Journal of Biomedical Engineering Research (대한의용생체공학회:의공학회지)

The Korean Society of Medical and Biological Engineering (대한의용생체공학회)
권호 표지
DOI QR코드

DOI QR Code

Development of an Explainable Multiclass Cognitive Impairment Classification Model Using Standardized Clinical Cognitive Assessments

표준화된 임상 인지 평가 지표를 이용한 설명 가능한 인지장애 다중분류 모델 개발

  • Moon-Hyeok Choi (Department of Electronic and Communication Engineering, Catholic Kwandog University) ;
  • Jea-Hyun Jo (Catholic Kwandong University Department of Digital Healthcare ) ;
  • Jin-Hyoung Jeong (Department of Medical Management, Catholic Kwandong University)
  • 최문혁 (가톨릭관동대학교 전자통신공학과) ;
  • 조재현 (가톨릭관동대학교 디지털헬스케어학과) ;
  • 정진형 (가톨릭관동대학교 의료경영학과)
  • Received : 2025.12.18
  • Accepted : 2026.02.02
  • Published : 2026.02.28
Download PDF
⟨ Previous Next ⟩

Abstract

This study aimed to propose a multiclass cognitive impairment classification framework based on conventional machine learning models using widely adopted clinical cognitive assessment tools, namely the Mini-Mental State Examination (MMSE) and the Clinical Dementia Rating (CDR), and to explore digital biomarkers through explainable artificial intelligence (XAI) techniques. The analysis was conducted using the publicly available Well-documented Alzheimer's Dataset from Kaggle, comprising a total of 436 participants (253 normal, 108 with mild cognitive impairment [MCI], and 75 with dementia). Feature vectors were constructed using age, sex, MMSE scores, and derived age-cognition interaction features. A comparative evaluation of Random Forest, Support Vector Machine (SVM), and XGBoost models under identical conditions demonstrated that the Random Forest model achieved the highest performance, with an accuracy of 0.95 (95% CI: 0.904-0.996). The model also exhibited consistently stable classification performance across ROC-AUC and confusion matrix analyses. Furthermore, SHAP-based explainable analysis revealed that MMSE, MMSE_Age_Ratio, age, and the Age×MMSE interaction were key contributors to cognitive stage prediction. These features effectively captured the progressive nature of cognitive decline and demonstrated strong potential as interpretable and quantitative digital biomarkers. Beyond performance improvement, this study is significant in that it presents a cognitive impairment classification model that simultaneously ensures high predictive accuracy and clinical interpretability.

Keywords

Acknowledgement

본 과제(결과물)는 2025년도 교육부 및 강원특별자치도의 재원으로 강원 RISE센터의 지원을 받아 수행된 지역혁신중심 대학지원체계(RISE)의 결과입니다(2025-RISE-10-001).

