针对癫痫FLAIR MRI中病灶稀少难建模，提出共享潜空间扩散模型提升生成稳定性。
SLIM-Diff: Shared Latent Image-Mask Diffusion with Lp loss for Data-Scarce Epilepsy FLAIR MRI
- 用双通道图像+掩码输入，共享瓶颈U-Net强耦合解剖结构与病灶形态。
- 使用x₀预测和L₁.₅损失可提升图像保真度，L₂损失更好保持病灶掩码形状。
- 适用于数据稀缺的医学图像生成，尤其适合癫痫病灶建模研究者。
癫痫FLAIR MRI中的局灶性皮质发育不良（FCD）病灶细微且稀少，导致图像-掩码联合生成模型易不稳定且产生记忆。本文提出SLIM-Diff，一种紧凑的联合扩散模型：(i) 采用单个共享瓶颈U-Net，从双通道图像+掩码表示中强制解剖结构与病灶几何的紧密耦合；(ii) 通过可调$L_p$目标函数优化损失几何。作为内部基线，包含标准DDPM风格目标（ε-预测，$L_2$损失），并在此匹配设置下分离预测参数化与$L_p$几何的影响。实验表明，$x_0$-预测在联合合成中始终最优，分数阶次小于二次的惩罚（$L_{1.5}$）提升图像保真度，而$L_2$更优保留病灶掩码形态。代码与模型权重见https://github.com/MarioPasc/slim-diff。
原文摘要 · Abstract (English)
Focal cortical dysplasia (FCD) lesions in epilepsy FLAIR MRI are subtle and scarce, making joint image--mask generative modeling prone to instability and memorization. We propose SLIM-Diff, a compact joint diffusion model whose main contributions are (i) a single shared-bottleneck U-Net that enforces tight coupling between anatomy and lesion geometry from a 2-channel image+mask representation, and (ii) loss-geometry tuning via a tunable $L_p$ objective. As an internal baseline, we include the canonical DDPM-style objective ($ε$-prediction with $L_2$ loss) and isolate the effect of prediction parameterization and $L_p$ geometry under a matched setup. Experiments show that $x_0$-prediction is consistently the strongest choice for joint synthesis, and that fractional sub-quadratic penalties ($L_{1.5}$) improve image fidelity while $L_2$ better preserves lesion mask morphology. Our code and model weights are available in https://github.com/MarioPasc/slim-diff
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。