Answers are grounded in this paper and cite their sources.
This study aimed to construct an effective model based on radiomics features and machine learning (ML) algorithms to predict hematoma expansion (HE) in spontaneous hypertensive intracerebral hemorrhage (SHICH) patients. This retrospective study included 85 patients with SHICH, of whom 64 had stable hematoma volume and 21 had HE. Initial non-contrast computed tomography scans were performed, and follow-up scans were conducted 24 hours later. Hematoma areas were segmented, and radiomics features were extracted using the PyRadiomics package. Feature selection was performed using p -value calculations and minimal redundancy and maximum relevance (mRMR) mehtod. The top 13 significant features were used to construct classification models using Naive bayes (NB), linear discriminant analysis (LDA), support vector machine (SVM), K-nearest neighbors (KNN), and random forest (RF) algorithms. Three ensemble methods combined the classifiers’ results. Sensitivity, specificity, and accuracy were evaluated for performance assessment. Among the classifiers, RF demonstrated the best performance with mean accuracy, sensitivity and specificity of 87.12%, 83.33% and 88.42%, respectively. The Stacking ensemble method improved overall classification accuracy to 98.4%, with a sensitivity of 93.33% and specificity of 100%. The proposed radiomics-based model combined with ML algorithms, particularly through the use of the stacking ensemble, significantly improves the prediction of HE in SHICH patients. This method holds potential for early identification and timely intervention, thereby improving patient outcomes.
No takes yet. Share an insight, caveat, or question.
Azizi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: