Machine learning applications in vascular neuroimaging for the diagnosis and prognosis of cognitive impairment and dementia: a systematic review and meta-analysis

Introduction Machine learning (ML) algorithms using neuroimaging markers of cerebral small vessel disease (CSVD) are a promising approach for classifying cognitive impairment and dementia. Methods We systematically reviewed and meta-analysed studies that leveraged CSVD features for ML-based diagnosis and/or prognosis of cognitive impairment and dementia. Results We identified 75 relevant studies: 43 on diagnosis, 27 on prognosis, and 5 on both. CSVD markers are becoming important in ML-based classifications of neurodegenerative diseases, mainly Alzheimer’s dementia, with nearly 60% of studies published in the last two years. Regression and support vector machine techniques were more common than other approaches such as ensemble and deep-learning algorithms. ML-based classification performed well for both Alzheimer’s dementia (AUC 0.88 [95%-CI 0.85–0.92]) and cognitive impairment (AUC 0.84 [95%-CI 0.74–0.95]). Of 75 studies, only 16 were suitable for meta-analysis, only 11 used multiple datasets for training and validation, and six lacked clear definitions of diagnostic criteria. Discussion ML-based models using CSVD neuroimaging markers perform well in classifying cognitive impairment and dementia. However, challenges in inconsistent reporting, limited generalisability, and potential biases hinder adoption. Our targeted recommendations provide a roadmap to accelerate the integration of ML into clinical practice.