-disordered breathing and the risk of Alzheimer’s Disease. Sleep Med Rev 55:101375

Bubu OM, Andrade AG, Umasabor-Bubu OQ, Hogan MM, Turner AD, de Leon MJ, Ogedegbe G, Ayappa I, Jean-Louis GG, Jackson ML, Varga AW, Osorio RS (2020) Obstructive sleep apnea, cognition and Alzheimer’s Disease: a systematic review integrating three decades of multidisciplinary research. Sleep Med Rev 50:101250

Sharma RA, Varga AW, Bubu OM, Pirraglia E, Kam K, Parekh A, Wohlleber M, Miller MD, Andrade A, Lewis C, Tweardy S, Buj M, Yau PL, Sadda R, Mosconi L, Li Y, Butler T, Glodzik L, Fieremans E, Babb JS, Blennow K, Zetterberg H, Lu SE, Badia SG, Romero S, Rosenzweig I, Gosselin N, Jean-Louis G, Rapoport DM, de Leon MJ, Ayappa I, Osorio RS (2018) Obstructive sleep apnea severity affects amyloid Burden in Cognitively Normal Elderly. A longitudinal study. Am J Respir Crit Care Med 197:933–943

Punjabi NM (2008) The epidemiology of adult obstructive sleep apnea. Proc Am Thorac Soc 5:136–143

Ooms S, Overeem S, Besse K, Rikkert MO, Verbeek M, Claassen JA (2014) Effect of 1 night of total sleep deprivation on cerebrospinal fluid β-amyloid 42 in healthy middle-aged men: a randomized clinical trial. JAMA Neurol 71:971–977

Lucey BP, Hicks TJ, McLeland JS, Toedebusch CD, Boyd J, Elbert DL, Patterson BW, Baty J, Morris JC, Ovod V, Mawuenyega KG, Bateman RJ (2018) Effect of sleep on overnight cerebrospinal fluid amyloid β kinetics. Ann Neurol 83:197–204

Benedict C, Blennow K, Zetterberg H, Cedernaes J (2020) Effects of acute sleep loss on diurnal plasma dynamics of CNS health biomarkers in young men. NEUROLOGY 94:e1181–e1189

Wei M, Zhao B, Huo K, Deng Y, Shang S, Liu J, Li Y, Ma L, Jiang Y, Dang L, Chen C, Wei S, Zhang J, Yang H, Gao F, Qu Q (2017) Sleep Deprivation Induced plasma Amyloid-β transport disturbance in healthy young adults. J Alzheimers Dis 57:899–906

Ju Y-ES, Ooms SJ, Sutphen C, Macauley SL, Zangrilli MA, Jerome G, Fagan AM, Mignot E, Zempel JM, Claassen JAHR, Holtzman DM (2017) Slow wave sleep disruption increases cerebrospinal fluid amyloid-β levels. Brain 140:2104–2111

Qiu H, Zhong R, Liu H, Zhang F, Li S, Le W (2016) Chronic sleep deprivation exacerbates Learning-Memory disability and Alzheimer’s Disease-Like pathologies in AβPP(swe)/PS1(∆E9) mice. J Alzheimers Dis 50:669–685

Zhao HY, Wu HJ, He JL, Zhuang JH, Liu ZY, Huang LQ, Zhao ZX (2017) Chronic sleep restriction induces cognitive deficits and cortical Beta-amyloid deposition in mice via BACE1-Antisense activation. CNS Neurosci Ther 23:233–240

Zhao B, Liu P, Wei M, Li Y, Liu J, Ma L, Shang S, Jiang Y, Huo K, Wang J, Qu Q (2019) Chronic sleep restriction induces Aβ Accumulation by disrupting the balance of Aβ production and clearance in rats. Neurochem Res 44:859–873

Xie Y, Ba L, Wang M, Deng SY, Chen SM, Huang LF, Zhang M, Wang W, Ding FF (2020) Chronic sleep fragmentation shares similar pathogenesis with neurodegenerative Diseases: Endosome-autophagosome-lysosome pathway dysfunction and microglia-mediated neuroinflammation. CNS Neurosci Ther 26:215–227

Vasciaveo V, Iadarola A, Casile A, Dante D, Morello G, Minotta L, Tamagno E, Cicolin A, Guglielmotto M (2023) Sleep fragmentation affects glymphatic system through the different expression of AQP4 in wild type and 5xFAD mouse models. Acta Neuropathol Commun 11:16

Nick H, Fenik P, Zhu Y, Veasey S (2022) Hypocretin/orexin influences chronic sleep disruption injury in the hippocampus. Front Aging Neurosci 14:1025402

Kent BA, Feldman HH, Nygaard HB (2021) Sleep and its regulation: an emerging pathogenic and treatment frontier in Alzheimer’s Disease. Prog Neurobiol 197:101902

Tabuchi M, Lone SR, Liu S, Liu Q, Zhang J, Spira AP, Wu MN (2015) Sleep interacts with aβ to modulate intrinsic neuronal excitability. Curr Biol 25:702–712

Ba L, Huang L, He Z, Deng S, Xie Y, Zhang M, Jacob C, Antonecchia E, Liu Y, Xiao W, Xie Q, Huang Z, Yi C, D’Ascenzo N, Ding F (2021) Does chronic sleep fragmentation lead to Alzheimer’s Disease in Young Wild-Type mice? Front Aging Neurosci 13:759983

Owen JE, Zhu Y, Fenik P, Zhan G, Bell P, Liu C, Veasey S (2021) Late-in-life neurodegeneration after chronic sleep loss in young adult mice. Sleep 44

Zhang F, Niu L, Zhong R, Li S, Le W (2023) Chronic sleep disturbances alters Sleep structure and tau phosphorylation in AβPP/PS1 AD mice and their wild-type littermates. J Alzheimers Dis 92:1341–1355

Zhu Y, Zhan G, Fenik P, Brandes M, Bell P, Francois N, Shulman K, Veasey S (2018) Chronic sleep disruption advances the temporal progression of Tauopathy in P301S Mutant mice. J Neurosci 38:10255–10270

Niu L, Zhang F, Xu X, Yang Y, Li S, Liu H, Le W (2022) Chronic sleep deprivation altered the expression of circadian clock genes and aggravated Alzheimer’s Disease neuropathology. Brain Pathol 32:e13028

Lucey BP (2020) It’s complicated: the relationship between sleep and Alzheimer’s Disease in humans. Neurobiol Dis 144:105031

Sato C, Barthélemy NR, Mawuenyega KG, Patterson BW, Gordon BA, Jockel-Balsarotti J, Sullivan M, Crisp MJ, Kasten T, Kirmess KM, Kanaan NM, Yarasheski KE, Baker-Nigh A, Benzinger TLS, Miller TM, Karch CM, Bateman RJ (2018) Tau kinetics in neurons and the Human Central Nervous System. Neuron 98:861–864

Yanamandra K, Patel TK, Jiang H, Schindler S, Ulrich JD, Boxer AL, Miller BL, Kerwin DR, Gallardo G, Stewart F, Finn MB, Cairns NJ, Verghese PB, Fogelman I, West T, Braunstein J, Robinson G, Keyser J, Roh J, Knapik SS, Hu Y, Holtzman DM (2017) Anti-tau antibody administration increases plasma tau in transgenic mice and patients with tauopathy. Sci Transl Med 9

Yamada K, Holth JK, Liao F, Stewart FR, Mahan TE, Jiang H, Cirrito JR, Patel TK, Hochgräfe K, Mandelkow EM, Holtzman DM (2014) Neuronal activity regulates extracellular tau in vivo. J Exp Med 211:387–393

Holth JK, Fritschi SK, Wang C, Pedersen NP, Cirrito JR, Mahan TE, Finn MB, Manis M, Geerling JC, Fuller PM, Lucey BP, Holtzman DM (2019) The sleep-wake cycle regulates brain interstitial fluid tau in mice and CSF tau in humans. Sci (New York NY) 363:880–884

Article   CAS   ADS   Google Scholar  

Irwin MR, Vitiello MV (2019) Implications of sleep disturbance and inflammation for Alzheimer’s Disease Dementia. Lancet Neurol 18:296–306

Wu R, Tripathy S, Menon V, Yu L, Buchman AS, Bennett DA, De Jager PL, Lim ASP (2022) Fragmentation of rest periods, astrocyte activation, and cognitive decline in older adults with and without Alzheimer’s disease. Alzheimers Dement

Kaneshwaran K, Olah M, Tasaki S, Yu L, Bradshaw EM, Schneider JA, Buchman AS, Bennett DA, De Jager PL, Lim ASP (2019) Sleep fragmentation, microglial aging, and cognitive impairment in adults with and without Alzheimer’s Dementia. Sci Adv 5:eaax7331

Bellesi M, de Vivo L, Chini M, Gilli F, Tononi G, Cirelli C (2017) Sleep loss promotes astrocytic phagocytosis and microglial activation in mouse cerebral cortex. J Neurosci 37:5263–5273

Kincheski GC, Valentim IS, Clarke JR, Cozachenco D, Castelo-Branco MTL, Ramos-Lobo AM, Rumjanek V, Donato J Jr., De Felice FG, Ferreira ST (2017) Chronic sleep restriction promotes Brain Inflammation and synapse loss, and potentiates memory impairment induced by amyloid-β oligomers in mice. Brain Behav Immun 64:140–151

Liu P, Zhao B, Wei M, Li Y, Liu J, Ma L, Shang S, Huo K, Wang J, Li R, Qu Q (2020) Activation of inflammation is Associated with Amyloid-β Accumulation Induced by Chronic Sleep restriction in rats. J Alzheimers Dis 74:759–773

Berk M, Williams LJ, Jacka FN, O’Neil A, Pasco JA, Moylan S, Allen NB, Stuart AL, Hayley AC, Byrne ML, Maes M (2013) So depression is an inflammatory Disease, but where does the inflammation come from? BMC Med 11:200

Siegel JM (2005) Clues to the functions of mammalian sleep. Nature 437:1264–1271

Wulff K, Gatti S, Wettstein JG, Foster RG (2010) Sleep and circadian rhythm disruption in psychiatric and neurodegenerative Disease. Nat Rev Neurosci 11:589–599

Mendoza J (2019) Circadian insights into the biology of depression: symptoms, treatments and animal models. Behav Brain Res 376:112186

Chellappa SL, Aeschbach D (2022) Sleep and anxiety: from mechanisms to interventions. Sleep Med Rev 61:101583

Mirchandaney R, Barete R, Asarnow LD (2022) Moderators of cognitive behavioral treatment for Insomnia on Depression and anxiety outcomes. Curr Psychiatry Rep 24:121–128

Soehner AM, Kaplan KA, Harvey AG (2014) Prevalence and clinical correlates of co-occurring insomnia and hypersomnia symptoms in depression. J Affect Disord 167:93–97

Zhou F, Li S, Xu H (2022) Insomnia, sleep duration, and risk of anxiety: a two-sample mendelian randomization study. J Psychiatr Res 155:219–225

Bean CAL, Ciesla JA (2021) Naturalistic partial sleep deprivation leads to Greater Next-Day anxiety: the moderating role of baseline anxiety and depression. Behav Ther 52:861–873

Palmer CA, Alfano CA (2020) Anxiety modifies the emotional effects of sleep loss. Curr Opin Psychol 34:100–104

Zhou L, Zhang H, Luo Z, Liu X, Yang L, Hu H, Wang Y, Wang C, Wang F (2020) Abnormal night sleep duration and inappropriate sleep initiation time are associated with elevated anxiety symptoms in Chinese rural adults: the Henan Rural Cohort. Psychiatry Res 291:113232

Riemann D (2007) Insomnia and comorbid psychiatric disorders. Sleep Med 8(Suppl 4):S15–20

Lopez R, Barateau L, Evangelista E, Dauvilliers Y (2017) Depression and Hypersomnia: a Complex Association. Sleep Med Clin 12:395–405

Luik AI, Zuurbier LA, Direk N, Hofman A, Van Someren EJ, Tiemeier H (2015) 24-HOUR ACTIVITY RHYTHM AND SLEEP DISTURBANCES IN DEPRESSION AND ANXIETY: A POPULATION-BASED STUDY OF MIDDLE-AGED AND OLDER PERSONS. Depress Anxiety 32:684–692

Bouwmans ME, Conradi HJ, Bos EH, Oldehinkel AJ, de Jonge P (2017) Bidirectionality between sleep symptoms and core depressive symptoms and their long-term course in Major Depression. Psychosom Med 79:336–344

Casement MD, Keenan KE, Hipwell AE, Guyer AE, Forbes EE (2016) Neural reward Processing mediates the relationship between insomnia symptoms and Depression in Adolescence. Sleep 39:439–447

Coles ME, Schubert JR, Nota JA (2015) Sleep, circadian rhythms, and anxious traits. Curr Psychiatry Rep 17:73

Horváth A, Montana X, Lanquart JP, Hubain P, Szűcs A, Linkowski P, Loas G (2016) Effects of state and trait anxiety on sleep structure: a polysomnographic study in 1083 subjects. Psychiatry Res 244:279–283

Short MA, Louca M (2015) Sleep deprivation leads to mood deficits in healthy adolescents. Sleep Med 16:987–993

Taraku B, Zavaliangos-Petropulu A, Loureiro JR, Al-Sharif NB, Kubicki A, Joshi SH, Woods RP, Espinoza R, Narr KL, Sahib AK (2023) White matter microstructural perturbations after total sleep deprivation in depression. Front Psychiatry 14:1195763

