Related Experiment Video
Updated: Aug 9, 2026
06:56
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
Self-supervised reconstruction framework via motion- and physics-informed learning for four-dimensional magnetic
Chenyang Liu1, Lu Wang2, Xiang Wang1
1Department of Health Technology and Informatics, The Hong Kong Polytechnic University, 999077, Hong Kong Special Administrative Region of China.
Medical Image Analysis
|August 7, 2026
Summary
A new self-supervised deep learning method, SS-4DMRF, reconstructs motion-resolved liver tissue maps faster and more accurately. This advances precision in liver cancer radiotherapy using quantitative MRI.
Area of Science:
- Medical Imaging
- Radiotherapy
- Artificial Intelligence in Medicine
Background:
- Four-dimensional magnetic resonance fingerprinting (4DMRF) offers precise, motion-resolved tissue quantification for liver cancer radiotherapy.
- Clinical adoption is limited by long reconstruction times and the lack of ground-truth 4D data for deep learning acceleration.
Purpose of the Study:
- To develop the first self-supervised reconstruction framework (SS-4DMRF) for 4DMRF, enabling motion-resolved tissue map reconstruction without supervised labels.
- To accelerate 4DMRF reconstruction time while maintaining or improving accuracy for clinical translation.
Main Methods:
- Proposed SS-4DMRF framework utilizing a temporal low-rank-constrained registration (TelReg) network for motion modeling.
- Employed self-supervision using undersampled k-space data and subspace images, leveraging respiratory motion's low-rank property.
- Implemented a physics-informed pattern matching (PiPM) network with Swin Transformers and Bloch-equation-guided denoising for motion-informed compensation and tissue quantification.
Main Results:
- SS-4DMRF demonstrated superior accuracy in tissue quantification and motion measurement compared to state-of-the-art methods on phantom and patient data.
- Achieved significantly reduced normalized root-mean-square error (NRMSE) in 4D tissue property quantification and improved inter-phase motion repeatability (p<0.001).
- Showcased a 10-fold acceleration in reconstruction time, with accurate tumor motion trajectory correlation (r=0.939±0.057).
Conclusions:
- SS-4DMRF provides rapid, precise, and motion-resolved quantitative MRI, enhancing liver cancer radiotherapy precision.
- Establishes a clinically feasible platform for accelerated abdominal quantitative MRI in oncology.
- Overcomes limitations of supervised deep learning by enabling self-supervised reconstruction from undersampled data.
Related Concept Videos
01:24Magnetic Resonance Imaging
Magnetic resonance imaging (MRI) is a noninvasive medical imaging technique based on a phenomenon of nuclear physics discovered in the 1930s, in which matter exposed to magnetic fields and radio waves was found to emit radio signals. In 1970, a physician and researcher named Raymond Damadian noticed that malignant (cancerous) tissue gave off different signals than normal body tissue. He applied for a patent for the first MRI scanning device in clinical use by the early 1980s. The early MRI...
01:14NMR Spectrometers: Resolution and Error Correction
When magnetic nuclei in a sample achieve resonance and undergo relaxation, the signal detected in NMR is an approximately exponential free induction decay. Fourier transform of an exponential decay yields a Lorentzian peak in the frequency domain. Lorentzian peaks in an NMR spectrum are defined by their amplitude, full width at half maximum, and position, where the peak width is governed by the spin-spin relaxation time alone. In real experiments, however, the applied magnetic field is rendered...