Detection of cerebral small vessel disease in health examination populations using machine learning

Motivation: The diagnosis of cerebral small vessel disease (CSVD) primarily relies on magnetic resonance imaging (MRI); however, its relatively high cost poses a challenge for implementing CSVD screening in the general population, particularly in low- and middle-income countries. Goal(s): To detect CSVD in the general population using routine health examination data. Approach: We developed and validated a machine learning (ML) model within a novel framework, termed Risk Assessment of CSVD in the General Population (RACGP). Results: The LightGBM model based on RACGP achieved area under the curve (AUC) values of 0.862 on the test set and 0.789 on the external validation set. Impact: Our ML model can identify CSVD patients within health examination populations in a low-cost manner, showing potential for CSVD screening.