Hu B, Liu C, Mou T, Luo F, Lv T, Qian C, Zhang J, Ye M, Liu Z (2021) Meta-analysis of Sleep Deprivation effects on patients with Depression. Front Psychiatry 12:783091

Dong L, Xie Y, Zou X (2022) Association between sleep duration and depression in US adults: a cross-sectional study. J Affect Disord 296:183–188

Tartar JL, Ward CP, Cordeira JW, Legare SL, Blanchette AJ, McCarley RW, Strecker RE (2009) Experimental sleep fragmentation and sleep deprivation in rats increases exploration in an open field test of anxiety while increasing plasma corticosterone levels. Behav Brain Res 197:450–453

Pires GN, Bezerra AG, Tufik S, Andersen ML (2016) Effects of experimental sleep deprivation on anxiety-like behavior in animal research: systematic review and meta-analysis. Neurosci Biobehav Rev 68:575–589

Finan PH, Quartana PJ, Smith MT (2015) The effects of Sleep Continuity disruption on positive Mood and Sleep Architecture in healthy adults. Sleep 38:1735–1742

Meerlo P, Mistlberger RE, Jacobs BL, Heller HC, McGinty D (2009) New neurons in the adult brain: the role of sleep and consequences of sleep loss. Sleep Med Rev 13:187–194

Clasadonte J, McIver SR, Schmitt LI, Halassa MM, Haydon PG (2014) Chronic sleep restriction disrupts sleep homeostasis and behavioral sensitivity to alcohol by reducing the extracellular accumulation of adenosine. J Neurosci 34:1879–1891

Arvin P, Ghafouri S, Bavarsad K, Hajipour S, Khoshnam SE, Sarkaki A, Farbood Y (2023) Administration of growth hormone ameliorates adverse effects of total sleep deprivation. Metab Brain Dis 38:1671–1681

Klumpers UM, Veltman DJ, van Tol MJ, Kloet RW, Boellaard R, Lammertsma AA, Hoogendijk WJ (2015) Neurophysiological effects of sleep deprivation in healthy adults, a pilot study. PLoS ONE 10:e0116906

Jordan W, Tumani H, Cohrs S, Eggert S, Rodenbeck A, Brunner E, Rüther E, Hajak G (2004) Prostaglandin D synthase (beta-trace) in healthy human sleep. Sleep 27:867–874

Zhang Y, Dai C, Shao Y, Peng J, Yang Y, Hou Y (2021) Decreased functional connectivity in the reward network and its relationship with negative emotional experience after total sleep deprivation. Front Neurol 12:641810

Bosch OG, Rihm JS, Scheidegger M, Landolt HP, Stämpfli P, Brakowski J, Esposito F, Rasch B, Seifritz E (2013) Sleep deprivation increases dorsal nexus connectivity to the dorsolateral prefrontal cortex in humans. Proc Natl Acad Sci U S A 110:19597–19602

Lopez-Rodriguez F, Wilson CL, Maidment NT, Poland RE, Engel J (2003) Total sleep deprivation increases extracellular serotonin in the rat hippocampus. Neuroscience 121:523–530

Wu JC, Gillin JC, Buchsbaum MS, Schachat C, Darnall LA, Keator DB, Fallon JH, Bunney WE (2008) Sleep deprivation PET correlations of Hamilton symptom improvement ratings with changes in relative glucose metabolism in patients with depression. J Affect Disord 107:181–186

Grassi Zucconi G, Cipriani S, Balgkouranidou I, Scattoni R (2006) One night’ sleep deprivation stimulates hippocampal neurogenesis. Brain Res Bull 69:375–381

Tartar JL, Ward CP, McKenna JT, Thakkar M, Arrigoni E, McCarley RW, Brown RE, Strecker RE (2006) Hippocampal synaptic plasticity and spatial learning are impaired in a rat model of sleep fragmentation. Eur J Neurosci 23:2739–2748

Grubac Z, Sutulovic N, Ademovic A, Velimirovic M, Rasic-Markovic A, Macut D, Petronijevic N, Stanojlovic O, Hrncic D (2019) Short-term sleep fragmentation enhances anxiety-related behavior: the role of hormonal alterations. PLoS ONE 14:e0218920

Nair D, Zhang SX, Ramesh V, Hakim F, Kaushal N, Wang Y, Gozal D (2011) Sleep fragmentation induces cognitive deficits via nicotinamide adenine dinucleotide phosphate oxidase-dependent pathways in mouse. Am J Respir Crit Care Med 184:1305–1312

Stepnowsky CJ, Palau JJ, Zamora T, Ancoli-Israel S, Loredo JS (2011) Fatigue in sleep apnea: the role of depressive symptoms and self-reported sleep quality. Sleep Med 12:832–837

Gong L, Liao T, Liu D, Luo Q, Xu R, Huang Q, Zhang B, Feng F, Zhang C (2019) Amygdala Changes in Chronic Insomnia and Their Association with sleep and anxiety symptoms: insight from shape analysis. Neural Plast 2019:8549237

Gong L, Shi M, Wang J, Xu R, Yu S, Liu D, Ding X, Zhang B, Zhang X, Xi C (2021) The abnormal functional connectivity in the Locus Coeruleus-Norepinephrine System Associated with anxiety Symptom in Chronic Insomnia Disorder. Front Neurosci 15:678465

Xu M, Wang Q, Li B, Qian S, Wang S, Wang Y, Chen C, Liu Z, Ji Y, Liu K, Xin K, Niu Y (2023) Cerebellum and hippocampus abnormalities in patients with insomnia comorbid depression: a study on cerebral blood perfusion and functional connectivity. Front Neurosci 17:1202514

Motomura Y, Katsunuma R, Ayabe N, Oba K, Terasawa Y, Kitamura S, Moriguchi Y, Hida A, Kamei Y, Mishima K (2021) Decreased activity in the reward network of chronic insomnia patients. Sci Rep 11:3600

Gong L, Yu S, Xu R, Liu D, Dai X, Wang Z, Hu Y, Yan B, Kui Y, Zhang B, Feng F (2021) The abnormal reward network associated with insomnia severity and depression in chronic insomnia disorder. Brain Imaging Behav 15:1033–1042

Rahmani M, Rahmani F, Rezaei N (2020) The brain-derived neurotrophic factor: missing Link between Sleep Deprivation, Insomnia, and Depression. Neurochem Res 45:221–231

Schmitt K, Holsboer-Trachsler E, Eckert A (2016) BDNF in sleep, insomnia, and sleep deprivation. Ann Med 48:42–51

Konakanchi S, Raavi V, Ml HK, Shankar Ms V (2022) Effect of chronic sleep deprivation and sleep recovery on hippocampal CA3 neurons, spatial memory and anxiety-like behavior in rats. Neurobiol Learn Mem 187:107559

Xia M, Li Z, Li S, Liang S, Li X, Chen B, Zhang M, Dong C, Verkhratsky A, Guan D, Li B (2020) Sleep Deprivation Selectively Down-Regulates Astrocytic 5-HT(2B) Receptors and Triggers Depressive-Like Behaviors via Stimulating P2X(7) Receptors in Mice. Neurosci Bull 36:1259–1270

Manchanda S, Singh H, Kaur T, Kaur G (2018) Low-grade neuroinflammation due to chronic sleep deprivation results in anxiety and learning and memory impairments. Mol Cell Biochem 449:63–72

Kiecolt-Glaser JK, Derry HM, Fagundes CP (2015) Inflammation: depression fans the flames and feasts on the heat. Am J Psychiatry 172:1075–1091

Liu H, Huang X, Li Y, Xi K, Han Y, Mao H, Ren K, Wang W, Wu Z (2022) TNF signaling pathway-mediated microglial activation in the PFC underlies acute paradoxical sleep deprivation-induced anxiety-like behaviors in mice. Brain Behav Immun 100:254–266

Merikanto I, Partonen T (2021) Eveningness increases risks for depressive and anxiety symptoms and hospital treatments mediated by insufficient sleep in a population-based study of 18,039 adults. Depress Anxiety 38:1066–1077

Stahl ST, Smagula SF, Rodakowski J, Dew MA, Karp JF, Albert SM, Butters M, Gildengers A, Reynolds CF 3rd (2021) Subjective Sleep Quality and trajectories of Interleukin-6 in older adults. Am J Geriatr Psychiatry 29:204–208

Selvi Y, Boysan M, Kandeger A, Uygur OF, Sayin AA, Akbaba N, Koc B (2018) Heterogeneity of sleep quality in relation to circadian preferences and depressive symptomatology among major depressive patients. J Affect Disord 235:242–249

Taylor BJ, Hasler BP (2018) Chronotype and Mental Health: recent advances. Curr Psychiatry Rep 20:59

Merikanto I, Kronholm E, Peltonen M, Laatikainen T, Vartiainen E, Partonen T (2015) Circadian preference links to depression in general adult population. J Affect Disord 188:143–148

Schapira AHV, Chaudhuri KR, Jenner P (2017) Non-motor features of Parkinson Disease. Nat Rev Neurosci 18:435–450

Stefani A, Högl B (2020) Sleep in Parkinson’s Disease. NEUROPSYCHOPHARMACOLOGY 45:121–128

Morawska MM, Moreira CG, Ginde VR, Valko PO, Weiss T, Büchele F, Imbach LL, Masneuf S, Kollarik S, Prymaczok N, Gerez JA, Riek R, Baumann CR, Noain D (2021) Slow-wave sleep affects synucleinopathy and regulates proteostatic processes in mouse models of Parkinson’s Disease. Sci Transl Med 13:eabe7099

Zuzuárregui JRP, During EH (2020) Sleep issues in Parkinson’s Disease and their management. Neurotherapeutics 17:1480–1494

Hermanowicz N, Jones SA, Hauser RA (2019) Impact of non-motor symptoms in Parkinson’s Disease: a PMDAlliance survey. Neuropsychiatr Dis Treat 15:2205–2212

Kay DB, Tanner JJ, Bowers D (2018) Sleep disturbances and depression severity in patients with Parkinson’s Disease. Brain Behav 8:e00967

Latif S, Jahangeer M, Maknoon Razia D, Ashiq M, Ghaffar A, Akram M, El Allam A, Bouyahya A, Garipova L, Ali Shariati M, Thiruvengadam M, Azam Ansari M (2021) Dopamine in Parkinson’s Disease. Clin Chim Acta 522:114–126

Dauvilliers Y, Schenck CH, Postuma RB, Iranzo A, Luppi PH, Plazzi G, Montplaisir J, Boeve B (2018) REM sleep behaviour disorder. Nat Rev Dis Primers 4:19

Miglis MG, Adler CH, Antelmi E, Arnaldi D, Baldelli L, Boeve BF, Cesari M, Dall’Antonia I, Diederich NJ, Doppler K, Dušek P, Ferri R, Gagnon JF, Gan-Or Z, Hermann W, Högl B, Hu MT, Iranzo A, Janzen A, Kuzkina A, Lee JY, Leenders KL, Lewis SJG, Liguori C, Liu J, Lo C, Ehgoetz Martens KA, Nepozitek J, Plazzi G, Provini F, Puligheddu M, Rolinski M, Rusz J, Stefani A, Summers RLS, Yoo D, Zitser J, Oertel WH (2021) Biomarkers of conversion to α-synucleinopathy in isolated rapid-eye-movement sleep behaviour disorder. Lancet Neurol 20:671–684

Si XL, Gu LY, Song Z, Zhou C, Fang Y, Jin CY, Wu JJ, Gao T, Guo T, Guan XJ, Xu XJ, Yin XZ, Yan YP, Zhang MM, Pu JL (2020) Different Perivascular Space burdens in Idiopathic Rapid Eye Movement Sleep Behavior Disorder and Parkinson’s Disease. Front Aging Neurosci 12:580853

Maggi G, Trojano L, Barone P, Santangelo G (2021) Sleep disorders and cognitive dysfunctions in Parkinson’s Disease: a Meta-Analytic Study. Neuropsychol Rev 31:643–682

Lin YQ, Chen SD (2018) RBD: a red flag for cognitive impairment in Parkinson’s Disease? Sleep Med 44:38–44

Lazar AS, Panin F, Goodman AO, Lazic SE, Lazar ZI, Mason SL, Rogers L, Murgatroyd PR, Watson LP, Singh P, Borowsky B, Shneerson JM, Barker RA (2015) Sleep deficits but no metabolic deficits in Premanifest Huntington’s Disease. Ann Neurol 78:630–648

Vas S, Nicol AU, Kalmar L, Miles J, Morton AJ (2021) Abnormal patterns of sleep and EEG power distribution during non-rapid eye movement sleep in the sheep model of Huntington’s Disease. Neurobiol Dis 155:105367

Cabanas M, Pistono C, Puygrenier L, Rakesh D, Jeantet Y, Garret M, Cho YH (2019) Neurophysiological and behavioral effects of anti-orexinergic treatments in a mouse model of Huntington’s Disease. Neurotherapeutics: The Journal of the American Society for Experimental NeuroTherapeutics 16:784–796

Hulisz D (2018) Amyotrophic Lateral Sclerosis: Disease state overview. Am J Manag Care 24:S320–s326

PubMed   Google Scholar  

Liu S, Ren Q, Gong G, Sun Y, Zhao B, Ma X, Zhang N, Zhong S, Lin Y, Wang W, Zheng R, Yu X, Yun Y, Zhang D, Shao K, Lin P, Yuan Y, Dai T, Zhang Y, Li L, Li W, Zhao Y, Shan P, Meng X, Yan C (2022) Hypothalamic subregion abnormalities are related to body mass index in patients with sporadic Amyotrophic Lateral Sclerosis. J Neurol 269:2980–2988

Gnoni V, Zoccolella S, Giugno A, Urso D, Tamburrino L, Filardi M, Logroscino G (2023) Hypothalamus and Amyotrophic Lateral Sclerosis: potential implications in sleep disorders. Front Aging Neurosci 15:1193483

Lo Martire V, Caruso D, Palagini L, Zoccoli G, Bastianini S (2020) Stress & sleep: a relationship lasting a lifetime. Neurosci Biobehav Rev 117:65–77

Basta M, Vgontzas AN, Fernandez-Mendoza J, Antypa D, Li Y, Zaganas I, Panagiotakis S, Karagkouni E, Simos P (2022) Basal cortisol levels are increased in patients with mild cognitive impairment: role of Insomnia and short sleep duration. J Alzheimers Dis 87:933–944

Gabery S, Ahmed RM, Caga J, Kiernan MC, Halliday GM, Petersén Å (2021) Loss of the metabolism and sleep regulating neuronal populations expressing orexin and oxytocin in the hypothalamus in Amyotrophic Lateral Sclerosis. Neuropathol Appl Neurobiol 47:979–989

Boentert M (2019) Sleep disturbances in patients with Amyotrophic Lateral Sclerosis: current perspectives. Nat Sci Sleep 11:97–111

Download references

Acknowledgements

We would like to thank Editage ( www.editage.cn ) for English language editing.

