Synthetic training improves detection of rare brain hemorrhages
Synthetic training for long-tail haemorrhagic lesion segmentation in data-scarce settings
Computer Vision and Pattern Recognition
Summary
Finding and marking small brain bleeds from medical images is hard because there are very few examples to learn from. The authors created a way to make realistic fake images with these bleeds based on what doctors know about their shapes and locations. They trained computer models using these fake images and got better results than older methods when tested on real patient scans. This approach could help computers spot rare brain bleeds more accurately when real data is scarce.
What this means in practice
- •For medical imaging teams: Train lesion detection models for rare brain hemorrhages without needing real annotated cases.
- •For clinical ai developers: Create brain lesion segmentation tools that work well despite limited training data for specific hemorrhagic lesions.
Authors
Yuan Cao, Sumeet Dash, Antonia Zachariadis, Stefanie Schreiber, Katja Neumann, Jose Bernal
Abstract
Cerebral microbleeds (CMBs) and cortical superficial siderosis (cSS) are imaging markers of cerebral small vessel disease, but their automated segmentation is limited by the scarcity of positive cases and voxel-level annotations. We propose a synthetic training framework for long-tail haemorrhagic lesion segmentation that requires no real lesion annotations for training and leverages radiological description of the lesions. Starting from anatomical brain parcellations, the framework applies spatial augmentation and voxel resampling, procedurally inserts cSS and CMB labels using clinical priors on lesion location and morphology, and synthesises images through randomised intensity assignment, blurring, and Rician noise simulation. Models were trained on dynamically generated image-label pairs and evaluated against manual delineations in 10 cSS cases and 13 CMB cases. The proposed configurations outperformed classical filter baselines. For cSS, the hypointensity constrained model achieved higher AUPRC and AUROC than the Frangi filter (AUPRC: 0.284 vs 0.083; AUROC: 0.907 vs 0.731). For CMBs, explicit synthesis of blood vessels as lesion mimics improved performance over the classical baseline (AUPRC: 0.538 vs 0.004; AUROC: 0.999 vs 0.968). These results support our proposal as a feasible strategy for data-scarce haemorrhagic lesion segmentation.