READi-Dem: ML-powered, web-interface tool for Robust, Efficient, Affordable Diagnosis of Dementia

• DOI: 10.1101/2023.10.23.23297405
• Publication Date: 2023-10-24T09:50:18Z
• Authors: Verena Klar; Elinor Thompson; Melvin Selim Atay; Peter Owoade; Sofia Toniolo; Amir Dehsarvi; Sanjay Rathee
• Keywords: Clinical Dementia Rating; Healthy ageing

Abstract Background Dementia screening tools typically involve face-to-face cognitive testing. Indeed, this introduces an increasing burden on the clinical staff, particularly in low-resource settings. The objective of our study is to develop integrated online platform for efficient dementia screening, using a brief and cost-effective assessment. Methods We used Longitudinal Ageing Study India dataset (LASI-DAD, n=2528) predict diagnosis based Clinical Rating (CDR). Using feature selection algorithms principal component analysis (PCA), we identified key predictive features. compared performance six machine learning (ML) classifiers that were trained 42 selected features (full model) two components by PCA (minimal model). best-performing model was web platform. Results Selected mapped onto distinct, interpretable domains: domain informant domain. first cumulatively explained 90.2% variance included questions from Mini-Mental State Exam (MMSE) Informant Questionnaire Cognitive Decline Elderly (IQCODE). Classifiers minimal performed par with full model, Support Vector Machine performing best (93.4%). did not reliably Parkinson’s disease (67% accuracy) or stroke (53.1% accuracy), suggesting specificity. respective MMSE IQCODE (27 items) incorporated into Conclusion built enabling end-to-end assessment prediction, patient caregiver reports. Web App code available at GitHub: https://github.com/sanjaysinghrathi/READi-Dem & link Page: https://researchmind.co.uk/readi-dem . For convenience researchers, video summarizing work also accessible Page YouTube Link