This study was supported by grants from the National Natural Science Foundation of China (No. 82371541), the project for the improvement of research skill in Anhui Medical University (No.2021xkjT003), Talent Training Program from the School of Basic Medical Sciences of Anhui Medical University (No. 2022YPJH201), Research Fund of Anhui Institute of translational medicine (No. 2022zhyx-C11) and National Training Project of Innovation and Entrepreneurship for Undergraduates of China (No. 202310366060 & No. 202310366037).

Author information

Xiao Xiao and Yimin Rui contributed equally to this work.

Authors and Affiliations

School of Mental Health and Psychological Sciences, Anhui Medical University, Hefei, Anhui, China

Xiao Xiao & Yimin Rui

Department of Pharmacology, School of Basic Medical Sciences, Anhui Medical University, Hefei, Anhui, China

Xiao Xiao, Yimin Rui, Yu Jin & Ming Chen

You can also search for this author in PubMed   Google Scholar

Contributions

All authors contributed to the manuscript. X.X. had the idea for the article. X.X., Y.M.R. and Y. J. performed the literature search, drafted the manuscript and created the figures. M.C. revised critically the manuscript. All authors read and approved the manuscript.

Corresponding author

Correspondence to Ming Chen .

Ethics declarations

Competing interests.

The authors have no relevant financial or non-financial interests to disclose.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Xiao, X., Rui, Y., Jin, Y. et al. Relationship of Sleep Disorder with Neurodegenerative and Psychiatric Diseases: An Updated Review. Neurochem Res 49 , 568–582 (2024). https://doi.org/10.1007/s11064-023-04086-5

Download citation

Received : 22 September 2023

Revised : 08 December 2023

Accepted : 09 December 2023

Published : 18 December 2023

Issue Date : March 2024

DOI : https://doi.org/10.1007/s11064-023-04086-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Sleep Disorder
• Alzheimer’s Disease
• Parkinson’s Disease
• Find a journal
• Publish with us
• Track your research

Skip to content

Explore CUIMC

Diversity, equity, and inclusion.

At CUIMC, we are committed to continuous improvement in providing culturally inclusive medical education and clinical care.

Columbia Vagelos College of Physicians and Surgeons is dedicated to developing the next generation of leaders in medicine

Patient Care

• Find a Doctor

Search for a provider by specialty, expertise, location and insurance. Schedule an appointment online.

Read the latest news stories about CUIMC faculty, research, and events

Newly Found Genetic Variant Defends Against Alzheimer’s Disease

Discovery could launch new types of drugs to prevent, treat the disease, share this page.

• Share on Facebook
• Share on X (formerly Twitter)
• Share on LinkedIn
• Share by email

Columbia researchers have discovered a genetic variant that reduces the odds of developing Alzheimer’s disease by up to 70% and may be protecting thousands of people in the United States from the disease. 

The discovery of the protective variant, which appears to allow toxic forms of amyloid out of the brain and through the blood-brain barrier, supports emerging evidence that the brain’s blood vessels play a large role in Alzheimer's disease and could herald a new direction in therapeutic development. 

“Alzheimer’s disease may get started with amyloid deposits in the brain, but the disease manifestations are the result of changes that happen after the deposits appear,” says Caghan Kizil, PhD , a co-leader of the study that identified the variant and associate professor of neurological sciences (in neurology and in the Taub Institute) at Columbia University Vagelos College of Physicians and Surgeons.  

“Our findings suggest that some of these changes occur in the brain’s vasculature and that we may be able to develop new types of therapies that mimic the gene’s protective effect to prevent or treat the disease.” 

An attractive drug target?

The protective variant identified by the study occurs in a gene that makes fibronectin, a component of the blood-brain barrier, a lining surrounding the brain’s blood vessels that controls the movement of substances in and out of the brain. 

Fibronectin is usually present in the blood-brain barrier in very minute amounts, but it is increased in large amounts in people with Alzheimer’s disease. The variant identified in the fibronectin gene seems to protect against Alzheimer’s disease by preventing the buildup of excess fibronectin at the blood-brain barrier. 

"It’s a classic case of too much of a good thing,” Kizil says. “It made us think that excess fibronectin could be preventing the clearance of amyloid deposits from the brain.” 

The researchers confirmed that hypothesis in a zebrafish model of Alzheimer’s disease and have additional studies in mice underway. They also found that reducing fibronectin in the animals increased amyloid clearance and improved other damage caused by Alzheimer’s disease. 

“These results gave us the idea that a therapy targeting fibronectin and mimicking the protective variant could provide a strong defense against the disease in people,” says study co-leader Richard Mayeux, MD , chair of neurology and the Gertrude H. Sergievsky Professor of Neurology, Psychiatry, and Epidemiology. 

The newest treatments for Alzheimer’s disease target the amyloid deposits directly and are very efficient at removing the deposits via the immune system. However, simply removing the deposits this way doesn’t improve symptoms or repair other damage.  

“We may need to start clearing amyloid much earlier and we think that can be done through the bloodstream,” Mayeux adds. “That's why we are excited about the discovery of this variant in fibronectin, which may be a good target for drug development.” 

Protective gene was found in people resilient to Alzheimer’s disease

The researchers discovered the protective variant in people who never developed symptoms but who had inherited the e4 form of the APOE gene, which significantly increases the risk of developing Alzheimer’s disease. 

“These resilient people can tell us a lot about the disease and what genetic and non-genetic factors might provide protection,” says study co-leader Badri N. Vardarajan, PhD , assistant professor of neurological science (in neurology, the Gertrude H. Sergievsky Center, and the Taub Institute), who is an expert in using computational approaches to discover Alzheimer’s disease genes. 

"We hypothesized that these resilient people may have genetic variants that protect them from APOEe4.” 

To find protective mutations, the Columbia researchers sequenced the genomes of several hundred APOEe4 carriers over age 70 of various ethnic backgrounds, including those with and without Alzheimer's disease. Many participants were residents of Northern Manhattan who were enrolled in the Washington Heights/Inwood Columbia Aging Project, an ongoing study that has been conducted by Columbia University’s Department of Neurology for more than 30 years.  

The study identified the fibronectin variant, and the Columbia team publicized their results in a preprint for other researchers to view. Based on the Columbia team’s observations, another group from Stanford and Washington universities replicated the study in an independent cohort of APOEe4 carriers, mostly of European origin.  

“They found the same fibronectin variant, which confirmed our finding and gave us even more confidence in our result,” Vardarajan says.  

The two groups combined the data on their 11,000 participants, which allowed them to calculate that the mutation reduces the odds of developing Alzheimer’s in APOE4 carriers by 71% and forestalls the disease by roughly four years in those who eventually develop the disease. 

The researchers estimate that 1% to 3% of APOEe4 carriers in the United States—roughly 200,000 to 620,000 people—may also carry the protective fibronectin mutation. 

Wide therapeutic potential

The fibronectin variant, though discovered in APOEe4 carriers, could protect against Alzheimer’s disease in people with other forms of APOE. 

“There’s a significant difference in fibronectin levels in the blood-brain barrier between cognitively healthy individuals and those with Alzheimer's disease, independent of their APOEe4 status,” Kizil says.  

“Anything that reduces excess fibronectin should provide some protection, and a drug that does this could be a significant step forward in the fight against this debilitating condition.” 

More information  

The study, “ Rare genetic variation in Fibronectin 1 (FN1) protects against APOEε4 in Alzheimer’s disease ,” was published online April 10 in the journal Acta Neuropathologica.  

The research and researchers were supported by the Carol and Gene Ludwig Family Foundation, the Agouron Institute, the U.S. National Institutes of Health (grants R01AG067501, RF1AG066107, R01AG072474, R35GM148348, R01AG061796, U19AG074879, U01AG066752, U01AG046139, R01AG060747, P50AG047366, and R00AG075238), a Schaefer Research Scholars Award, Taub Institute Grant for Emerging Research (TIGER), the Thompson Family Foundation Program for Accelerated Medicine Exploration in Alzheimer’s Disease and Related Disorders of the Nervous System, and an Alzheimer’s Association Zenith Fellows Award.   

Richard Mayeux, MD, is also director of the Gertrude H. Sergievsky Center and co-director of the Taub Institute for Research on Alzheimer's Disease and the Aging Brain at Columbia University Vagelos College of Physicians and Surgeons and neurologist-in-chief at NewYork-Presbyterian/Columbia University Irving Medical Center.  

Other contributors and funding sources are listed in the paper.  

• U.S. Department of Health & Human Services

• Virtual Tour
• Staff Directory
• En Español

You are here

News releases.

Media Advisory

Thursday, April 11, 2024

Study reveals no causal link between neurodevelopmental disorders and acetaminophen exposure before birth

NIH-funded research in siblings finds previously reported connection is likely due to other underlying factors.

Acetaminophen exposure during pregnancy is not linked to the risk of developing autism, ADHD, or intellectual disability, according to a new study of data from more than 2 million children in Sweden. The collaborative research effort by Swedish and American investigators, which appears in JAMA , is the largest of its kind and was funded by the National Institute of Neurological Disorders and Stroke (NINDS), part of the National Institutes of Health (NIH).

Scientists compared siblings — who share genetics and other variables such as parental health, environmental exposures, and socioeconomic factors — and were able to limit the influence of other potential risk factors. This allowed them to focus specifically on, and eliminate, the risk associated with acetaminophen. The study design was unique due to the size of the population captured in the Swedish Medical Birth Register and the Swedish Prescribed Drug Register. Before siblings were considered, there appeared to be a small increase in risk for neurodevelopmental disorders in children exposed to acetaminophen, which was noted in previous studies.

Acetaminophen is commonly used as a pain reliever and fever reducer and is found in a variety of medicines available over the counter and via prescription. It is often taken during pregnancy instead of nonsteroidal anti-inflammatory drugs, known as NSAIDs, which can cause low levels of amniotic fluid, according to the Food and Drug Administration . The reasons pregnant people might take acetaminophen, including fever, or conditions such as chronic migraine, could be, and in some cases are, associated with an increased risk for later neurodevelopmental disorders following pregnancy. 

One limitation of this study is that it relies on data from prescribed acetaminophen and from self-reporting from pregnant people during prenatal care. It may not capture all use or dosage in all people, particularly over-the-counter medicines. However, the number of patients included in the study sample and the ability to control for many other confounding factors support the conclusion that acetaminophen is not directly linked to an increase link of autism, ADHD, or intellectual disability.  

To inform best preventative strategies, additional research is required to fully understand the genetic and non-genetic factors that increase the risk of autism, ADHD, and intellectual disability.

This study was supported by NINDS (NS107607).

Vicky Whittemore, Ph.D., program director, NINDS

Ahlqvist VH, et al. “Acetaminophen use during pregnancy and children’s risk of autism, ADHD, and intellectual disability.” JAMA. April 9, 2024. DOI: 10.1001/jama.2024.3172

NINDS  is the nation’s leading funder of research on the brain and nervous system. The mission of NINDS is to seek fundamental knowledge about the brain and nervous system and to use that knowledge to reduce the burden of neurological disease.

About the National Institutes of Health (NIH): NIH, the nation's medical research agency, includes 27 Institutes and Centers and is a component of the U.S. Department of Health and Human Services. NIH is the primary federal agency conducting and supporting basic, clinical, and translational medical research, and is investigating the causes, treatments, and cures for both common and rare diseases. For more information about NIH and its programs, visit www.nih.gov .

NIH…Turning Discovery Into Health ®

Connect with Us

• More Social Media from NIH

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Diagnostics (Basel)

Modern Methods of Diagnostics and Treatment of Neurodegenerative Diseases and Depression

Natalia shusharina.

