J Neurol. 2026 May 22;273(6):331. doi: 10.1007/s00415-026-13875-1.
ABSTRACT
BACKGROUND: Accurate prognostication in deep supratentorial intracerebral hemorrhage (ICH) remains challenging. Hematoma volume is a key determinant of outcome and a component of the ICH Score. Whether radiomics-derived shape features provide prognostic information beyond volumetric burden and clinical risk scores remains unclear. We investigated the association between radiomic shape characteristics of basal ganglia hemorrhage (BGH) and intraventricular hemorrhage (IVH) with in-hospital mortality and unfavorable functional outcome.
METHODS: We retrospectively analyzed 50 patients with deep supratentorial ICH and IVH. Semi-automated segmentations were used to extract radiomic features (roundness, elongation, flatness, Feret diameter, surface area) and volumetric parameters. Logistic regression models were constructed for in-hospital mortality and unfavorable functional outcome (mRS 5-6 at discharge; secondary definition mRS 4-6), with collinearity assessment guiding feature selection. Model discrimination was evaluated using receiver operating characteristic curves with DeLong comparisons. Secondary models incorporated the ICH Score. Calibration and internal validation were assessed using the Hosmer-Lemeshow test and bootstrap resampling (1000 iterations).
RESULTS: For in-hospital mortality, BGH roundness showed discrimination comparable to BGH volume (AUC 0.742 vs 0.738). In multivariable analysis, lower roundness remained independently associated with mortality after adjustment for the ICH Score (OR 0.20 per 0.1 increase; 95% CI 0.04-0.66; p = 0.021), whereas volume was no longer significant. The combined model achieved an AUC of 0.866 versus 0.813 for the ICH Score (DeLong p = 0.28). IVH-derived shape parameters were not associated. For unfavorable outcome, BGH volume remained strongest predictor.
CONCLUSIONS: Radiomic hemorrhage roundness may provide complementary prognostic information beyond hematoma volume and the ICH Score in deep ICH with IVH.
PMID:42171773 | DOI:10.1007/s00415-026-13875-1