用无监督扩散模型预训练，让医学图像分割更准更稳。
Unsupervised Anatomical Feature Learning via Diffusion Models: Enhanced Medical Image Segmentation with Denoising Diffusion Probabilistic Models
- 用21个未标注腹部CT训练扩散模型，提取解剖结构特征。
- 肝脏分割Dice达0.93，边界精度提升74%，低数据下仍稳定。
- 无需标签也能保持80%性能，适合标注稀缺的医疗场景。
医学图像分割的像素级标注是严重瓶颈。传统U-Net虽有效，但仅学习局部纹理，缺乏全局解剖结构感知，在小样本情况下常出现边界分割失败。本文提出利用无监督去噪扩散概率模型（DDPM）提取解剖特征。在21个未标注腹部CT上训练DDPM以学习结构表征，并将编码器权重迁移至下游分割任务，评估使用BTCV多器官数据集。扩散预训练显著提升肝脏分割：Dice从0.75±0.36提升至0.93±0.16（p < 5.33×10⁻²⁶，Cohen's d = 0.529），平均表面距离降低66%，95%分位数豪斯多夫距离减少45%。肾脏分割Dice从0.90±0.19升至0.95±0.10（p < 4.01×10⁻¹¹）。多器官综合表现显示方差降低68%，边界精度提升74%（Dice: 0.95±0.07）。关键发现：冻结编码器模型在未见标签时仍保持超过80%微调性能，证明已习得解剖先验。在低数据场景中，仅需50%（肝脏Dice: 0.92，肾脏: 0.94）、25%和10%（肝脏: 0.89，肾脏: 0.71）标注数据，模型依然稳健。利用未标注图像进行扩散预训练，成功嵌入鲁棒解剖特征，使U-Net具备解剖意识。
原文摘要 · Abstract (English)
Acquiring pixel-level annotations for medical image segmentation is a severe bottleneck. Traditional U-Net architectures, while effective, learn local texture patterns and lack awareness of global anatomical structures, leading to boundary delineation failures in low-data regimes. This research paper proposes utilizing unsupervised Denoising Diffusion Probabilistic Models (DDPMs) to extract anatomical features. We train a DDPM on 21 unlabeled abdominal CT scans to learn structural representations, transferring the encoder weights to a downstream segmentation task evaluated on the BTCV multi-organ dataset. Diffusion pretraining significantly improved liver segmentation: Dice increased from $0.75\pm0.36$ to $0.93\pm0.16$ ($p < 5.33\times10^{-26}$, 0.529 Cohen's d), Average Surface Distance (ASD) decreased by 66%, and 95th-percentile Hausdorff Distance (HD95) reduced by 45%. For kidney segmentation, Dice improved from $0.90\pm0.19$ to $0.95\pm0.10$ ($p < 4.01\times10^{-11}$). Multi-organ pooled performance showed a 68% variance reduction and a 74% improvement in boundary precision (Dice $0.95\pm0.07$). Crucially, frozen encoder models retained > 80% of fine-tuned performance without exposure to segmentation labels, proving the existence of learned anatomical priors. In low-data scenarios, diffusion-pretrained models maintained robust performance with only 50% (Dice: 0.92 liver, 0.94 kidney), 25%, and even 10% (Dice: 0.89 liver, 0.71 kidney) of labeled data. Using unlabeled images for diffusion-based pretraining successfully embeds robust anatomical features prior to human supervision, transforming U-Nets into anatomy-aware systems.
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。