1 Baltic Center for Neurotechnologies and Artificial Intelligence, Immanuel Kant Baltic Federal University, 236041 Kaliningrad, Russia

Denis Yukhnenko

2 Department of Social Security and Humanitarian Technologies, N. I. Lobachevsky State University of Nizhny Novgorod, 603022 Nizhny Novgorod, Russia

Stepan Botman

Viktor sapunov, vladimir savinov, gleb kamyshov, dmitry sayapin, igor voznyuk.

3 Department of Neurology, Pavlov First Saint Petersburg State Medical University, 197022 Saint Petersburg, Russia

Associated Data

No new data were created or analyzed in this study. Data sharing is not applicable to this article.

This paper discusses the promising areas of research into machine learning applications for the prevention and correction of neurodegenerative and depressive disorders. These two groups of disorders are among the leading causes of decline in the quality of life in the world when estimated using disability-adjusted years. Despite decades of research, the development of new approaches for the assessment (especially pre-clinical) and correction of neurodegenerative diseases and depressive disorders remains among the priority areas of research in neurophysiology, psychology, genetics, and interdisciplinary medicine. Contemporary machine learning technologies and medical data infrastructure create new research opportunities. However, reaching a consensus on the application of new machine learning methods and their integration with the existing standards of care and assessment is still a challenge to overcome before the innovations could be widely introduced to clinics. The research on the development of clinical predictions and classification algorithms contributes towards creating a unified approach to the use of growing clinical data. This unified approach should integrate the requirements of medical professionals, researchers, and governmental regulators. In the current paper, the current state of research into neurodegenerative and depressive disorders is presented.

1. Neurodegenerative Disorders

Neurodegenerative disorders are a group of slowly progressing irreversible conditions characterized by neuronal death and subsequent atrophy of certain areas of the brain. These disorders predominantly manifest in old age and include Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, Pick’s disease, amyotrophic lateral sclerosis, dementia with Lewy bodies, and other similar other associated conditions [ 1 ]. All of them are characterized by a decrease in cognitive abilities, severe motor impairment, impaired social functioning, and significant difficulties in one’s day-to-day tasks. Currently, neurodegenerative disorders are not curable, and their treatments are aimed at slowing down the disease’s progression, improving the quality of life of patients, and addressing comorbid disorders. The number of people with known neurodegenerative disorders has increased over the past decades. This increase has been predominantly associated with an increase in life expectancy (so, individuals are more likely to develop a disorder during their lifetime) and an increase in the effectiveness of early diagnostics (more cases became known to healthcare systems) [ 2 ]. According to recent data, up to 30% of all people over 85 suffer from Alzheimer’s disease, and 5% of people over 65 suffer from Parkinson’s disease [ 3 ]. This increase in the number of patients as well as the need to utilize a comprehensive approach necessitates the high standards for medical data systems and machine learning algorithms deployed on these systems [ 4 ].

1.1. The Assessment of Neurodegenerative Disorders

The effectiveness of a given therapy in slowing down the progression of neurodegenerative disorders depends on their timely detection. Early detection through screening will allow initiation of treatment before the onset of severe clinical symptoms and significantly delay their onset. In recent decades, pre-clinical assessment has become a major focus of research into neurodegenerative disorders [ 5 , 6 ]. The assessment might take into account behavioral symptoms, biomarkers from blood and cerebrospinal fluid, neuroimaging data (PET and MRI), conductivity and electrical activity of the brain (TMS and EEG), and the results of neuropsychological tests. Often, differential diagnosis is required for depression and conditions that could cause a potentially reversible cognitive decline [ 7 ]. Thus, the number of instrumentally determined markers and parameters is quite large. A feature of most neurodegenerative disorders is the low specificity of individual assessment methods and markers. This often leads to the inability to establish the correct diagnosis without multifaceted and often costly examinations [ 8 ]. Below, we will describe the main groups of diagnostic methods used both to detect the disease and to monitor its progression.

1.2. Biomarkers in Blood and Cerebrospinal Fluid

The main biomarkers used in the differential diagnostics of neurodegenerative disorders, in particular Alzheimer’s disease, are the levels of beta-amyloids and tau proteins in the cerebrospinal fluid of patients [ 9 ]. Using the ratio of the amount of beta-amyloid 1-42 to the total tau index makes it possible to identify patients with neurodegenerative disorders with high accuracy as well as to distinguish patients with Alzheimer’s disease from patients with other forms of dementia with moderate accuracy [ 10 ]. The level of tau protein phosphorylated at position P217 (P217 tau) in blood plasma also demonstrated the potential to be used in differential diagnostics of neurodegenerative disorders, primarily Alzheimer’s disease [ 11 ]. Tau 217 values outperformed other blood biomarkers, as well as MRI markers, in a study of 1402 patients from 3 independent cohorts that included patients with dementia, a cognitive decline of a different nature, and healthy subjects [ 12 ]. The accuracy of the assessment using blood biomarkers in this study was comparable to the accuracy of diagnosis using PET markers. Other studies have also demonstrated that tRNAs could serve as potential biomarkers to detect various neurodegenerative disorders at an early stage [ 13 ].

1.3. Psychological and Neuropsychological Assessment

Psychological (including experimental pathopsychological assessment as practiced in post-USSR countries) and neuropsychological methods for assessing the cognitive deficit present in neurodegenerative disorders are also used to monitor the progression of the disorder and the effectiveness of treatment. The most sensitive to specific cognitive deficits caused by neurodegenerative disorders are short-term memory assessments (e.g., the Visual Short-Term Memory Binding Test) and executive functions assessments. A meta-analysis of 142 trials of sorting tests evaluating executive functions (Wisconsin Card Sorting Test [WCST] and Delis-Kaplan Executive Function System [DKEFS-ST]) demonstrated their high diagnostic validity in detecting neurodegenerative disorders and vascular dementias [ 14 ]. The results of memory and executive function tests, conducted in isolation or as part of standardized neuropsychological batteries, correlate with pathological changes in the medulla as well as with the patient’s overall level of maladjustment. Their use, therefore, represents a cost-effective and non-invasive way to track the progression of the disease. Neuropsychological testing has also demonstrated the potential for the pre-clinical diagnosis of neurodegenerative disorders, especially Alzheimer’s disease. Composite score from the ADCS-PACC scale, which is derived from several neuropsychological tests of memory and executive functions, showed a correlation with the level of beta-amyloids in the cerebrospinal fluid of patients in two studies conducted in the USA and Australia [ 15 ]. These preliminary research results demonstrate that sufficiently sensitive neuropsychological batteries might be used as a non-invasive method for the early detection of neurodegenerative disorders.

1.4. Neuroimaging

The regularity of the spread of the degenerative process in the brain is one of the main criteria for the differential diagnosis of various neurodegenerative disorders. The formation of tau links, as well as a decrease in the volume of certain areas of the brain, correlates with the prognosis and clinical symptoms. These changes are determined using MRI and PET methods. MRI can detect damage to the white matter and a decrease in the local volume of various parts of the brain [ 16 ]. PET allows for assessing changes in glucose metabolism and the presence of neurofibrillary tangles in the brain [ 17 ]. Machine learning methods are currently actively applied to neuroimaging data to minimize diagnostic errors and automate the process. Currently, MRI and PET methods are quite expensive and their use as methods for pre-clinical diagnosis of neurodegenerative diseases is severely limited.

1.5. Connectivity

Transcranial magnetic stimulation (TMS) demonstrated some potential for differential diagnosis in patients with neurodegenerative disorders of various etiologies. In a blinded study of 273 patients (with 421 controls) with Alzheimer’s disease, Lewy body dementia, and frontotemporal dementia, the assessment of cortical connectivity with TMS was conducted [ 18 ]. The researchers then trained a prediction model in the form of an ensemble of the binary decision three classifiers. The model showed extremely high classification accuracy when discriminating between disorders. This study demonstrates the potential of combining TMS and machine learning methods in the non-invasive differential diagnosis of neurodegenerative disorders.

1.6. Electrical Activity of the Brain

Abnormal patterns of electrophysiological activity of the brain in neurodegenerative disorders result from several factors including disturbed functional connections between cortical zones, axon pathology, cholinergic deficiency, and others [ 19 ]. Typical EEG patterns used in screening have been described in the scientific and practical literature [ 20 ]. However, at the moment, the described EEG patterns have not demonstrated high diagnostic accuracy. Potentially, their diagnostic accuracy can be significantly improved through the application of deep learning methods to the analysis of neurophysiological data. Such studies are already being actively carried out [ 21 ].

Thus, the development of algorithms for pre-clinical non-invasive assessment of neurodegenerative disorders and their differential diagnosis remains a promising area of research. The main challenge is to ensure the generalizability and universality of developed algorithms, so they can by deployed using various EEG devices. The other promising direction of research is the application of machine learning methods to the differential diagnosis between neurodegenerative disorders and mild cognitive impairments of different etiology [ 22 ].

1.7. Application of Machine Learning

The process of establishing a diagnosis varies depending on the data about a patient available to a medical practitioner at a particular moment. The results of psychosocial and behavioral assessments are often presented as a small set of tabular data and/or several composite indicators. In this case, the practitioner makes a decision based often on his subjective clinical experience. The evaluation of clinical tests (biomarkers of blood, cerebrospinal fluid, etc.) is often also carried out by the medical practitioner who is provided with the generally accepted reference values. That is, for each set of parameters, there are specific ranges that are the “norm” for a healthy person and numerical indicators of the level of deviations associated with potential health problems. This form of data presentation is easy to interpret and convenient for the practitioner.

However, when interpreting the results of neuroimaging and the analysis of the electrical activity of the brain, not only the amount of data is increased substantially, but the format of the data presentation changes. Instead of concise tabular values, a medical practitioner sees images of nerve tissues or a graphical representation of the electrical signals of the brain. In this form, the data are hard to interpret by a person directly, and to establish a diagnosis, preliminary decoding based on complex groups of instrumental parameters is necessary. The situation complicates further if the assessments are performed periodically and there is a need to track the progression of the disease. In this case, the complexity of a patient’s data increases significantly along with the probability of a medical error, simply because of the large number of parameters examined. Therefore, although the use of neuroimaging (PET and SPECT in particular) is one of the most effective and accurate methods for detecting neurodegenerative disorders at the pre-clinical stage, these methods are rarely employed as part of routine medical screenings [ 23 ]. The problem of analysis of patient data that are too large in volume and too complex for human perception has a potential solution: automation of processing using mathematical pre-processing and machine learning technologies. Modern algorithms can already analyze large volumes of data and produce summary reports with a list of detected signs of a patient’s health problems. Of course, this method has disadvantages: difficulties in creating a training sample of medical data of sufficient size and quality, “opacity” of algorithm predictions (a medical practitioner cannot know why the machine made this or that decision), technological challenges (deployment of such systems in the infrastructure of a medical institution could be a difficult task), social perception (distrust of automatic “assistants” by medical practitioners and patients), and the presence of controversial ethical aspects regarding the use of such technologies in healthcare [ 24 ].

To date, diagnostic systems based on machine learning are rarely used in clinical practice, since the listed disadvantages outweigh their potential advantages. However, a large body of research and development in this area demonstrates that such advanced systems will play an important role in the delivery of palliative care for patients with neurodegenerative disorders. This statement is supported by a significant increase in the number of relevant publications: almost an order of magnitude over the period 2014–2019. It should be noted that at the same time, the proportion of research devoted to deep learning algorithms has noticeably increased, which is illustrated by Figure 1 [ 4 ].

The number of publications on the applications of machine learning ( right ) and deep learning ( left ) to diagnostics of Parkinson’s disease (yellow graph) and Alzheimer’s disease (purple graph) by year (reproduced from [ 4 ]).

In support of this argument, we will further report on some of the applications of machine learning to the assessment and correction of the disorders under consideration. Assistive technologies and methodological approaches will be discussed separately in the relevant sections below.

The use of machine learning for the analysis of blood and cerebrospinal fluid biomarkers is relatively uncommon, and the algorithms used, in most cases, are quite simple from a mathematical point of view. For example, the method of linear discriminant analysis (LDA) has been used to stratify patients by their blood work [ 25 ]. In total, 377 people participated in the study: individuals without symptoms of neurodegenerative diseases (97 individuals), with signs of Alzheimer’s disease (35 + 41 with mild cognitive impairment), with parkinsonism (57 + 29 with cognitive impairment in addition to the usual damage to motor functions), with dementia due to Parkinson’s disease (87), and with frontotemporal dementia (31). Five blood markers were assessed: the levels of beta-amyloids 42 and 40, the total level of tau protein (t-Tau) phosphorylated at position 181 (p-Tau181), and alpha-synucleins. LDA was used to build 2D and 3D linear models from blood work data, and subsequent classification was performed using a random forest (RF) algorithm. The average accuracy of differentiation by categories (Alzheimer’s disease, parkinsonism, and frontotemporal dementia) was 76%, and by subgroups: 83% in the spectrum of Alzheimer’s disease and 63% in the spectrum of Parkinson’s disease. The researchers in another study used cerebrospinal fluid to differentiate patients with Alzheimer’s disease from healthy controls [ 26 ]. They examined the levels of beta-amyloid 42, t-Tau, and p-Tau. The inflammatory proteins were determined by immunoassay. The authors used the LASSO algorithm (linear model with L1 regularization) for feature selection and logistic regression for classification, which reached 0.91 AUC (quantitative interpretation of receiver operating characteristic curve [ROC]). Another study used an alternative set of indicators from cerebrospinal fluid: the presence of elements in the fluid (As, Fe, Mg, Ni, Se, Sr, etc.). Their presence was determined with optical emission spectrometry systems with inductively coupled plasma and sector mass spectrometry. The data were then processed by a support vector machine (SVM) algorithm with a radial kernel [ 27 ]. As a result, the researchers were able to achieve an AUC of 0.76 for the diagnosis of Parkinson’s disease (binary classification versus healthy people) in a sample of 187 people (82 patients with parkinsonism). However, the out-of-sample prediction performance of the resulting model turned out to be quite unusual. At first, the model showed high accuracy when fitted on local patients and extremely low accuracy in people living in geographically remote areas. This was due to the characteristics of the training sample: SVM was trained on the sample of individuals living in the same area. After the model was re-trained using cross-validation with different territories serving as references, the average accuracy of the algorithm reached AUC 0.78, regardless of where the patients came from. The results of the latter work highlight the importance of applying proper training procedures and the need to use a sufficient data sample, even for relatively straightforward algorithms.

