Prediction of etiology and prognosis based on hematoma location of spontaneous intracerebral hemorrhage

Abstract

Background The characteristics of the hemorrhagic location of spontaneous intracerebral hemorrhage (sICH) is pivotal for both identifying its etiology and prognosis. While empirical conclusions have been obtained in clinical practice, a comprehensive and quantitative modeling approach has yet to be thoroughly explored. Methods We employed lesion-symptom mapping to extract the location features of sICH. We registered patients' non-contrast computed tomography image and hematoma masks with standard human brain templates to identify specific affected brain regions. Then, we generated hemorrhage probabilistic maps of different etiologies and prognoses. By integrating radiomics and clinical features into multiple logistic regression models, we developed and validated optimal etiological and prognostic models across three centers, comprising 1162 sICH patients. Results Hematomas of different etiology have unique spatial distributions. Location features robustly categorized the etiology sICH (mean area under the curve (AUC) = 0.825) across different datasets), yielding clear add on value in models (fusion model mean AUC = 0.915) compared to clinical features (mean AUC = 0.828). In the prognostic analysis, patients with more extensive hematoma had a worse prognosis, the location (mean AUC = 0.762) and radiomic features (mean AUC = 0.837) also providing reliable add on value effects (fusion model mean AUC = 0.873) compared to clinical features alone (mean AUC = 0.771). Interpretation Our results show that location features were more intrinsically robust, generalizable relative, strong interpretability to the complex modeling of radiomics, our approach demonstrated a novel interpretable, streamlined, comprehensive etiologic classification and prognostic prediction framework for sICH.