References

  1. Fowler NR, Partrick KA, Taylor J, Hornbecker M, Kelleher K, Boustani M, Cummings JL, MacLeod T, Mielke MM, Brosch JR, Lee J, Shobin E, Galvin JE, Fillit H, Udeh-Momoh C, Willis DR. Implementing early detection of cognitive impairment in primary care to improve care for older adults. J Intern Med. 2025;298(1):31-45. https://doi.org/10.1111/joim.v298.1
  2. Morris JC. Mild cognitive impairment (MCI) represents earlystage Alzheimer disease. JAMA Neurol. 2001;58(3):487-493.
  3. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology. 1993;43(11):2412-2414. https://doi.org/10.1212/WNL.43.11.2412-a
  4. Tombaugh TN, McIntyre NJ. The Mini-Mental State Examination: a comprehensive review. J Am Geriatr Soc. 1992; 40(9):922-935. https://doi.org/10.1111/jgs.1992.40.issue-9
  5. Grassi M, Rouleaux N, Caldirola D, Loewenstein DA, Schruers K, Perna G. Machine learning approaches to identify cognitive impairment: A systematic review. Front Aging Neurosci. 2018;10:1-15. https://doi.org/10.3389/fneur.2019.00756
  6. Sarica A, Cerasa A, Quattrone A. Random forest algorithm for the classification of neuroimaging data in Alzheimer's disease: A systematic review. Front Aging Neurosci. 2017;9:329. https://doi.org/10.3389/fnagi.2017.00329
  7. Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Adv Neural Inf Process Syst. 2017;30: 4765-4774.
  8. Tjoa E, Guan C. A survey on explainable artificial intelligence (XAI): Toward medical XAI. IEEE Trans Neural Netw Learn Syst. 2021;32(11):4793-4813. https://doi.org/10.1109/TNNLS.2020.3027314
  9. Soladoye AA, Aderinto N, Osho D, Olawade DB. Explainable machine learning models for early Alzheimer's disease detection using a comprehensive Kaggle Alzheimer's dataset. Neurocomputing. 2025;525:127549.
  10. Perneczky R, Wagenpfeil S, Komossa K, Grimmer T, Diehl J, Kurz A. Mapping scores onto stages: Mini-Mental State Examination and Clinical Dementia Rating. Am J Geriatr Psychiatry. 2006;14(2):139-144. https://doi.org/10.1097/01.JGP.0000192478.82189.a8
  11. Olatunde OO, Oyetunde KS, Han J, Khasawneh MT, Yoon H. Multiclass classification of Alzheimer's disease clinical stages using ensemble manifold regularization. Neuropsychologia. 2024;182:108957.
  12. Sokolova M, Lapalme G. A systematic analysis of performance measures for classification tasks. Inf Process Manage. 2009;45(4):427-437. https://doi.org/10.1016/j.ipm.2009.03.002
  13. Lima MR, Capstick A, Geranmayeh F, Nilforooshan R, Matarić M, Vaidyanathan R, Barnaghi P. Explainable machine learning models for ADRD screening and severity prediction: evaluation with 95% confidence intervals. arXiv 2025;2501.18731.
  14. Jung Y, Park Y, Jo J, Jeong J. MMSE-Based Dementia Prediction: Deep vs. Traditional Machine Learning Models. Life (Basel). 2025;15(10):1544.
  15. Petersen RC. Mild cognitive impairment. Continuum (Minneap Minn). 2016;22(2):404-418. https://doi.org/10.1212/CON.0000000000000313
  16. Grundman M, Petersen RC, Ferris SH, Thomas RG, Aisen PS, Bennett DA, Foster NL, Jack CR, Galasko D, Doody R, Kaye J, Sano M, Mohs R, Gauthier S, Kim HT, Jin S, Schultz AN, Schafer K, Mulnard R, Van Dyck CH. Mild cognitive impairment can be distinguished from Alzheimer disease and normal aging for clinical trials. Arch Neurol. 2004;61(1):59-66. https://doi.org/10.1001/archneur.61.1.59
  17. Yi F, Yu H, Yang H, Chen D, Qin Y, Han H, Cui J, Bai W, Ma Y, Zhang R. XGBoost-SHAP-based interpretable diagnostic framework for Alzheimer's disease multiclassification. BMC Medical Informatics and Decision Making. 2023;23:137. https://doi.org/10.1186/s12911-023-02238-9
  18. Jo T, Nho K, Saykin AJ. Machine learning approaches in Alzheimer's disease neuroimaging: A survey. NeuroImage: Clinical. 2019; 21:101-112.
  19. Holzinger A, Langs G, Denk H, Zatloukal K, Müller H. Causability and explainability of artificial intelligence in medicine. Wiley Interdiscip Rev Data Min Knowl Discov. 2019;9(4): e1312. https://doi.org/10.1002/widm.2019.9.issue-4
  20. Pellegrini E, Ballerini L, Valdés Hernández MdC, Chappell FM, González-Castro V, Anblagan D, Danso S, Muñoz-Maniega S, Job DE, Pernet C, Mair G, MacGillivray T, Trucco E, Wardlaw J. Machine learning of neuroimaging for assisted diagnosis of cognitive impairment and dementia: A systematic review. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring. 2018;10:519-535. https://doi.org/10.1016/j.dadm.2018.07.004
  21. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. Chapman & Hall/CRC; 1994.
  22. Varoquaux G, Cheplygina V. Machine learning for medical imaging: methodological failures and recommendations for the future. NPJ Digit Med. 2022;5:48. https://doi.org/10.1038/s41746-022-00592-y
  23. Lipton ZC. The mythos of model interpretability. Queue. 2018;16(3):31-57. https://doi.org/10.1145/3236386.3241340
  24. Crum RM, Anthony JC, Bassett SS, Folstein MF. Population-based norms for the Mini-Mental State Examination by age and educational level. JAMA. 1993;269(18):2386-2391. https://doi.org/10.1001/jama.1993.03500180078038
  25. Sperling RA, Aisen PS, Beckett LA, Bennett DC, Jack CR Jr, Morris JC, Petersen RC, Klunk WE, Green RC, Hyman BT. Toward defining the preclinical stages of Alzheimer's disease: Recommendations from the National Institute on Aging–Alzheimer's Association workgroups. Alzheimers Dement. 2011;7(3):280-292.
  26. Yi F, Yu H, Yang H, Chen D, Qin Y, Han H, et al. XGBoostSHAP-based interpretable diagnostic framework for Alzheimer's disease multiclassification. BMC Med Inform Decis Mak. 2023;23:137. https://doi.org/10.1186/s12911-023-02238-9
  27. Ko JH, Oh S, Urtnasan E, Key J. Smartwatch-based ensemble learning for depression prediction and feature analysis using explainable artificial intelligence. J Biomed Eng Res. 2024;45(6):379-390.