The diversity of changes in brain structures in neurodegenerative disorders, including those linked to genetic factors, makes creating computational models a challenge. The performance of the model can sharply decrease if the preliminary classification of pathological changes used to train additional models was performed erroneously. The possibility of automating pre-classification has been explored in detail by A. Young et al. [ 28 ]. The researchers made an attempt to combine algorithms that identify subtypes and the progression of neurodegenerative disorders into a single ensemble of learning models. In addition to a very detailed pre-classification of images by the type of brain changes, the authors created a model that estimated the stage of disease progression. To do this, they introduced time granulation related to the severity of a particular observed parameter within each pre-classified type of brain change. The model was also supplemented with data on genetically caused symptoms (Genetic Frontotemporal Dementia Initiative [GENFI]) that significantly affect the course of the disease. The mathematical model itself was an ensemble of simple linear z-score-based models that performed clustering tasks. For genetically determined frontotemporal dementia, the SuStaIn model was able not only to identify the genotype based on MRI images alone but also to resolve further heterogeneity within specific genotypes. For Alzheimer’s disease, the model was able to identify three subtypes with unique progressions over time. A more detailed specification of the structure of each subtype of disease according to physiological characteristics (e.g., localization and severity of atrophic changes) could be considered a fairly promising approach and, in addition to the work noted above, is covered in [ 29 ].

Peran and Barbagallo [ 30 ] compared MRI image classification algorithms for distinguishing cases of multiple system atrophy and Parkinson’s syndrome in a mixed dataset (29 patients with MSA and 26 with Parkinson’s syndrome) using multimodal MRI. Comparing the algorithms of discriminant analysis, voxel-based morphometry (supervised learning), and Kohonen networks (unsupervised learning), the authors concluded that the higher number of MRI modalities leads to better classification performance regardless of the processing method used.

For Alzheimer’s disease, the published studies [ 31 , 32 ] demonstrated the advantage of analyzing morphological and physiological changes in the brain with deep learning methods using MRI data. For example, Xinyang Feng et al. [ 31 ] compared the performance of a convolutional neural network to the classification based on morphometry markers, cognitive tests, and PET scans obtained from a multimodal dataset containing 975 images of patients with Alzheimer’s syndrome and 1943 images of healthy controls. The accuracy of the model based on the convolutional neural network turned out to be very high (AUC = 0.973). In addition, the developers were able to accurately determine specific anatomical features that turned out to be significant for the classification. It was due to the visually illustrative way of how information is represented in convolutional networks. This feature of convolutional networks was implemented in [ 32 ] as an independent diagnostic tool directly used by a medical professional. The researchers proposed not only to classify MRI data, but also to visually display the detected significant “abnormal” anatomical zones. Their relationship maps with morphological signs of atrophy correlated well with the results from earlier studies. This approach makes the application of the algorithm much more transparent for medical practitioners and increases confidence in its estimates.

To improve the speed of machine learning processing, Leonie Henschela et al. [ 33 ] proposed an algorithm for rapid segmentation of the entire brain into 95 classes using a convolutional neural network. The FastSurfer algorithm proposed by the authors has been commonly used as a means of preliminary data markup. This machine learning pipeline can be used as a standalone classifier and can also be integrated into more complex architectures.

There is an apparent lack of studies into the application of deep learning methods (when the model classifies “raw” data without access to human-designed features) to the analysis of blood and cerebrospinal fluid biomarkers. This can be explained both by the comparative simplicity of this data modality (a limited set of numerical values that is easily interpreted by a person), which eliminates the need to use “complex” solutions, and by insufficient research on this type of biomarker in the context of neurodegenerative disorder diagnostics in general. For example, in [ 34 ], a simple method for analyzing the concentration of p-tau181 in a patient’s blood is proposed, which provides accuracy and specificity sufficient for the preliminary screening of patients with suspected Alzheimer’s disease. However, the specificity of such a single marker is insufficient for a reliable final diagnosis, which makes it necessary to obtain additional tests. At some point, the amount of diverse data necessitates the use of more powerful machine learning methods.

It is important to note that the results of comparative studies show the advantages of deep learning models compared to simpler models in the diagnostics of various symptoms. The accuracy and specificity of deep learning models are higher than those of classical models [ 35 ].

An extensive comparative study of the accuracy and specificity of different algorithms for the detection of Parkinson’s disease was carried out by Mei et al. [ 36 ]. The authors concluded that the classification accuracy of machine learning algorithms was, on average, about ~94% for SPECT, ~86% for PET, and ~87% for MRI (including fMRI). The most used when working with neuroimaging data and the most accurate classifiers were SVM algorithms and artificial neural networks. SVM was used more frequently (50–70% for SPECT or PET and ~60% for MRI) compared to neural networks (22–53% for SPECT or PET and ~23% for MRI). Figure 2 demonstrates different data types and ML models described in the literature.

Data type, machine learning models applied, and accuracy. ( A ) Accuracy achieved in individual studies and average accuracy for each data type. Error bar: standard deviation. ( B ) Distribution of machine learning models applied per data type. MRI, magnetic resonance imaging; SPECT, single-photon emission computed tomography; PET, positron emission tomography; CSF, cerebrospinal fluid; SVM, support vector machine; NN, neural network; EL, ensemble learning; k-NN, nearest neighbor; regr, regression; DT, decision tree; NB, naïve Bayes; DA, discriminant analysis; other: data/models that do not belong to any of the given categories (reproduced from [ 36 ]).

Machine learning methods were also successfully employed in the studies into genetic markers of neurodegenerative disorders. For example, when studying miRNAs and non-coding RNAs, Ángela García-Fonseca et al. [ 37 ] noted the complex relationships of identified sites with specific disorders and outlined the possibility of modelling these relationships. The most effective algorithms were shown to be SVM and various types of decision trees. An example combining manual feature selection and deep learning was provided by Yuan Sh et al. [ 38 ]. As in the previous study, the authors confirmed, using statistical methods of data analysis, the presence of a strong correlation between miRNA expression and ratios of various blood cells. The authors selected the most informative features for their machine learning model. The selected features were associated with the presence of Alzheimer’s disease, mild cognitive disorders, and were also predictive of the disease’s progression. The final model was based on a non-linear neural network and demonstrated the out-of-sample classification accuracy of 91.59%. The area under the curve (AUC) for the control group, the mild cognitive impairment group, and the Alzheimer’s group was 0.97, 0.95, and 0.98, respectively.

Similarly, Jianhu Zhang et al. [ 39 ] suggested working with blood biomarkers. As in the studies described above, feature selection was performed manually based on statistical data. The neural network-based classifier demonstrated an accuracy of about 98%, which is roughly in line with the results shown in previous work.

The general mechanisms linking neurodegenerative and other mental disorders are important to consider for theories that aim to describe the relation between genetic markers and the progression of pathological changes in the brain [ 40 ]. The discovery of common mechanisms for neurodegenerative and mental changes might lead to new research avenues in omics technologies for diagnostics.

The growing popularity of computational approaches in bioinformatics, the relative simplicity of analyzing biological fluids statistically, and the good interpretability of the obtained results will likely lead to a steady increase in the number of studies on the assessment of neurodegenerative disorders using machine learning applied to genetic data. In addition, the possibility of using genetic data with instrumental and sociodemographic variables is attracting more and more attention [ 41 ].

The nature of the data obtained during the neuropsychological examination does not encourage the use of deep learning, so we will focus on classical algorithms. In one study, the use of 9 standardized psychological tests (including the assessment of memory, speech, daily activities, and general cognition) in combination with a test of spatial attention (also to evaluate cognitive abilities) in a sample of 28 patients with Alzheimer’s disease and 50 healthy controls made it possible to achieve a classification accuracy of 91.08% for the SVM algorithm [ 42 ]. The authors also evaluated several other algorithms: RF, gradient boosting (GB), and its variation adaptive boosting (AB). However, the performance of these algorithms was noticeably lower: 81.75%, 85.92%, and 80.75%, respectively. A slightly more sophisticated approach was implemented in another study, where the authors classified the patients into three groups (healthy controls, possible Alzheimer’s disease, and confirmed Alzheimer’s disease) with a two-stage algorithm based on additional modality data [ 43 ]. Initially, the authors identified patients at risk based on biomarkers: levels of cholesterol, blood glucose, high-density lipoprotein, glycosylated hemoglobin, and blood pressure, which was performed using SVM. In the second step, the results of a standardized cognitive ability test (CAT) of the selected individuals were classified by a polynomial logistic regression (LR) algorithm. The accuracy of the SVM classifier was 0.86 AUC, and the addition of LR increased it to 0.89 AUC; these results were achieved on a sample of 2361 patients.

Unfortunately, these and other similar studies are not of particular interest with regard to the objective diagnosis of neurodegenerative disorders. The data on the cognitive and psychological state of a person underlying their predictions have little specificity: similar symptoms (decreased cognitive abilities, lowered mood, etc.) are observed in various disorders with different etiologies. The more promising approach might be to combine data from neuropsychological assessments with data on pathophysiological changes in the brain (e.g., recorded using neuroimaging methods). This would allow not only to maintain the high sensitivity of neuropsychological data but also to significantly increase the specificity of diagnosis. For example, combining the results from the Addenbrooke’s Cognitive Examination, INECO Frontal Screening, and several parameters calculated from MRI data (in particular, volumes of gray matter in different areas of the brain) made it possible to carry out differential diagnosis of Alzheimer’s disease and frontotemporal dementia with an average accuracy of more than 0.91 AUC [ 44 ]. Moreover, despite the use of classic machine learning algorithms (the k-means method for primary feature clustering with LR for the final classification), this approach also demonstrated an outstanding generalizability: the model trained on patient data from one country (Argentina) retained an average accuracy above 0.9 AUC when validated on patients from other countries (Colombia and Australia). In another study, the authors examined different combinations of cognitive and neurophysiological markers to estimate which would perform the best for the differential diagnosis of Alzheimer’s disease (171 patients) and frontotemporal dementia (72 patients). They used a genetic machine learning algorithm for selecting optimal features for subsequent classification using SVM [ 45 ]. The final values of the f-measure (similar to AUC, but more reliable for data with a strong asymmetry in the distribution of class labels) were as follows: only for cognitive factors (memory, speech, attention, etc.), 0.882; only neurophysiological (hypermetabolism foci in various areas of the brain), 0.921; and the optimal combination of traits, 0.949.

The implementation of machine learning methods will likely reduce costs by enabling faster diagnostics as well as reducing the likelihood of medical errors. There are three main areas of application of modern methods of data analysis in the diagnosis of neurodegenerative disorders:

• 1. Increasing the accuracy and coverage of pre-clinical diagnostics. Identification of the pathological process before the onset of clinically observable symptoms will allow treatment to be started in advance, slowing down the progression of the disease and improving the quality of life of the patient.
• 2. Differential diagnosis of various neurodegenerative disorders and comorbidities. Increasing the accuracy of diagnosis will improve the selection of treatment and correction of symptoms.
• 3. Increasing the accuracy of forecasts for the progression of the disease after its onset. The application of machine learning methods to the analysis of longitudinal data will make it possible to build dynamic models for symptom monitoring.

1.8. The Treatment of Neurodegenerative Disorders

Due to the irreversible nature of the degeneration of the nervous tissue, the therapy of neurodegenerative disorders is aimed at slowing down the progression of degeneration and improving patients’ quality of life. Pharmacotherapy slows down degenerative processes through changes in the metabolism of neurons and glial cells. In addition, pharmacological interventions can be aimed at compensating for the functions of destroyed neuron populations. Non-pharmacological interventions are aimed at improving adaptation to the environment and improving the quality of life of the patient as well as helping patients to perform their household tasks. The main approach to non-pharmacological interventions is comprehensive rehabilitation that combines aerobic physical activity, cognitive training, and occupational therapy [ 46 ]. It is also possible to additionally use methods of brain stimulation.

1.9. Pharmacological Therapy

The main drugs used in the pharmacotherapy of neurodegenerative disorders are cholinesterase inhibitors (donepezil, rivastigmine), NMDA receptor antagonists (memantine), and their combinations [ 47 ]. Dopamine receptor agonists (apomorphine), dopamine precursors (levodopa), and monamine oxidase inhibitors (MAO-B) are used to reduce the intensity of motor disorders in parkinsonism (clinical syndrome) [ 48 ]. At the moment of publication, studies into potential neuroprotective drugs are being conducted, but there is no consensus on the use of such drugs in practice.

