Projection Guided Personalized Federated Learning for Low Dose CT Denoising

TL;DR

ProFed introduces a projection space personalization framework for federated low-dose CT denoising, effectively separating scanner noise from patient anatomy and outperforming existing methods.

Contribution

It proposes a dual-level personalization in projection space with multi-constraint losses and uncertainty-guided aggregation, advancing federated CT reconstruction.

Findings

Achieves 42.56 dB PSNR with CNN backbones.

Outperforms 11 federated baselines, including SCAN-PhysFed.

Demonstrates significant noise reduction and image quality improvement.

Abstract

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.