A research project by Professor Hu Xiangyun's team from the School of Remote Sensing and Information Engineering at Av性爱 (WHU) has been accepted for presentation at MICCAI 2025, one of the world's leading conferences in medical image computing.
VQ-SCD framework: Self-supervised denoising of Low-Dose CT images under unknown scanning conditions.
The paper, titled VQ-SCD: Vector Quantization Meets Unknown Scan Condition Self-supervised Low-Dose CT Denoising, introduces an innovative approach to CT image denoising under unknown scan conditions.
The team proposes a novel self-supervised method, VQ-SCD, which enables effective denoising using only normal-dose CT (NDCT) data for training. The approach is the first to leverage vector quantization to unify feature extraction across diverse scan settings. It also introduces a lightweight diffusion model that enhances image detail reconstruction.
VQ-SCD outperforms existing supervised and state-of-the-art self-supervised methods in both quantitative metrics and visual quality, while achieving fast inference – just 0.25 seconds per image.
The paper's first author is PhD student Su Bo, under the joint supervision of Professor Hu Xiangyun and Academician Li Jiancheng.
