Objective:

To outline the evolution of imaging-based prediction methods for hematoma expansion (HE) following acute intracerebral hemorrhage (ICH).

Approach:

• Core Concepts: Defines traditional HE, revised HE (rHE), and ultra-early hematoma growth (uHG).
• Traditional Imaging Markers: Summarizes studies using traditional imaging markers like CTA spot sign and NCCT signs.
• AI-Driven Methodologies: Focuses on radiomics, deep learning, and multi-task learning for HE prediction.
• Multimodal Data Fusion: Discusses precision prediction through the fusion of clinical, laboratory, and imaging data.
• Challenges and Future Directions: Addresses challenges in model interpretability, generalizability, and suggests future research directions.

Key Findings:

• Approximately 20-30% of patients experience HE within 24 hours post-ICH.
• Revised HE (rHE) improves predictive sensitivity for poor outcomes by incorporating intraventricular hemorrhage growth.
• Ultra-early hematoma growth (uHG) provides a dynamic assessment for risk stratification based on the rate of bleeding.

Interpretation:

The review discusses the transition from traditional imaging methods to advanced AI-driven approaches for predicting hematoma expansion.

Limitations:

• Challenges in model generalizability and interpretability.
• Need for standardization in clinical translation.
• Persistent translational gaps in applying predictive models.

Conclusion:

This review provides a structured framework for understanding imaging advancements in predicting hematoma expansion.