Related Experiment Video
Updated: Feb 16, 2026
08:22
3D Cine Magnetic Resonance Imaging of Respiratory Motion in Mechanically Ventilated Mice and Rats
Published on: September 19, 2025
1.1K
Respiratory motion correction for free-breathing 3D abdominal MRI using CNN-based image registration: a feasibility
Jun Lv1, Ming Yang2, Jue Zhang1,3
11 Academy for Advanced Interdisciplinary Studies, Peking University , Beijing , China.
The British Journal of Radiology
|December 21, 2017
Summary
A novel convolutional neural network (CNN) method accurately registers free-breathing abdominal MRI scans, significantly reducing motion artifacts and improving image quality. This approach drastically cuts registration time, showing promise for clinical applications.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Image Registration
Background:
- Free-breathing abdomen imaging necessitates non-rigid motion registration to correct for respiratory motion in undersampled 3D datasets.
- Existing methods struggle with accurate motion correction, impacting image quality and diagnostic utility.
Purpose of the Study:
- To introduce and evaluate a convolutional neural network (CNN) based image registration method for obtaining motion-free abdominal images throughout the respiratory cycle.
- To compare the CNN approach against non-motion corrected (NMC) and local affine registration (LREG) methods.
Main Methods:
- Abdominal MRI data were acquired from 10 volunteers at 1.5 T.
- Respiratory signals were extracted to bin data into three respiratory phases for retrospective reconstruction using non-Cartesian iterative SENSE.
- A CNN was trained to generate displacement vector fields for registration, and its performance was compared with NMC and LREG.
Main Results:
- CNN-based registration resulted in sharper and more consecutive blood vessels and clearer liver contours compared to LREG and NMC.
- CNN achieved the highest image quality, with superior Signal-to-Noise Ratio (SNR) and visual scores compared to NMC and LREG.
- CNN reduced registration time from approximately 1 hour to 1 minute, a significant improvement over LREG.
Conclusions:
- The CNN-based registration method is feasible and outperforms NMC and LREG methods in abdominal imaging.
- This technique offers a substantial reduction in registration time, indicating strong potential for clinical adoption.
- This study represents the first application of a CNN-based registration method for abdominal images.
