Predicting incident dementia in cerebral small vessel disease: comparison of machine learning and traditional statistical models.

Change log
Harshfield, Eric L 
Bell, Steven 
Burkhart, Michael 
Tuladhar, Anil M 

BACKGROUND: Cerebral small vessel disease (SVD) contributes to 45% of dementia cases worldwide, yet we lack a reliable model for predicting dementia in SVD. Past attempts largely relied on traditional statistical approaches. Here, we investigated whether machine learning (ML) methods improved prediction of incident dementia in SVD from baseline SVD-related features over traditional statistical methods. METHODS: We included three cohorts with varying SVD severity (RUN DMC, n = 503; SCANS, n = 121; HARMONISATION, n = 265). Baseline demographics, vascular risk factors, cognitive scores, and magnetic resonance imaging (MRI) features of SVD were used for prediction. We conducted both survival analysis and classification analysis predicting 3-year dementia risk. For each analysis, several ML methods were evaluated against standard Cox or logistic regression. Finally, we compared the feature importance ranked by different models. RESULTS: We included 789 participants without missing data in the survival analysis, amongst whom 108 (13.7%) developed dementia during a median follow-up of 5.4 years. Excluding those censored before three years, we included 750 participants in the classification analysis, amongst whom 48 (6.4%) developed dementia by year 3. Comparing statistical and ML models, only regularised Cox/logistic regression outperformed their statistical counterparts overall, but not significantly so in survival analysis. Baseline cognition was highly predictive, and global cognition was the most important feature. CONCLUSIONS: When using baseline SVD-related features to predict dementia in SVD, the ML survival or classification models we evaluated brought little improvement over traditional statistical approaches. The benefits of ML should be evaluated with caution, especially given limited sample size and features.

Cerebral small vessel disease, Dementia, Machine learning, Prediction
Journal Title
Cereb Circ Cogn Behav
Conference Name
Journal ISSN
Volume Title
Elsevier BV
Alzheimer's Society (573 (AS-RF-21-017))
British Heart Foundation (RG/F/22/110052)
British Heart Foundation (RE/18/1/34212)
National Institute for Health and Care Research (IS-BRC-1215-20014)
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by a British Heart Foundation (BHF) programme grant [grant number RG/F/22/110052] and infrastructural support was provided by the Cambridge British Heart Foundation Centre of Research Excellence [grant number RE/18/1/34212] and the Cambridge University Hospitals NIHR Biomedical Research Centre [grant number BRC-1215-20014]. HSM is supported by an NIHR Senior Investigator Award, and a number of peer reviewed funders including Medical Research Council, EU, Alzheimer’s Society, Stroke Association, BHF. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. RL is supported by a PhD scholarship awarded by Trinity College, University of Cambridge. ELH is supported by Cambridge BHF Centre of Research Excellence [grant number RE/18/1/34212]; Alzheimer’s Society [grant number AS- RF-21-017]; BHF programme grant [grant number RG/F/22/110052]; Cambridge NIHR Biomedical Research Centre [grant number BRC-1215-20014]. SB is supported by BHF. AMT is supported by Dutch Heart Foundation [grant number 2016T044]. Wellcome Trust [grant number 081589] provided initial funding for SCANS study. SDJM is supported by Wellcome Trust [grant number WT088134/Z/09/A]. JMW is supported by Wellcome Trust, Row Fogo Trust, and Medical Research Council. CC is supported by National Medical Research Council of Singapore. ZK is supported by Wellcome Trust and Alan Turing Institute.