{"title":"Anatomical structure-guided joint spatiotemporal graph embedding framework for magnetic resonance fingerprint reconstruction","authors":"Peng Li , Jianxing Liu , Yue Hu","doi":"10.1016/j.media.2025.103816","DOIUrl":null,"url":null,"abstract":"<div><div>Highly undersampled acquisition schemes in magnetic resonance fingerprinting (MRF) typically introduce aliasing artifacts, degrading the accuracy of quantitative imaging. While state-of-the-art graph-based reconstruction methods have shown promise in addressing this challenge by leveraging non-local and non-linear correlations in MRF data, they often face two critical limitations: high computational costs associated with large-scale graph structure estimation and limited capacity to capture complex spatiotemporal dynamics. To overcome these challenges, this study proposes an anatomical structure-guided joint spatiotemporal graph embedding framework for MRF reconstruction. By integrating anatomical segmentation and homogeneity clustering, our framework partitions MRF data into spatially contiguous regions and groups them into clusters based on tissue homogeneity. Subgraphs are then constructed for each cluster, capturing non-local spatial correlations while preserving fine-grained temporal signal dynamics. The hierarchical graph embedding architecture enables efficient focusing on critical correlations, significantly improving reconstruction performance and reducing computational complexity. Numerical experiments on both simulated and <em>in vivo</em> MRF datasets demonstrate that our method outperforms state-of-the-art methods, achieving a <span><math><mo>∼</mo></math></span>2 dB higher signal-to-noise ratio (SNR) in reconstructed data and a <span><math><mo>∼</mo></math></span>70% reduction in reconstruction time. The source code is publicly available at <span><span>https://github.com/bigponglee/SP_GE_MRF</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":18328,"journal":{"name":"Medical image analysis","volume":"107 ","pages":"Article 103816"},"PeriodicalIF":11.8000,"publicationDate":"2025-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Medical image analysis","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1361841525003627","RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
Highly undersampled acquisition schemes in magnetic resonance fingerprinting (MRF) typically introduce aliasing artifacts, degrading the accuracy of quantitative imaging. While state-of-the-art graph-based reconstruction methods have shown promise in addressing this challenge by leveraging non-local and non-linear correlations in MRF data, they often face two critical limitations: high computational costs associated with large-scale graph structure estimation and limited capacity to capture complex spatiotemporal dynamics. To overcome these challenges, this study proposes an anatomical structure-guided joint spatiotemporal graph embedding framework for MRF reconstruction. By integrating anatomical segmentation and homogeneity clustering, our framework partitions MRF data into spatially contiguous regions and groups them into clusters based on tissue homogeneity. Subgraphs are then constructed for each cluster, capturing non-local spatial correlations while preserving fine-grained temporal signal dynamics. The hierarchical graph embedding architecture enables efficient focusing on critical correlations, significantly improving reconstruction performance and reducing computational complexity. Numerical experiments on both simulated and in vivo MRF datasets demonstrate that our method outperforms state-of-the-art methods, achieving a 2 dB higher signal-to-noise ratio (SNR) in reconstructed data and a 70% reduction in reconstruction time. The source code is publicly available at https://github.com/bigponglee/SP_GE_MRF.
期刊介绍:
Medical Image Analysis serves as a platform for sharing new research findings in the realm of medical and biological image analysis, with a focus on applications of computer vision, virtual reality, and robotics to biomedical imaging challenges. The journal prioritizes the publication of high-quality, original papers contributing to the fundamental science of processing, analyzing, and utilizing medical and biological images. It welcomes approaches utilizing biomedical image datasets across all spatial scales, from molecular/cellular imaging to tissue/organ imaging.