The Role of Artificial Intelligence and Machine Learning in the Assessment, Diagnosis, and Prediction of Cerebral Small Vessel Disease

Abstract

Cerebral small vessel disease (CSVD) contributes substantially to ischemic stroke and vascular cognitive impairment but remains difficult to detect with conventional diagnostics. Recent advances in artificial intelligence (AI), including machine learning (ML) and deep learning (DL), have improved neuroimaging analysis, early risk stratification, and clinical decision support in CSVD-related stroke, while raising questions about generalizability, interpretability, and ethics.

This review aims to narratively synthesize how AI supports neuroimaging analysis, early detection, clinical decision-making, and prognostication in stroke with an emphasis on CSVD, and to summarize limitations, bias, and implementation challenges.

This narrative review synthesized evidence from 122 studies. AI showed strong performance across stroke care with an emphasis on CSVD: intracerebral hemorrhage (ICH) detection on noncontrast CT (sensitivity = 93%, specificity = 92%); 18-25-minute reductions in door-to-needle time; superior prediction of 90-day disability versus clinician assessment (89% vs. 72%); reduced inter-rater variability for white matter hyperintensities (WMHs) segmentation; ~94% accuracy for enlarged perivascular spaces (EPVS) classification on MRI; and faster team notification and time-to-treatment, with mixed evidence for improved 90-day functional independence. However, performance was weaker in older and diabetic cohorts, underscoring limited generalizability, scarce prospective validation, and risks of bias.

AI augments stroke care across imaging-based diagnosis, risk stratification, and rehabilitation, with growing utility in CSVD. Translation into routine care requires robust external validation, bias mitigation, model interpretability, and clear governance around safety, liability, and cost.