Published July 1, 2024
| Accepted
Journal Article
Open
Learning Task-Specific Strategies for Accelerated MRI
Abstract
Compressedsensing magnetic resonance imaging (CS-MRI) seeks to recover visual information from subsampled measurements for diagnostic tasks. Traditional CS-MRI methods often separately address measurement subsampling, image reconstruction, and task prediction, resulting in a suboptimal end-to-end performance. In this work, we propose Tackle as a unified co-design framework for jointly optimizing subsampling, reconstruction, and prediction strategies for the performance on downstream tasks. The naïve approach of simply appending a task prediction module and training with a task-specific loss leads to suboptimal downstream performance. Instead, we develop a training procedure where a backbone architecture is first trained for a generic pre-training task (image reconstruction in our case), and then fine-tuned for different downstream tasks with a prediction head. Experimental results on multiple public MRI datasets show that Tackle achieves an improved performance on various tasks over traditional CS-MRI methods. We also demonstrate that Tackle is robust to distribution shifts by showing that it generalizes to a new dataset we experimentally collected using different acquisition setups from the training data. Without additional fine-tuning, Tackle leads to both numerical and visual improvements compared to existing baselines. We have further implemented a learned 4×-accelerated sequence on a Siemens 3 T MRI Skyra scanner. Compared to the fully-sampling scan that takes 335 seconds, our optimized sequence only takes 84 seconds, achieving a four-fold time reduction as desired, while maintaining high performance.
Copyright and License
© 2024 IEEE.
Acknowledgement
The authors would like to thank Xinyi Wu for her assistance as a volunteer in testing our learned MRI sequence and collecting data.
This work was sponsored by NSF Award 2048237, NIH Projects 5R01AG064027, 5R01AG070988, R21EB029641, R01HD099846, R01HD085813, Heritage Medical Research Fellowship, S2I Clinard Innovation Award, and Rockley Photonics. Z. Wu was sponsored by the Kortschak Fellowship, Amazon AI4Science Fellowship, and Amazon AI4Science Partnership Discovery Grant.
Files
Learning_Task-Specific_Strategies_for_Accelerated_MRI.pdf
Files
(56.8 MB)
Name | Size | Download all |
---|---|---|
md5:c2a3dcc23c3b904713c5594acd138ca4
|
12.3 MB | Preview Download |
md5:85adb54ec8bd9470c414c6358b300409
|
44.5 MB | Download |
Additional details
- National Science Foundation
- CCF-2048237
- National Institutes of Health
- 5R01AG064027
- National Institutes of Health
- 5R01AG070988
- National Institutes of Health
- R21EB029641
- National Institutes of Health
- R01HD099846
- National Institutes of Health
- R01HD085813
- California Institute of Technology
- Heritage Medical Research Institute
- California Institute of Technology
- S2I Clinard Innovation Award Caltech Center for Sensing to Intelligence
- Rockley Photonics
- California Institute of Technology
- Kortschak Scholars Program
- Amazon (United States)
- Caltech groups
- Heritage Medical Research Institute, Caltech Center for Sensing to Intelligence (S2I)