Feasibility Exploration of High-Resolution MRI Plaque Features for Assessing Outcomes of Intracranial Angioplasty and Stenting in Ischemic Stroke Patients.

Abstract

Objective: To evaluate the feasibility of plaque-based radiomics extracted from high-resolution magnetic resonance imaging (HR-MRI) data for assessing the short-term outcomes of endovascular treatment in patients with symptomatic intracranial artery stenosis.

Methods: HR-MRI was performed on patients with symptomatic intracranial artery stenosis. Plaque-based radiomics describing the morphological features and pixel value of the image were extracted from the HR-MRI data. Demographic features were also collected. The short-term favorable outcome was defined by a postoperative residual stenosis rate <35% with the absence of perioperative complications. Univariate analysis was conducted to identify features associated with favorable outcomes. Based on the results of this analysis, a prediction model was developed using logistic regression. The performance of both clinical and radiomic models was evaluated using the receiver operating characteristic (ROC) curve and the area under the curve (AUC).

Results: From January 2022 to December 2023, 42 consecutive patients with symptomatic intracranial artery stenosis were enrolled. Digital subtraction angiography (DSA) revealed a more than 70% stenosis rate in these patients. The stents were implemented in all 42 patients; 21 (50%) of these were male, and the mean age of all patients was 52.74 ± 13.02 years. Thirty-five patients (83.33%) had impaired sensory or motor function of the limbs. In the univariate analysis, 11 morphologic or first-order radiomics features and five clinical features were initially identified as potentially associated with short-term favorable outcomes. Logistic multivariate analysis further indicated that shape-flatness (p = 0.04, Odd ratio (OR) = 169.02, 95% CI: 1.30-22,026.5) and first-order-minimum (p = 0.02, OR = 94.63, 95% CI: 1.93-4592.5) might be independently related to post-stenting outcomes. A prediction model constructed based on the above morphologic and first-order features showed an AUC of 0.82 in this small cohort.

Conclusion: Plaque-based radiomic features, which describe the shape and voxel characteristics extracted from HR-MRI data, are associated with the short-term outcomes of patients treated with stent implementation.