CT-based prediction of hematoma expansion and adverse outcomes after intracerebral hemorrhage: evidence appraisal, artificial intelligence translation, and GeroScience perspectives.
Authors
Affiliations (4)
Affiliations (4)
- School of Medical Science and Engineering, Beijing Institute of Technology, Beijing, 100081, China.
- Zhengzhou Academy of Intelligent Technology, Beijing Institute of Technology, Beijing, 450003, China.
- School of Medical Science and Engineering, Beijing Institute of Technology, Beijing, 100081, China. [email protected].
- Zhengzhou Academy of Intelligent Technology, Beijing Institute of Technology, Beijing, 450003, China. [email protected].
Abstract
Spontaneous intracerebral hemorrhage (ICH) is a highly lethal and disabling form of stroke, in which hematoma expansion (HE) is a major and potentially modifiable determinant of early neurological deterioration and poor functional outcome. Computed tomography (CT) remains the first-line imaging modality for acute ICH and provides essential information for early HE risk stratification. However, current evidence is dispersed across conventional CT signs, composite scores, radiomics, machine learning, and deep learning approaches, many of which have been reported descriptively without sufficient comparison of clinical utility, validation quality, or translational readiness. This review critically evaluates CT-based prediction of HE and adverse outcomes after spontaneous ICH. We compare contrast-enhanced markers, including the spot sign, leakage sign, and iodine sign, with non-contrast CT markers such as the blend sign, black hole sign, island sign, satellite sign, hypodensity sign, swirl sign, hematoma shape, and density heterogeneity, focusing on sensitivity, specificity, reproducibility, availability, and clinical applicability. We further assess composite prediction models and artificial intelligence approaches, emphasizing limitations related to small cohorts, overfitting, dataset heterogeneity, insufficient external validation, interpretability, and workflow integration. Given the scope of GeroScience, we also discuss how vascular aging, cerebral amyloid angiopathy, frailty, anticoagulant exposure, and age-associated vulnerability to secondary injury may influence HE risk and outcome prediction. Future progress will require interpretable, multimodal, prospectively validated models that integrate CT imaging, clinical variables, biomarkers, and aging-related factors to support individualized management of ICH.