Quantitative Shape Irregularity and Density Heterogeneity Predict Hematoma Expansion in Patients With Intracerebral Hemorrhage.

Purpose: This study aimed to explore the association between quantitative shape irregularity and density heterogeneity of hematomas and hematoma expansion (HE) for intracerebral hemorrhage (ICH) patients.

Methods: This cohort study included patients arriving within 24 h of symptom onset between August 2021 and July 2022 as the derivation cohort and those between July 2023 and February 2024 as the external validation cohort. HE is defined as a hematoma increase of > 6 mL or > 33% from the baseline to the follow-up CT scan between 24 and 48 h. The least absolute shrinkage and selection operator (LASSO) regression was applied to select the traditional image signs to fit the logistic regression as Model 1. Afterwards, the surface regularity index (SRI) and density coefficient of variation (DCV) of hematoma were added to form Model 2. Finally, we used the SRI and DCV to replace the selected traditional image signs as Model 3. The performance and clinical utilities were evaluated and compared in the external validation cohort.

Result: The three models demonstrated good discrimination in both the derivation cohort and the validation cohort, with Model 2 and Model 3 showing significant improvements in area under the receiver operating characteristic curve (AUROC) and in clinical utility compared to Model 1 (Model 2 AUROC: 0.859 [95% CI: 0.802, 0.926] vs. Model 1 AUROC: 0.713 [95% CI: 0.625, 0.814], Delong test p < 0.001; Model 3 AUROC: 0.840 [95% CI: 0.776, 0.912] vs. Model 1 AUROC: 0.713 [95% CI: 0.625, 0.814], p = 0.006). The SRI and DCV can improve the prediction of HE based on traditional clinical indicators and imaging signs, also serving as possible alternatives to traditional imaging signs.

Conclusions: The SRI and DCV can serve as effective substitutes for traditional imaging signs in predicting hematoma expansion.