1.10. Cognitive Training

Exercises aimed at training memory, attention, and thinking have been extensively used as a therapy for individuals with neurodegenerative disorders [ 46 ]. Examples of such training include episodic memory capacity training with mnemonic techniques and visual information processing speed training [ 49 ]. The meta-analysis of 32 studies on the effectiveness of cognitive training in individuals with neurodegenerative disorders showed moderate effectiveness of training for improving cognitive functions [ 50 ]. However, many of the studies included in the meta-analysis were of low quality and had small sample sizes, so the authors cautioned against drawing definitive conclusions. At the moment, the effectiveness of cognitive training for the therapy of neurodegenerative disorders has not been sufficiently studied.

1.11. Physical Exercises

Decreased levels of neurotrophic factors—mainly brain-derived neurotrophic factor (BDNF) and its receptors—are one of the most common physiological consequences of various neurodegenerative disorders. The meta-analysis of 18 randomized trials showed an association of exercise with neurotrophic factor levels in patients with neurodegenerative disorders. Physical exercise increases the level of a neurotrophic factor in blood plasma [ 51 ]. High concentrations of neurotrophic factors reduce the toxic effect of nerve cell death [ 52 , 53 ]. In the studies, an increase in the levels of neurotrophic factors in plasma occurred regardless of the type of exercise: aerobic, strength, or combined exercise programs had similar effects.

1.12. Ergotherapy

Lifestyle counselling, socialization assistance, and re-training of everyday skills are critical for slowing down the progression of neurodegenerative disorders and improving patients’ general quality of life [ 54 ]. In some European countries and the US, occupational therapy is often provided by specialized employees of medical and social institutions (occupational therapists). In Russia, occupational therapy is usually provided by clinical psychologists.

1.13. Brain Stimulation

Methods of TMS and deep brain stimulation (DMS) have not demonstrated significant effects on clinical outcomes in neurodegenerative disorders [ 55 ]. One exception is the use of the DMS method for the correction of motor symptoms in Parkinson’s disease. The DMS stimulation led to a clinically significant decrease in tremor and muscle rigidity in patients.

1.14. Prevention of the Associated Psychological Problems

Affective symptoms (including anxiety and depression) can occur in patients with neurodegenerative disorders. The symptoms could occur as a direct result of dementia, and independently as a reaction to one’s disease or social situation (e.g., as a reaction to impaired communication with relatives) [ 56 ]. Psychosocial interventions could be aimed at facilitating social support and managing the symptoms of anxiety and depression.

Early pre-clinical diagnostics of neurodegenerative disorders will enable the early start of pharmacological therapy, which would slow down the degeneration of the brain matter, as well as cognitive training, which would slow down the development of cognitive deficits. Currently, pre-clinical diagnostics of neurodegenerative disorders using neuroimaging, brain stimulation, and analysis of brain electrical activity is the most promising area for the application of machine learning and data analysis methods.

1.15. Application of Machine Learning

Since the changes in brain matter that occur in neurodegenerative disorders are often irreversible, their therapy is aimed primarily at maintaining the quality of life of patients. Machine learning might be employed to assist with the therapy in several ways: the evaluation of the effectiveness of therapeutic interventions, dynamic monitoring of the disease progression, and the optimization of brain stimulation methods.

When conventional DBS therapy is conducted, continuous pulsation is delivered without adjustment for the patient’s current condition. This approach could lead to sub-optimal results. One possible way to optimize the delivery of pulsation would be to use real-time data on the patient’s condition to select the right moment for stimulation. These data can include information obtained from chemical sensors (e.g., dopamine levels), accelerometers, and ECoG and EMG data [ 57 ]. For example, the occurrence of tremors can be detected in real time by applying a pre-trained classifier to the EMG signal [ 58 ]. In the case of DBS, this approach is called adaptive DBS.

Other methods of pharmacological therapy and physiotherapy used in individuals with neurodegenerative disorders are less invasive compared to DBS. However, since these methods’ effects are not immediate, evaluating their effectiveness can be more challenging. This problem can be solved by adapting the approach used for diagnostics: automated gait analysis [ 59 ] using wearable devices or computer vision systems. A similar approach is used to prevent fall injuries by continuously monitoring and evaluating the probability of falling using a trained regression model [ 60 ]. The analogous solutions for the assessment of neurodegenerative disorders are discussed in the corresponding section.

In conclusion, it is important to note the partial intersection of the proposed methods both in the feature space and in the algorithms being used. There is an emerging approach to aggregate individual markers and even intermediate diagnostic models into higher-level hierarchical ensembles. This approach entails an end-to-end system for processing instrumental data, whose results can then be used in different processes including diagnostics and disease progression monitoring (as depicted in Figure 3 ).

A multilevel approach to working with data on neurodegenerative disorders [ 4 ].

Such an approach, however, is characterized by high computational, technological, and regulatory complexity. To develop such multichannel systems, it is also necessary to address the issues of collecting and pre-processing data, cross-checking the fit of models, and verifying the effectiveness of the analyses.

2. Depressive Disorders

Around 15% of the world population has been diagnosed with a depressive disorder at least once in their life [ 61 ]. The average annual prevalence of depressive disorders among the adult population, according to the World Health Organization, is 5% [ 62 ]. Depression is one of the most common mental disorders. Depression negatively impacts the ability to work, decreases the quality of life, and constitutes a major risk factor for suicide [ 63 ] and other adverse health outcomes [ 64 ]. Subclinical depression may also precede the onset of neurodegenerative disorders [ 65 ].

2.1. Diagnostics of Depressive Disorders

Depressive symptoms can have many etiologies, and their potential neurophysiological mechanisms have not been well understood yet. The current consensus is to consider depressive states as a consequence of poorly differentiated functional, neurotransmitter, and metabolic changes in the brain [ 64 ]. Since there is no specific localized pathological process in the nervous tissue in depressive disorders, neuroimaging methods for the diagnosis of depression have been rarely used in practice. Biomarker studies of blood and cerebrospinal fluid have been also rarely used. Diagnosis of major depressive disorder, minor depressive disorder, or bipolar depression is usually based on the results of clinical interviews and standardized questionnaires. The potential use of brain electrical activity data to diagnose depression and predict therapeutic responses to antidepressants is currently the subject of active research.

2.1.1. Use of Standardized Questionnaires

In addition to the clinical interview, standardized questionnaires are used to assess the severity of depressive symptoms. In Russia, among others, the following are used for this: the Beck Depression Inventory (BDI) [ 66 ], the Hamilton Scale (HAM-D) [ 67 ], and the Hospital Anxiety and Depression Scale (HADS) [ 68 ]. The diagnosis of a depressive disorder is not made solely based on questionnaires. Their results serve as a guide for a psychiatrist in determining the severity of symptoms and finalizing the diagnosis. The final diagnosis is based on the combination of identified symptoms that fit the diagnostic criteria of the International Classification of Diseases. However, the periodic completion of questionnaires by the patient can be used as a relatively objective way to track the dynamics of depressive symptoms to adjust the course of treatment.

2.1.2. Analysis of the Functional Activity of the Brain (EEG)

Methods for analyzing brain activity can be used both to diagnose depressive states (in particular, to assess the severity of depression) and to predict the response to pharmacological treatment. For example, a decrease in interhemispheric coherence of the frontal leads, observed regardless of the etiology of the depressive state, has been identified as a potential predictor of resistance to pharmacotherapy [ 69 ]. Identification of predictors of drug resistance and response to drugs is one of the directions of research on the use of EEG methods in the diagnosis of depression [ 70 ].

2.1.3. Application of Machine Learning

Attempts have been made to use neuroimaging data (MRI, functional MRI, and diffusion MRI) [ 71 ] and brain electrical activity (EEG) data [ 72 ] to diagnose depressive disorders. The recorded data were analyzed with classical machine learning algorithms (e.g., the support vector machine) and deep learning methods [ 73 ].

The main types of data sources and methods used in the assessment of various mental disorders, including depression, were described in the review by Chung and Teo [ 74 ]. The authors analyzed several works on the assessment of anxiety and depression based on written texts, voice recordings, data from MRI scans, survey methods, and their combinations. The paper noted that different algorithms were better suited for different types of data (they examined gradient boosting, random forest, and neural networks), but the average accuracy of the best-performing algorithms did not exceed 80%. The authors attribute the low results both to the small amount of data and to incomplete or imbalanced training datasets. Chen et al. [ 75 ] also noted the need to use additional data channels, including voice, activity, sleep, questionnaire, and instrumental methods (e.g., MRI) to potentially improve the accuracy of diagnosis. Additional data channels can also represent different aspects of everyday activity to add behavioral patterns into the data. The authors emphasized the promise of multimodal datasets, since various mental disorders are often accompanied by external signs progressing with time ( Table 1 ).

Multimodal data features and their uses in mental health [ 75 ].

The use of multimodal data potentially makes it possible not only to diagnose a wide range of mental disorders, but also to monitor the dynamics of the development of the disease, since the observed signs become more pronounced at later stages. However, the collection of multimodal data is often associated with a number of difficulties, in particular with the need to observe the patient for a long time, as well as to confirm the observed signs by cross-checking on historical data performed by an experienced doctor. Big data collected from mobile devices reflect the patient’s social interactions and fine motor skills features such as taps duration, typing speed and rhythm, reaction speed, etc. It is worth noting that in order to automatically collect such an amount of data, the mobile application installed on the patient’s device should have access to the correspondence and system software, which can become a security issue. However, such solutions can be easily implemented on the architecture of social interaction platforms such as social networks. The authors proposed the implementation of platform solutions for big data as the main processing unit. The units then are used as a basis for more complex models. In addition, the authors proposed a machine learning model that can be used to evaluate the effectiveness of treatment and to monitor the patient’s condition using biofeedback.

One of the features of mental disorders, when compared to other disorders, is that their assessment relies heavily on the subjective experience of a patient subjectively described by him or her. The quality of the assessment is highly dependent on the clinical experience of the medical professionals conducting the assessment. To systematize the data collection process of mental illness clinical signs, a general methodology for filling the database is needed. The same methodology, on the other hand, forces doctors to use standardized data collection forms, which positively affects the entire dataset. The fact is that the different experiences of doctors and the peculiarities of the method of presenting the picture of the disease cause certain markup anomalies in the database [ 76 ]. Often, this feature does not allow one to reasonably assert that the sample data are not biased. In addition, the amount of relevant data differs from channel to channel, which often requires the deployment of different cleaning and pre-processing algorithms.

In their review of machine learning methods for neuroimaging, Quaak et al. [ 77 ] pointed out the need for a more conservative approach to testing the quality of trained models, since their results on the test set often do not reflect the model’s real performance. In addition, the authors noted the popularity of the EEG as a source of data for the diagnosis of mental disorders. MRI/EEG databases for diagnostics of depression, which are used to train machine learning algorithms, can be recorded in different modes: when a patient is performing certain tasks (game), responding to external stimuli (video, images, and music), and in a resting state. The most common approach is to receive data from a person in a resting state (eyes open or closed).

The information about the patient’s depression is often considered protected private information, which leads to very few datasets being openly available. The information on openly available datasets for diagnostics of depression by EEG is presented in Table 2 .

Available open datasets for diagnostics of depression by EEG.

The common problem with openly available datasets is that they often include a limited number of participants (the average sample size of patients with depressive disorder is around 30 people). This problem calls into question the diagnostic value of studies conducted using the datasets. This problem can be mitigated either by combining existing datasets [ 82 ] or by the collection of more data. The latter option requires substantial resources.

Convolutional neural networks are the most common types of neural networks for classifying the presence or absence of a depressive disorder. They rely on a two-dimensional or one-dimensional convolution operation, or hybrid models of convolutional and recurrent neural networks [ 83 ].

The input of the neural network is either the raw EEG signal (after pre-processing in the form of filtering and denoising) or its converted version. Given the existence of well-known models of neural networks used for image classification, the transformation often consists of the formation of an “image” from the EEG signal, which is then fed to the input of a two-dimensional convolutional network. To form an “image”, the power values of brain rhythms can be used, which are spatially projected onto a plane, following the location of the electrodes [ 84 ].

Since deep learning algorithms are demanding towards an amount of training data, and the existing samples of training data are often of limited size, the augmentation method is applied. Augmentation is a method of artificially increasing the amount of data used for training [ 85 ]. The use of data augmentation in training deep neural networks can reduce the effect of overfitting and improve accuracy and stability. EEG data augmentation uses sliding window data sampling, data from generative models, noise addition (generally Gaussian noise is used), sampling, segment recombination, and Fourier transform [ 86 ]. The architectures of neural networks and the achieved accuracy of depression detection according to EEG data are presented in Table 3 .

EEG classification accuracy for depressive disorder.

It should be noted that the claimed high accuracy of depression classification, exceeding 90%, may be associated with testing models on a limited set of initial data and possible incorrect partitioning of training data by patients, which might have led to the implicit leakage of training data into the test subset. Leakage of this kind leads to overestimated accuracy and reduces the generalizability of the model due to the effect of overfitting [ 99 ].

The application of methods for automated assessment of depressive disorder severity involves the selection of relevant features of the received signal and the development of complex metrics of the depressive state based on them. Isolation of non-linear spectral characteristics of EEG signals in combination together with the support vector machine methods demonstrated high classification accuracy in several studies [ 100 ]. The combination of linear discriminant analysis and genetic algorithms also demonstrated high discrimination performance when classifying patients with the depressive disorder [ 101 ]. This line of research can be developed further by incorporating new diagnostic categories as comparison groups and the development of classifiers with higher ecological validity.

