# Synthetic training for long-tail haemorrhagic lesion segmentation in data-scarce settings

arXiv: https://arxiv.org/abs/2610.01542 · HTML: https://mlsanity.com/paper/2610.01542/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 18 references that are in the map.
- First posted on arXiv 1 Oct 2026.

## Builds on (most cited first, 8 of 8)

- [SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining](https://mlsanity.com/paper/2107.09559/) (2021)
- [SynthStrip: Skull-Stripping for Any Brain Image](https://mlsanity.com/paper/2203.09974/) (2022)
- [Robust machine learning segmentation for large-scale analysis of heterogeneous clinical brain MRI datasets](https://mlsanity.com/paper/2209.02032/) (2022)
- [Anatomy-aware and acquisition-agnostic joint registration with SynthMorph](https://mlsanity.com/paper/2301.11329/) (2023)
- [Quantifying white matter hyperintensity and brain volumes in heterogeneous clinical and low-field portable MRI](https://mlsanity.com/paper/2312.05119/) (2023)
- [Where is VALDO? VAscular Lesions Detection and segmentatiOn challenge at MICCAI 2021](https://mlsanity.com/paper/2208.07167/) (2022)
- [Synthetic Data for Robust Stroke Segmentation](https://mlsanity.com/paper/2404.01946/) (2024)
- [SIAM: Head and Brain MRI Segmentation from Few High-Quality Templates via Synthetic Training](https://mlsanity.com/paper/2605.02737/) (2026)

## Abstract (arXiv)

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.