Deep learning-based automated segmentation of intracerebral haemorrhage, intraventricular haemorrhage and perihaematomal oedema on non-contrast CT

Key Points

• The research aims to evaluate a deep learning model for automating the segmentation of brain hemorrhages and related edema on CT images.
• Developed a deep learning model for image analysis
• Applied the model to non-contrast CT scans
• Evaluated segmentation accuracy for intracerebral hemorrhage, intraventricular hemorrhage, and perihaematomal edema
• Achieved accurate segmentation for intracerebral hemorrhage
• Segmentation of intraventricular hemorrhage and perihaematomal edema remains challenging
• Automated segmentations can assist in the research phase but need visual verification for accuracy.

Abstract