Noninvasive model for predicting future ischemic strokes in patients with silent lacunar infarction using radiomics

Background: This study aimed to investigate integrating radiomics with clinical factors in cranial computed tomography (CT) to predict ischemic strokes in patients with silent lacunar infarction (SLI). Methods: Radiomic features were extracted from baseline cranial CT images of patients with SLI. A least absolute shrinkage and selection operator (LASSO)-Cox regression analysis was used to select significant prognostic factors based on ModelC with clinical factors, ModelR with radiomic features, and ModelCR with both factors. The Kaplan-Meier method was used to compare stroke-free survival probabilities. A nomogram and a calibration curve were used for further evaluation. Results: Radiomic signature (p < 0.01), age (p = 0.09), dyslipidemia (p = 0.03), and multiple infarctions (p = 0.02) were independently associated with future ischemic strokes. ModelCR had the best accuracy with 6-, 12-, and 18-month areas under the curve of 0.84, 0.81, and 0.79 for the training cohort and 0.79, 0.88, and 0.75 for the validation cohort, respectively. Patients with a ModelCR score < 0.17 had higher probabilities of stroke-free survival. The prognostic nomogram and calibration curves of the training and validation cohorts showed acceptable discrimination and calibration capabilities (concordance index [95% confidence interval]: 0.7864 [0.70-0.86]; 0.7140 [0.59-0.83], respectively). Conclusions: Radiomic analysis based on baseline CT images may provide a novel approach for predicting future ischemic strokes in patients with SLI. Older patients and those with dyslipidemia or multiple infarctions are at higher risk for ischemic stroke and require close monitoring and intensive intervention.