CT Radiomics Model Predicts 3-Month Poor Outcome After Basal Ganglia Hemorrhage

A CT radiomics-based random forest model predicted poor 3-month functional outcome (mRS ≥4) in conservatively managed basal ganglia hemorrhage with an AUC of 0.867 (95% CI 0.797–0.929) after bootstrap correction. Key predictors: admission GCS and wavelet radiomic features. Exter…

• Retrospective study of 254 patients with conservatively managed basal ganglia hemorrhage; 14 features (mostly wavelet radiomic) selected via LASSO.
• Random forest outperformed nine other classifiers including a SuperLearner ensemble on discrimination and calibration.
• SHAP analysis identified admission GCS and wavelet radiomic markers as most influential; model not yet ready for clinical use.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.