What question did this study set out to answer?

The study aims to evaluate clot composition in ischemic stroke using radiomics from non-contrast CT images.

January 18, 2026Open Access

Clot Composition Profiling in Large Vessel Occlusion Stroke Via Radiomics

Key Points

The study aims to evaluate clot composition in ischemic stroke using radiomics from non-contrast CT images.
Conducted a prospective study comparing radiomic features with histopathology of thrombi from mechanical thrombectomy.
Thrombi were analyzed using micro-CT and matched with histological slices for composition analysis.
Extracted radiomic features were assessed for associations with composition using multivariate logistic regression and Spearman’s correlation.
Performed receiver operating characteristic analysis on NCCT radiomic features to evaluate discrimination of clot composition.
Applied radiomics to a retrospective NCCT dataset involving different stroke etiologies.
Analyzed 10 thrombi using micro-CT, finding significant associations between total energy and RBC composition (OR = 1.35).
Radiomic features correlated well with RBC composition, indicating effectiveness in identifying clot components.
In the retrospective analysis of 150 patients, 72% of cardioembolic strokes showed RBC-predominant composition compared to 30% of LAA and 50% of cryptogenic strokes.

Abstract

Objective Clot composition may offer insights into the mechanism of ischemic stroke. Radiomics, a noninvasive imaging technique, enables tissue characterization through radiomic features (RFs). We aimed to evaluate clot composition using radiomics on non‐contrast computed tomography (NCCT). Methods In the first phase, we conducted a prospective study comparing RFs with histopathology in thrombi retrieved via mechanical thrombectomy (MT). Thrombi were imaged using micro‐computed tomography (micro‐CT) and analyzed histologically. Matched micro‐CT and histological slices identified red blood cells (RBCs) and fibrin‐rich regions. RFs were extracted, and multivariate logistic regression identified features associated with each component. Spearman's correlation was used to assess associations between RFs and percentage composition. The same clots were localized on pre‐MT NCCT, and RFs were extracted. Micro‐CT and NCCT RFs were correlated to enable histology‐informed interpretation. Receiver operating characteristic analysis evaluated the ability of NCCT RFs to discriminate clot composition. In the second phase, radiomics was applied to a retrospective NCCT dataset from patients with ischemic stroke with varying etiologies. Results Ten thrombi were analyzed using micro‐CT. Total energy (odds ratio OR = 1.35, 95% confidence interval CI = 1.20–1.54, p 70% RBCs composition on NCCT (Rho = 0.752 and Rho = 0.815). Subsequently, 150 NCCT scans were analyzed, including 50 cardioembolic, 50 large artery atherosclerosis (LAA), and 50 cryptogenic strokes. Radiomic analysis indicated RBCs‐predominant composition in 72% of cardioembolic, 30% of LAA, and 50% of cryptogenic clots. Interpretation Radiomics is a promising, noninvasive technique for characterizing clot composition. ANN NEUROL 2026

Ask AI

Helpful

Bookmark

View Full Paper