The development and validation of composite diagnostic indices, as well as ensembles of algorithms for solving specific diagnostic problems, are most often carried out on small samples in the absence of independent external validation. This creates a high risk of overfitting the algorithms and reduces the generalizability of the results. To increase the validity in the development of EEG-based diagnostic indices, attention should be paid to the composition of a training sample, its size, and its quality. The use of the combination of neurophysiological data labelled according to a single protocol from different clinical sites could be an optimal solution.

2.2. Treatment of Depressive Disorders

The main methods of treatment of depressive disorders include pharmacological therapy and psychosocial interventions (psychoeducation and group and individual psychotherapy) [ 102 ]. Biofeedback and TMS are less commonly used. In individuals with depression, relevant therapeutic targets include maladaptive thoughts and beliefs, lowered mood, quality of life, and others. Reducing the risk of suicide in severe depression is also an important therapeutic goal.

2.2.1. Pharmacological Therapy

The most common medication used for treating depression is serotonin reuptake inhibitors (e.g., sertraline and paroxetine), tricyclic antidepressants (citalopram and fluoxetine), and monoamine oxidase inhibitors (moclobemide and pirlindole). The mechanism of action of antidepressants is based on a change in the concentration of neurotransmitters available for binding in the brain, which leads to long-term potentiation or depression of synaptic connections [ 103 ]. The choice of medication is made based on their safety and the observed therapeutic effect in a particular patient. Doses are adjusted by the medical professional during treatment. Antipsychotics can also be used in depressive states that have arisen as part of psychotic disorders (e.g., bipolar disorder) [ 104 ]. Pharmacological therapy is usually only required for moderate to severe depressive disorders. The combination of pharmacological therapy and psychotherapy is optimal for the treatment of such forms of depression [ 105 ].

2.2.2. Psychosocial Interventions

The main psychosocial intervention for working with patients with mild depression is psychoeducation, providing information about the symptoms and the possible progression of the disorder, recommendations for self-help, and lifestyle changes (if necessary). Other suitable psychosocial interventions for patients with depressive disorders are group and individual psychotherapy. The most studied approaches for the treatment of depression are cognitive behavioral therapy and its variations, psychodynamic therapy, schema therapy, and decision-oriented therapy [ 106 , 107 ]. In general, psychotherapeutic approaches do not differ significantly in effectiveness, so the choice of a particular approach in each case depends on the availability of a specialist and the individual preferences of the patient. Besides psychotherapy, regular physical exercise could also be considered a part of a psychosocial intervention. Regular exercise has been shown to be effective in reducing depressive symptoms [ 108 ]. Positive changes may result from the metabolic and hormonal changes that accompany regular exercise.

2.2.3. Biofeedback

Biofeedback methods have also been shown to be effective in reducing depressive symptoms. A meta-analysis of 14 randomized controlled trials in 794 subjects showed significant improvements in symptoms in patients with depression [ 109 ]. Neurofeedback using EEG and fMRI demonstrated promising results in reducing depressive symptoms, but the number of published studies is still limited [ 110 , 111 ].

2.2.4. Brain Stimulation

Studies demonstrated that transcranial magnetic stimulation had a potentiating effect on antidepressant intake and was associated with improved clinical outcomes [ 112 ]. TMS is recommended by several national medical agencies for use in the treatment of depression as a procedure with potentially positive results and no side effects [ 113 ]. However, there is currently insufficient data to draw definitive conclusions about the effectiveness of TMS in reducing depressive symptoms.

2.2.5. Application of Machine Learning

Machine learning is also used to assess the treatment effectiveness in depressive disorders [ 114 ]. The effectiveness of depression treatment using antidepressants or TMS can be assessed by using and applying classical machine learning methods (support vector machine [ 115 ] and random forest [ 116 ]) or deep neural network [ 97 ] algorithms to EEG data [ 117 ].

The machine learning methods used and the accuracy achieved in predicting depression treatment based on EEG data are presented in Table 4 .

EEG predictive accuracy of treatment for depressive disorder.

Studies on the predictive treatment of depression, as well as studies on the diagnosis of depression, are characterized by a limited number of participants, which does not allow us to assert that the methods used have sufficient generalizability. At the moment, the most promising areas for the development of applied methods of machine learning for the therapy of depression are dynamic monitoring of the symptoms and treatment of pharmacoresistant depression.

3. The Use of Auxiliary Indicators

Along with well-established instrumental methods, such as MRI, fMRI, and EEG, it became possible to automatize the use of several auxiliary diagnostics indicators. These indicators had been used as visually observable markers of a disorder and were usually determined by a medical professional during direct contact with a patient. At the same time, the evaluation of the severity of these indicators and their association with the symptoms were always performed by medical professionals based on their subjective experience. Such auxiliary indicators include both observed disturbances in motor functions (gait, gross and fine motor skills, the ability to speak and swallow, etc.) as well as their various combinations with cognitive impairments. Automating the processing of such indicators is a promising direction of work for machine learning specialists since relevant training datasets are easy to collect and label, e.g., based on neuroimaging methods. For some types of motor disorders, researchers have already obtained promising results [ 120 , 121 , 122 , 123 ].

Instrumental methods for determining gait disorders, which are quite pronounced, in particular for Parkinson’s disease cases, are based on data from video cameras and sensors of wearable devices. Wearable devices are relatively easy to construct, and it is possible to use them outside the clinical setting to collect more data. On the other hand, small-sized three-axis accelerometers and inertial systems most often used for such solutions are subject to external interference because the device is loosely attached to the body and interacts with clothing and furniture. The daily activity data obtained from such devices are poorly interpreted outside the context of the situation. Nevertheless, the results of comparative experiments using such devices and machine learning methods look quite promising.

For example, Dante Trabassi et al. [ 122 ] provided a comparative analysis of various machine learning algorithms on a dataset from inertial sensors of a wearable device to detect gait anomalies which are common for Parkinson’s disease. The trials were conducted with 81 subjects with Parkinson’s disease and a control group of 80 subjects. The test protocol was the same for the test and control groups and included walking along a building corridor which was about 30 meters long with a wearable device in the form of a sensor on the belt. The collected data was pre-processed manually and then used to train several different machine learning models, such as decision trees, random forests, k-nearest neighbors, support vector machines, and artificial neural networks. The best result with an accuracy of 0.86 was achieved using an algorithm based on the support vector machine. At the same time, artificial neural networks showed a lower accuracy, comparable to a random forest.

In contrast, in an article by Robbin Romijnders et al. [ 123 ], the best result (classification accuracy of about 98% and classification completeness >95%) was shown using artificial neural networks. In this publication, a different approach to sensor placement was used, and gait features were reconstructed based on data from accelerometers and position sensors placed on the lower leg and ankle. The study used a larger dataset of 157 subjects with several disorders (Parkinson’s disease, multiple sclerosis, recent stroke, and chronic low back pain) and a larger test program that included the passage of a five-meter segment at a randomly defined slow, medium, and fast pace. Deep learning using temporal convolutional networks was chosen as the primary algorithm for this research. The high result shown allows us to conclude that this approach is suitable both for the preliminary screening of patients and for monitoring the dynamics of disease development.

Jinglin Sun et al. [ 120 ] proposed a similar approach for the diagnosis of Alzheimer’s disease, but eye movements were used as the primary source. A comparative study was conducted for classical machine learning methods (support vector machine and k-nearest neighbors) and well-established neural network architectures VGG-16 and Resnet-18, and for the architecture proposed by the authors based on a dual autoencoder module with a separate final classifier in the form of a fully connected three-layer network. Mean accuracy values of 0.87 ± 0.04 and recall values of 0.89 ± 0.04 were demonstrated, respectively.

Chonghua Xue [ 124 ] proposed to use and analyze patients’ voice recordings by convolutional and recurrent neural networks for the detection of Alzheimer’s disease. This architecture is well suited for time-distributed data; therefore, it is often used in voice analysis.

A number of works proposed an analysis of behavioral responses for the diagnosis of depression [ 125 , 126 ]. In such experiments, a patient is asked to determine the emotional coloring of various words and images that are demonstrated by a special program. The accuracy of the classifier based on bagged decision trees, trained on a sample with manual feature selection, reached 0.76, which does not allow us to speak of a stable result.

The issue of predicting and early diagnosis of cognitive dysfunctions in the elderly using machine learning methods on various data, including sociodemographic data, electronic medical records, clinical and psychometric studies, and neuroimaging data, is considered in a review paper [ 127 ]. Sarah Graham et al. highlighted the need to use multimodal datasets and maintain the transparency of the algorithms. They also noted the direct comparative analysis of the effectiveness of clinical methods for the early diagnosis of cognitive impairment and new methods based on machine learning have not yet been carried out.

A growing number of studies on the potential application of machine learning highlights the need of a general approach to evaluate such potential applications. Without determining the existing barriers to the introduction of technology, its features, strengths, and weaknesses, and understanding the methods for ensuring its safe use, the implementation of such methods in clinical practice will be difficult [ 128 ]. For most countries, the procedures of state regulators regarding the introduction of new medical instruments, as well as ethical issues related to the low transparency of the results of the model and the specific features of collecting and processing medical data, significantly slow down the introduction of systems based on machine learning. This was especially evident during the COVID-19 pandemic when many predictive, diagnostic, and clinical models were created, but only a few point developments reached successful practical use in medical practice. From a technical point of view, when developing solutions based on machine learning, it is necessary to carefully check all the main stages of data collection and preparation, justify the choice of a specific algorithm for solving the problem, and explain the results obtained [ 129 ].

Machine learning methods are gradually becoming an increasingly mature research tool in areas such as radiology, immunology, and the synthesis of new drugs [ 130 ]. There are studies on the application of machine learning methods in general, and in relation to the tasks of diagnosing neurodegenerative and mental diseases. In particular, there are a number of inter-related works that are being actively studied.

The issue of achieving consensus between several experts when labeling data and the features of training models on such a set is covered in [ 131 ]. The authors managed to get the convolutional neural network to reproduce the manual data markup features that are specific to each expert. Thus, the algorithm imitates the consensus of experts in the form of an ensemble of layers trained on separate markups.

The phenomenon of sustainable overestimation of the effectiveness of machine learning methods developed for use in psychiatry was studied in [ 132 ]. The authors noted that in the Predictive Analytics Competition (PAC-2018), in which participants were asked to identify depression from MRI scans, the results were between 60% and 65% on a large independent dataset, while on smaller datasets, the accuracy reached 80% or more. A more detailed study showed a steady overestimation of the accuracy of the model when testing on small sample sizes. For simple methods, in which the number of hyperparameters is small, with the correct tuning, the model is trained quite quickly, and with an increase in the size of the training sample, no further noticeable increase in accuracy and recall is observed. At the same time, many researchers split the dataset in such a way that as much of the data as possible is used in the training process. Given the relatively small number of publicly available datasets for training, the question of the realism of the declared high performance of the developed models when tested on a new larger independent dataset becomes increasingly relevant.

Based on the foregoing information, machine learning methods in no way can be considered a universal recipe for a diagnostician or therapist. A systematic approach has yet to be formulated and, at the moment, there are a number of shortcomings that are characteristic of not fully formed methodologies. Functionally, the first group can be attributed to “data and methods”, and the second to “the interpretability and trust”.

First, there is no generally accepted method for collecting information on the course of neurodegenerative diseases, on the basis of which it would be possible to collect a uniform and consistent dataset. As a result, available datasets described in the literature are often fragmentary and unbalanced or contain a limited number of respondents. The average sample size in terms of the number of respondents is about 500 people. The datasets themselves may contain more detailed information about each patient, but for a relatively small number of respondents. Thus, the dataset in [ 27 ] was collected from 277 patients. The largest dataset by the number of patients includes several thousand images but belonging to different patients [ 31 ]. It is worth noting the difficulty of tracking disease development because of data format limitations since they reflect the clinical situation only at a specific point in time.

To use the data in training models, it is required to clean them, balance by classes, and correctly split them into training and test sets. In addition, manual feature engineering based on the experience of the investigator and known patterns becomes important. Thus, in heterogeneous datasets, the features are difficult to validate and augment with data from other patients to increase the number of features. Augmentation with synthetic data is a promising and effective method, applicable for image [ 85 ] and time-series EEG data [ 86 ]. The selection of features for machine learning also depends on the characteristics of the dataset. A comparative study of some open datasets is given in [ 114 ]. Comparison of datasets, model’s accuracy, and participant-to-feature ratio is shown in Figure 4 .

Balanced accuracy and participant-to-feature ratio in published machine learning studies of outcome prediction in the treatment of MDD. The x-axis plots the ratio of participants to predictive features. The y-axis plots the mean balanced accuracy within each study. Studies predicting response, remission, and TRD are plotted as circles, diamonds, and triangles, respectively. Adequate quality studies are highlighted with large, filled symbols. The dark gray horizontal dashed line shows the mean balanced accuracy of the eight adequate quality studies. The pale gray horizontal dashed line shows the average balanced accuracy of the other 45 studies (reproduced from [ 114 ]).

After data preparation, the model learns on the samples in the training set. As data differ in structure and volume in different sets, selection of a machine learning model becomes a non-trivial task. In particular, it is difficult to control the quality of learning in small datasets. In addition, the generalization ability of models is often overestimated due to the small sample size and possible overfitting, which was described in detail above. As a result, researchers offered a large number of models, with different architecture and characteristics. Such a cardinal difference does not allow for an adequate quantitative comparison for ensembled models, despite the simplicity of models such as decision trees or k-nearest neighbors.

