MCH-Guard: Multimodal Machine Learning Framework for Risk Stratification of Cerebral Microhemorrhage Risk in the Alzheimer's Disease Neuroimaging Initiative

MCH-Guard is a multimodal machine learning framework developed to stratify the risk of cerebral microhemorrhages (MCH) using data from the ADNI initiative in a sample of 813 patients. Nested models integrated clinical history, biomarkers, and imaging to predict MCH presence, incidence, and stability. The comprehensive model detected baseline MCH with high accuracy (AUC 0.86), while the minimal model (M1) using only demographics and clinical history achieved robust performance (AUC 0.82). Longitudinal models predicted time-to-onset (R2=0.68) and stratified four-year risk. The research identified a transient vascular instability phenotype where MCH status fluctuates, which was strongly predicted by hepatic factors. MCH-Guard provides a flexible clinical decision-support tool for optimizing MCH and ARIA-H monitoring, with the performance of the clinical-only M1 model supporting equitable risk assessment in resource-limited settings.