Researchers at MIT have developed a deep learning model for motion correction in brain MRI scans. MRI scans are highly sensitive to motion, resulting in image artifacts and potential misdiagnoses. The model, titled "Data Consistent Deep Rigid MRI Motion Correction," constructs a motion-free image from motion-corrupted data without changing the scanning procedure. The researchers aim to combine physics-based modeling and deep learning to ensure consistency between the image output and the actual measurements. The development could help improve patient outcomes, especially for those with neurological disorders like Alzheimer's or Parkinson's disease. Motion affects approximately 15% of brain MRIs and results in significant hospital expenditures. Future work could explore more sophisticated types of motion and expand to other body parts. The research was awarded best oral presentation at the Medical Imaging with Deep Learning conference.
Thu, 17 Aug 2023 13:00:00 GMT | MIT News