arXiv：cs.LG· Weidong Zhang, Yongchan Jung, Shafayat Mowla Anik, Furen Xiao, Vasudevan Janarthanan, Enkhzaya Chuluunbaatar, Marco Ho, Byeong Kil Lee, Jeeho Ryoo·· 13 小时前
Lesion-DDPM：面向 MS MRI 合成的病灶增强 3D 扩散模型
Lesion-DDPM: Lesion-Enhanced 3D Diffusion for MS MRI Synthesis
AI 导读
Lesion-DDPM 是一个用于病灶感知 FLAIR 合成的 3D 条件扩散框架，通过多层级解剖掩码注入和病灶加权重建损失，在 MSLesSeg 数据集子集上取得所有对比方法中最低的病灶区域重建误差。仅用其生成扫描训练的 3D U-Net 在真实 MRI 上病灶分割 Dice 达 0.616，高于最佳竞品合成数据的 0.569；将生成图像加入真实训练集后 Dice 进一步提升至 0.685。
正文
Abstract:3D FLAIR MRI is widely recommended as one of the standard MRI sequences for brain imaging in multiple sclerosis (MS), but publicly available MS datasets remain relatively small and vary across scanners, acquisition protocols, and lesion patterns. This scarcity and variability hinder the development of robust neuroimaging machine learning models and are particularly challenging for generative models that aim to synthesize images while preserving small, sparse lesions. We propose Lesion-DDPM, a 3D conditional diffusion framework for lesion-aware FLAIR synthesis that incorporates multi-level anatomical mask injection together with a lesion-weighted reconstruction loss to emphasize lesion voxels while maintaining global brain structure. Using a curated subset of the MSLesSeg dataset, we compare Lesion-DDPM with representative state-of-the-art GAN- and diffusion-based models, assessing both image-generation metrics and downstream 3D U-Net segmentation. In our experiments, Lesion-DDPM achieved the lowest lesion-region reconstruction error among all methods. In a downstream 3D U-Net lesion segmentation task, a model trained only on Lesion-DDPM-generated scans and evaluated on real MRIs reached a Dice score of 0.616 compared with 0.569 for the best competing synthetic dataset. When Lesion-DDPM images were added to the real training set, the Dice score further increased to 0.685.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2606.15457 [cs.CV] |
| (or arXiv:2606.15457v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.15457 arXiv-issued DOI via DataCite |
Submission history
From: Jeeho Ryoo [view email]
[v1]
Sat, 13 Jun 2026 20:21:20 UTC (389 KB)
[v2]
Thu, 8 Oct 2026 18:14:32 UTC (389 KB)
来源：arXiv：cs.LG · arxiv.org