Related Experiment Video
Updated: Mar 16, 2026
02:09
Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function
Published on: April 12, 2024
1.1K
NeuroMix-DL: Improving imaging quality of a fast multiparametric MRI protocol using deep learning.
Amirhossein Sanaat1, Johannes Hugo Decker2, Ramy Hussein2
1Division of Nuclear Medicine and Molecular Imaging, Geneva University Hospital, Geneva, Switzerland; Department of Radiology, Stanford University, Stanford, CA 94305, USA.
European Journal of Radiology
|March 14, 2026
Summary
Deep learning significantly enhances fast brain MRI quality. The Swin U-Net Transformer (SwinUNETR) model improves image resolution and reduces artifacts in multi-contrast sequences, offering a better cost-benefit ratio for MRI.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Radiology
Background:
- Fast multi-contrast MRI protocols are crucial for efficient brain imaging.
- Deep learning offers potential for enhancing image quality in accelerated MRI sequences.
- Artifacts and resolution limitations can impact diagnostic accuracy in rapid MRI.
Purpose of the Study:
- To improve the image quality of a fast multi-contrast brain MRI protocol using deep learning.
- To evaluate the effectiveness of a Swin U-Net Transformer (SwinUNETR) model for enhancing NeuroMix sequences.
Main Methods:
- Retrospective analysis of 350 patients undergoing fast multi-contrast (NeuroMix) and conventional MRI.
- Application of a SwinUNETR model to enhance T1-weighted, T2-weighted, and FLAIR images (NeuroMix-DL).
- Quantitative (RMSE) and qualitative (clinical assessment scale) evaluation of image quality.
Main Results:
- SwinUNETR significantly reduced RMSE by 42% (T1w), 33% (T2w), and 33% (FLAIR) (p < 0.001).
- Clinical readers rated NeuroMix-DL images higher than original NeuroMix images.
- Improvements were observed in reducing common artifacts like motion and inhomogeneity.
Conclusions:
- The SwinUNETR model effectively enhances fast multi-contrast MRI quality.
- This deep learning approach mitigates artifacts and improves the cost-benefit ratio of MRI.
- SwinUNETR presents a viable solution for high-quality, accelerated brain MRI acquisition.
