# Improving Risk Stratification for Transient Ischaemic Attacks and Ischaemic Stroke in Patients with Coronary Artery Disease: A Combined Radiomics Analysis of Multimodal Adipose Tissue

## TL;DR

This study improves risk prediction for brain-related issues in heart disease patients by combining imaging and clinical data.

## Contribution

A novel combined radiomics model using coronary and cervical CTA data enhances risk stratification for cerebral ischaemic events in CAD patients.

## Key findings

- Higher prevalence of high-risk plaque in coronary and cervical arteries was observed in patients with cerebral ischaemia.
- RCA FAI and PFD were confirmed as independent risk factors for ischaemic events.
- Model 5, integrating clinical and imaging data, achieved the highest predictive accuracy (AUC: 0.821).

## Abstract

Background/Objectives: Patients with combined cardiovascular and cerebrovascular disease face poorer prognoses. Early, accurate assessment of the risk of cerebral ischaemic events (including transient ischaemic attacks (TIAs) and ischaemic strokes (ISs)) in patients with coronary artery disease (CAD) is therefore vital for clinical guidance. This study aims to develop a comprehensive risk assessment model for early warning in this population. Methods: In this study, we conducted a retrospective multicentre recruitment of CAD patients undergoing concurrent coronary CTA and cervical CTA (n = 326), with follow-up to observe the occurrence of cerebral ischaemic events. We performed an analysis of high-risk plaque (HRP) characteristics and subcomponent plaque in coronary and cervical arteries, measured the pericoronary fat attenuation index (FAI) and cervical perivascular fat density (PFD), and extracted corresponding radiomic features. Five models were constructed to identify the CAD patients who developed IS/TIA, respectively: Model 1—clinical characteristics; Model 2—coronary CTA parameters + Radscorecoronary; Model 3—cervical CTA parameters + Radscorecervical; Model 4—Model 1 + Model 2; Model 5—Model 1 + Model 2 + Model 3. Results: In the cerebral ischaemia group, the prevalence of coronary and/or cervical HRP was higher than in the non-ischaemia group (28.0% vs. 26.1%, 57.0% vs. 44.0%, p = 0.02). Multivariate logistic regression confirmed that RCA FAI and PFD remained significant independent risk factors for IS/TIA (all p < 0.05). The model prediction results showed that progressively incorporating coronary and cerebral vascular risk factors into the clinical features gradually improved model performance (Model 4 vs. Model 5, AUC: 0.711 [0.645–0.777] vs. 0.821 [0.769–0.873]). Model 5 achieved a sensitivity of 0.788 [0.485–0.909] and specificity of 0.827 [0.385–0.923], demonstrating the best overall clinical benefit. Conclusions: RCA FAI and PFD are independent predictors of cerebral ischaemic events. By integrating clinical characteristics, coronary CTA and cervical CTA parameters, combined with Radscorecoronary and Radscorecervical, the risk stratification capability for IS/TIA in CAD patients can be significantly enhanced.