Machine Learning-Enhanced TCAB Score for Predicting Postoperative Ischemic Stroke After CABG.

BACKGROUND

Postoperative acute ischemic stroke remains a critical complication of coronary artery bypass grafting. This study aimed to develop a novel Total Cerebral Atherosclerosis Burden (TCAB) score for predicting the risk of AIS post-coronary artery bypass grafting.

METHODS

A prospective cohort of patients undergoing coronary artery bypass grafting was enrolled. The TCAB score was calculated by summing stenosis severity grades (0: <50%, 1: 50-69%, 2: 70-99%, 3: 100%) across all intracranial and extracranial artery segments. Primary outcome was in-hospital ischemic stroke. Multivariable logistic regression models adjusted for key clinical covariates were used to evaluate the association between TCAB and clinical outcomes.

RESULTS

Among 909 included patients, the mean TCAB score was significantly higher in patients with in-hospital ischemic stroke compared with those without (8 versus 2, P < 0.001). A TCAB score >3 predicted in-hospital ischemic stroke with an area under the curve of 0.756. Across all 3 multivariable models, higher TCAB scores remained independently associated with in-hospital ischemic stroke (Model 3: odds ratio [OR]=1.089, P=0.011), 1-year ischemic stroke (OR=1.093, P=0.011), and 1-year major adverse cardiovascular and cerebrovascular events (OR=1.068, P=0.020). The gradient boosting machine achieved the most stable predictive ability (area under the curve=0.8736 for in-hospital ischemic stroke; 0.8575 for 1-year ischemic stroke; 0.7475 for 1-year major adverse cardiovascular and cerebrovascular events).

CONCLUSIONS

The TCAB score, enhanced by machine learning, effectively predicted in-hospital ischemic stroke, 1-year ischemic stroke, and 1-year major adverse cardiovascular and cerebrovascular events post-coronary artery bypass grafting. It offers a practical tool for guiding preoperative revascularization and intraoperative embolic protection.