CT Radiomics Model Predicts Outcomes After Aneurysmal Subarachnoid Hemorrhage

Prognostic assessment after aneurysmal subarachnoid hemorrhage (aSAH) commonly uses clinical variables and CT findings. In this multicenter cohort, researchers analyzed 879 patients from four hospitals in the PROSAH-MPC registry. They developed XGBoost models using clinical data, CT radiomic features, or both. The primary-center cohort was divided into training (n=637) and internal validation (n=160) groups; 82 patients from three other centers formed the external validation cohort. The study also examined CSF proteins in a separate prospective cohort, grouping patients according to the combined model’s risk scores.

The combined clinical-radiomics model had an AUC of 0.892 in the internal test cohort and 0.820 in external validation. Risk reclassification analyses suggested that radiomic features added prognostic information. SHAP interpretation highlighted radiomic heterogeneity and clinical variables as important predictors. Exploratory proteomic analyses found candidate protein and pathway differences between high- and low-risk groups.

The findings support further evaluation of CT radiomics as a risk-stratification approach, but do not establish that using the model improves patient outcomes or should guide care. The CSF analyses are exploratory, and the external validation cohort was relatively small. The study was retrospectively registered and reports no competing interests.