Su et al. in [ 41 ] confirmed the complex nature of data mining and processing task for genetic and transcriptomic data on PD. The authors summarize the remaining limitations and challenges, and accordingly discuss potential future directions which may lead to promising machine learning approaches to address the discovered issues. Table 5 demonstrates several common points about dealing with mixed datasets and additional methods, applied in sequence both for the data and the model. The separate point on the interdisciplinary issue highlights the need for better collaboration and better domain expertise, as an expert’s opinion is still often the best ‘ground truth’ source.

Summary points of challenges and potential future directions to address them [ 41 ].

The issues of interpretability and trust to algorithms, listed in Table 5 , can be broadened in several aspects. First, the regulatory framework for the use of machine learning models in clinical practice is rather incomplete. This is inherent not only in the diagnostics and treatment of neurodegenerative diseases, but rather in the clinical approach as a whole. As noted in [ 76 ], the methodology for processing mental disorders data for precision medicine approach is only being formed and it will take time to develop it. Second, whenever we use an artificial intelligence approach in healthcare, it is always related to making a decision. Disease detection, disease prediction, drug repurposing, precision medicine, medical resource allocation, and much more can be shown as evidence for it. In all of them, a machine learning algorithm either makes a decision or at least supports a decision. The system works as a complex visualization and analyzing tool to catch patterns in medical data [ 129 ]. The true reason why the recommendation system supports one decision or another is often hidden from the user. So, the questions of ethics and potential software issues arise and stay unresolved before medical systems become more mature and common in clinical practice. Quality and integrity checks of data, models, and protocols can help to speed up the development of ML-based medical systems and discover potential faults during development and trial stages.

In a study by Vinny et al. [ 133 ], the authors critically examined the main issues of quality assessment and the possibility of implementing solutions based on machine learning models. The authors noted the need for a consistent multistage analysis of each of the components of the solutions offered to physicians, starting with the dataset, its size, completeness, representativeness, method and protocol of collection, and ending with the re-checking of the demonstrated characteristics of the model and the method of their determination. Separately, the authors noted the importance of transparency, the interpretability of the applied model, and the “common sense” check. The authors formulated the main provisions in 14 questions as shown in Table 6 .

Summary of key questions in critical appraisal of a machine learning research paper [ 133 ].

4. Conclusions

The promising areas of research outlined in this review are currently being explored by various scientific groups around the world. The main factor limiting research in the identified areas is the lack of data of adequate quality and volume for training and validating machine learning models. The introduction of common standards for collecting, labeling, and storing data in shared databanks will allow the development of methods for the early diagnostics and correction of neurodegenerative and depressive disorders.

Funding Statement

This work was performed within the scope of the Agreement #FZWM-2020-0013.

Author Contributions

Conceptualization, N.S. and I.V.; methodology, D.Y. and D.S.; validation, S.B., D.S., N.S. and D.Y.; data curation, S.B., V.S. (Viktor Sapunov), D.Y.; writing—original draft preparation, D.Y., S.B., V.S. (Viktor Sapunov) and V.S. (Vladimir Savinov); writing—review and editing, S.B., D.Y., D.S., G.K., N.S. and I.V.; supervision, N.S. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

Conflicts of interest.

The authors declare no conflict of interest.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Stirring up emotions – Parkinson’s disease alters emotion-related bodily sensations

Researchers at the University of Turku and Turku University Hospital in Finland have shown that bodily sensations related to emotions are altered by the neurological disorder Parkinson's disease.

Emotions have a major impact on the way we act, and they regulate our hormones and many of the body’s vital functions. Emotions can also be associated with strong physical reactions and sensations, such as an increase in heart rate and blood pressure when a berry picker encounters a bear in the woods or butterflies in the stomach when going on a first date. Emotions are also reflected in the symptoms of many neurological and psychiatric disorders, with negative emotions in particular often increasing the symptoms of the disease.

Parkinson's disease is a neurological movement disorder that is characterised by motor symptoms, such as slowness, stiffness, and tremor. Parkinson's disease is also associated with a number of non-motor symptoms, such as depression, anxiety, and dysfunction of the autonomic nervous system. Dysfunction of the autonomic nervous system has an influence, for example, on the blood circulation and gastrointestinal tract function.

Doctoral Researcher and physician specialising in neurology, MD Kalle Niemi and his colleagues investigated the bodily sensations of basic emotions (anger, disgust, fear, happiness, sadness, surprise, and neutral) in Finnish Parkinson's disease patients.

The subjects were asked to identify their symptoms and bodily sensations associated with different emotions by drawing them on an electronic human body map using a computer mouse.

The descriptions of the Parkinson's disease patients on their emotion-related bodily sensations differed from those of the control subjects. Photo: Clinical Neurosciences, University of Turku

People with Parkinson's disease were found to have significant differences in all bodily sensations related to basic emotions when compared with the control subjects. The differences were most pronounced in the bodily sensations of anger, which in healthy people are focused in the chest area. In people with Parkinson's disease, the bodily sensation of anger in the chest was reduced and seemed to shift more to the abdominal region as the disease progressed, consistent with the dysfunction of the autonomic nervous system associated with Parkinson’s disease.

“In recent years, there has been a growing realisation that the non-motor symptoms of Parkinson's disease have a significant impact on the patients’ quality of life. The results of our study highlight yet another non-motor phenomenon," says Niemi.

Emotional abnormalities are common in psychiatric disorders, but this study is the first to show abnormalities in the emotion-related bodily sensations in a neurological disorder. The results may open up new perspectives into the symptoms and possibly even treatment of symptoms in neurological disorders.

“The results of our study raise many interesting questions about the role of emotions in the symptoms of Parkinson's disease. Extending our research method to other diseases offers new possibilities for neurology research," summarises Juho Joutsa , Professor of Neurology at the University of Turku and principal investigator of the study.

The research article Bodily maps of symptoms and emotions in Parkinson's disease  has been published on 8 April 2024 in Movement Disorders , the most prestigious journal in movement disorder neurology.

More information:

Doctoral Researcher Kalle Niemi University of Turku [email protected]

Professor Juho Joutsa  University of Turku [email protected]

Image for media:

The photos are free to use in news concerning the topic. Please credit the photographer or the publication when using the photos.

> The descriptions of the Parkinson's disease patients on their emotion-related bodily sensations differed from those of the control subjects . Photo: Clinical Neurosciences, University of Turku

Read the original research article

You might also be interested in.

Increased Risk of Parkinson's Disease in Patients with Schizophrenia

Parkinson’s Patient Skin Samples Provide Clues to Disease Mechanism and a Possible Clinical Test

Apr. 12, 2024

Rice team demonstrates miniature brain stimulator in humans, tiny skull implant places rice lab at the forefront of neurotech research.

Rice University engineers have developed the smallest implantable brain stimulator demonstrated in a human patient. Thanks to pioneering magnetoelectric power transfer technology, the pea-sized device developed in the Rice lab of Jacob Robinson in collaboration with Motif Neurotech and clinicians Dr. Sameer Sheth and Dr. Sunil Sheth can be powered wirelessly via an external transmitter and used to stimulate the brain through the dura ⎯ the protective membrane attached to the bottom of the skull.

The device, known as the Digitally programmable Over-brain Therapeutic (DOT), could revolutionize treatment for drug-resistant depression and other psychiatric or neurological disorders by providing a therapeutic alternative that offers greater patient autonomy and accessibility than current neurostimulation-based therapies and is less invasive than other brain-computer interfaces (BCIs).

"In this paper we show that our device, the size of a pea, can activate the motor cortex, which results in the patient moving their hand,” said Robinson, a professor of electrical and computer engineering and of bioengineering at Rice. “In the future, we can place the implant above other parts of the brain, like the prefrontal cortex, where we expect to improve executive functioning in people with depression or other disorders.”

Existing implantable technologies for brain stimulation are powered by relatively large batteries that need to be placed under the skin elsewhere in the body and connected to the stimulating device via long wires. Such design limitations require more surgery and subject the individual to a greater burden of hardware implantation, risks of wire breakage or failure and the need for future battery replacement surgeries.

“We eliminated the need for a battery by wirelessly powering the device using an external transmitter,” explained Joshua Woods , an electrical engineering graduate student in the Robinson lab and lead author on the study published in Science Advances. Amanda Singer, a former graduate student in Rice’s applied physics program who is now at Motif Neurotech, is also a lead author.

The technology relies on a material that converts magnetic fields into electrical pulses. This conversion process is very efficient at small scales and has good misalignment tolerance, meaning it does not require complex or minute maneuvering to activate and control. The device has a width of 9 millimeters and can deliver 14.5 volts of stimulation.

“Our implant gets all of its energy through this magnetoelectric effect,” said Robinson, who is founder and CEO of Motif , a startup formed through the Rice Biotech Launch Pad that is working to bring the device to market. “The physics of that power transfer makes this much more efficient than any other wireless power transfer technologies under these conditions.”

Motif is one of several neurotech companies that are probing the potential of BCIs to revolutionize treatments for neurological disorders.

“Neurostimulation is key to enabling therapies in the mental health space where drug side effects and a lack of efficacy leave many people without adequate treatment options,” Robinson said.

The researchers tested the device temporarily in a human patient, using it to stimulate the motor cortex ⎯ the part of the brain responsible for movement ⎯ and generating a hand movement response. They next showed the device interfaces with the brain stably for a 30-day duration in pigs.

“This has not been done before because the quality and strength of the signal needed to stimulate the brain through the dura were previously impossible with wireless power transfer for implants this small,” Woods said.

Robinson envisions the technology being used from the comfort of one’s home. A physician would prescribe the treatment and provide guidelines for using the device, but patients would retain complete control over how the treatment is administered. “Back home, the patient would put on their hat or wearable to power and communicate with the implant, push ‘go’ on their iPhone or their smartwatch and then the electrical stimulation from that implant would activate a neuronal network inside the brain,” Robinson said.

Implantation would require a minimally invasive 30-minute procedure that would place the device in the bone over the brain. Both the implant and the incision would be virtually invisible, and the patient would go home the same day. “When you think about a pacemaker, it’s a very routine part of cardiac care,” said Sheth, professor and vice-chair of research, McNair Scholar and Cullen Foundation Endowed Chair of Neurosurgery at the Baylor College of Medicine. “In neurological and psychiatric disorders, the equivalent is deep brain stimulation (DBS), which sounds scary and invasive. DBS is actually quite a safe procedure, but it’s still brain surgery, and its perceived risk will place a very low ceiling on the number of people who are willing to accept it and may benefit from it. Here’s where technologies like this come in. A 30-minute minor procedure that is little more than skin surgery, done in an outpatient surgery center, is much more likely to be tolerated than DBS. So if we can show that it is about as effective as more invasive alternatives, this therapy will likely make a much larger impact on mental health.”

For some conditions, epilepsy for example, the device may need to be on permanently or most of the time, but for disorders such as depression and OCD, a regimen of just a few minutes of stimulation per day could suffice to bring about the desired changes in the functioning of the targeted neuronal network.

In terms of next steps, Robinson said that on the research side he is “really interested in the idea of creating networks of implants and creating implants that can stimulate and record, so that they can provide adaptive personalized therapies based on your own brain signatures.” From the therapeutic development standpoint, Motif Neurotech is in the process of seeking FDA approval for a long-term clinical trial in humans. Patients and caregivers can sign up on the Motif Neurotech website to learn when and where these trials will begin. The work was supported in part by The Robert and Janice McNair Foundation, the McNair Medical Institute, DARPA and the National Science Foundation.

Miniature battery-free epidural cortical stimulators | Science Advances | DOI: 10.1126/sciadv.adn0858 Authors: Joshua Woods, Amanda Singer, Fatima Alrashdan, Wendy Tan, Chufeng Tan, Sunil Sheth, Sameer Sheth and Jacob Robinson https://doi.org/10.1126/sciadv.adn0858

https://youtu.be/jhAEpAJGKSE (Video by Brandon Martin/Rice University)

https://news-network.rice.edu/news/files/2024/04/240213_Implant_Fitlow_152-5b1ea18ebb9fdfc8.jpg CAPTION: Rice University’s Jacob Robinson and his team of researchers have developed the smallest implantable brain stimulator demonstrated in a human patient that could revolutionize treatment for drug-resistant depression and other psychiatric or neurological disorders. (Photo by Jeff Fitlow/Rice University) https://news-network.rice.edu/news/files/2024/04/JWF_8628-39b4feadc43c1c91.jpg CAPTION: Joshua Woods (from left), Jacob Robinson and Fatima Alrashdan (Photo by Jeff Fitlow/Rice University) https://news-network.rice.edu/news/files/2024/04/JWF_8429-88e28c285ed94f0d.jpg CAPTION: Rice University engineers have developed the first miniaturized brain stimulator shown to work in a human patient. (Photo by Jeff Fitlow/Rice University) Box Link: https://rice.box.com/s/il8b8lxzkekxxodzb1peqpjwcpv97a3c

Robinson lab: www.robinsonlab.com

Sheth lab: https://www.bcm.edu/research/faculty-labs/functional-and-cognitive-neurophysiology-laboratory

Motif Neurotech: www.motifneuro.tech

Rice Neuroengineering Initiative: neuroengineering.rice.edu

Rice Department of Electrical and Computer Engineering: eceweb.rice.edu Rice Department of Bioengineering: https://bioengineering.rice.edu/

George R. Brown School of Engineering: engineering.rice.edu