通过投影空间双层次个性化，提升低剂量CT去噪效果。
Projection Guided Personalized Federated Learning for Low Dose CT Denoising
- 在投影空间实现患者与扫描仪双重个性化建模
- 相比基线方法提升1.42 dB，最高达44.83 dB PSNR
- 适合医疗联邦学习中需区分噪声与解剖结构的场景
低剂量CT可降低辐射暴露，但会引入与协议相关的噪声和伪影，且不同机构间差异显著。联邦学习可在不集中患者数据的前提下协同训练，但现有方法在图像空间进行个性化，难以分离扫描仪噪声与患者解剖结构。本文提出ProFed（投影引导的个性化联邦学习），在噪声产生的投影空间中实现双层级个性化，同时结合协议与解剖特征。ProFed引入：(i) 解剖与协议感知网络，分别适应患者与扫描仪特性；(ii) 多约束投影损失，确保重建结果符合原始测量数据；(iii) 不确定性引导的选择性聚合，按预测置信度加权客户端贡献。在Mayo Clinic 2016数据集上的实验表明，使用CNN骨干网络时达到42.56 dB PSNR，使用Transformer时达44.83 dB PSNR，优于11种联邦学习基线方法，包括物理信息驱动的SCAN-PhysFed，提升达+1.42 dB。
原文摘要 · Abstract (English)
Low-dose CT (LDCT) reduces radiation exposure but introduces protocol-dependent noise and artifacts that vary across institutions. While federated learning enables collaborative training without centralizing patient data, existing methods personalize in image space, making it difficult to separate scanner noise from patient anatomy. We propose ProFed (Projection Guided Personalized Federated Learning), a framework that complements the image space approach by performing dual-level personalization in the projection space, where noise originates during CT measurements before reconstruction combines protocol and anatomy effects. ProFed introduces: (i) anatomy-aware and protocol-aware networks that personalize CT reconstruction to patient and scanner-specific features, (ii) multi-constraint projection losses that enforce consistency with CT measurements, and (iii) uncertainty-guided selective aggregation that weights clients by prediction confidence. Extensive experiments on the Mayo Clinic 2016 dataset demonstrate that ProFed achieves 42.56 dB PSNR with CNN backbones and 44.83 dB with Transformers, outperforming 11 federated learning baselines, including the physics-informed SCAN-PhysFed by +1.42 dB.
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。