Motion Artifact Removal in Pixel-Frequency Domain via Alternate Masks and Diffusion Model

Authors: Jiahua Xu, Dawei Zhou, Lei Hu, Jianfeng Guo, Feng Yang, Zaiyi Liu, Nannan Wang, Xinbo Gao

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Quantitative experiments are performed on datasets from different tissues and show that our method achieves superior performance on several metrics. Qualitative evaluations with radiologists also show that our method provides better clinical feedback.
Researcher Affiliation Academia 1State Key Laboratory of Integrated Services Networks, Xidian University, Xi an, China 2Department of Radiology, Guangdong Provincial People s Hospital, Southern Medical University, Guangzhou, China 3Department of Radiology, Xiangyang No. 1 People s Hospital, Hubei University of Medicine, Xiangyang, China 4Chongqing Key Laboratory of Image Cognition, Chongqing University of Posts and Telecommunications, Chongqing, China
Pseudocode Yes Algorithm 1: PFAD for motion artifact removal Input: xori, x T N (0, I), Φh, Φl, a, Dθ, { αi}T 1 , {mj}T 1 . Output: The motion-free image ex0. 1: for i = T to 1 do xi 1 Dθ(xi) 3: ωi 1 αi 4: Mi ωi mi 5: Φl(fx i 1) Φl(fxori) 6: Φh(fx i 1) Φh(fxori) Mi + Φh(fxi 1) (1 Mi) 7: x i 1 |F 1(Φh(fx i 1) + Φl(fx i 1))| Frequency domain 8: ϵ N(0, I) 9: xfor i 1 αi xori + 1 αi ϵ 10: x i 1 xfor i 1 Mi + xi 1 (1 Mi) Pixel domain 11: γi a e i T + 1 12: exi 1 γi x i 1 + (1 γi) x i 1 Dual domain balance 13: xi 1 exi 1 14: end for 15: return ex0
Open Source Code Yes Code https://github.com/medcx/PFAD
Open Datasets Yes We utilize two public datasets and one private dataset for the evaluation and validation of PFAD. The first dataset is the Human Connectome Project (HCP) data (Van Essen et al. 2012), which is a dataset containing MRI data of the human brain, and the second dataset is the knee MRI data from fast MRI (Zbontar et al. 2018).
Dataset Splits No The paper mentions using public datasets (HCP, fast MRI) and a private one, and refers to "test dataset" in table captions, but it does not specify the exact percentages, sample counts, or methodology for training, validation, and test splits within the main text.
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments, such as GPU or CPU models.
Software Dependencies No The paper does not provide specific software dependencies with version numbers needed to replicate the experiment.
Experiment Setup Yes In our experiments, we choose the value of a for the best case of the total metric, where a is equal to 0.7.