Stratifying vascular disease patients into homogeneous subgroups using machine learning and FLAIR MRI biomarkers

Abstract

This study proposes a framework to stratify vascular disease patients based on brain health and cerebrovascular disease (CVD) risk using regional FLAIR biomarkers. Intensity and texture biomarkers were extracted from FLAIR volumes of 379 atherosclerosis patients. K-Means clustering identified five homogeneous subgroups. The 15 most important biomarkers for subgroup differentiation, identified via Random Forest classification, were used to generate biomarker profiles. ANOVA tests showed age and white matter lesion volume were significantly (p < 0.05) different across subgroups, while Fisher's tests revealed significant (p < 0.05) differences in the prevalence of several vascular risk factors across subgroup. Based on biomarker and clinical profiles, Subgroup 4 was characterized with neurodegeneration unrelated to CVD, Subgroup 3 identified patients with high CVD risk requiring aggressive intervention, and Subgroups 1, 2, and 5 identified patients with varying levels of moderate risk, suitable for long-term lifestyle interventions. This study supports personalized treatment and risk stratification based on FLAIR biomarkers.

Keywords: Brain imaging; Image processing.

Fig. 1. Visuals of the ROIs in this work.

A Sample mean texture maps in ACA, MCA, and PCA BST regions. B Sample microstructural integrity maps of WM tract regions from samples slices of a volume, lower to higher slices from left to right and top to bottom. C Middle slice of original FLAIR volume (left), and WML penumbra regions P1 (green) to P5 (light pink) delineated on FLAIR NABM slices.

Fig. 2. Regional and composite FLAIR biomarker profiles for clustered subgroups.

A Z-scored FLAIR biomarker profiles for each subgroup of top 15 important biomarkers. Integrity biomarkers are shown in blue and Damage biomarkers are shown in red. B Composite Z-scored damage (right) and integrity (left) biomarkers by subgroup.

Fig. 3. Characterization of subgroups using z-scored demographic/clinical variables and CVD risk scores.

A Z-scored profiles of each subgroup.* indicate variables with significant differences (p < 0.05) between groups. B Compound CVD risk score shown by subgroup. High/low score thresholds are shown as dashed lines.

Fig. 4. Heatmap of R correlation coefficients between clinical variable and FLAIR biomarker.

* indicate significant correlations (p < 0.05).