Deblurring for spiral real-time MRI using convolutional neural networks

Yongwan Lim, Shrikanth Narayanan, Krishna Nayak

Keywords: Artifact correction, CNN, deblurring, off-resonance, real-time MRI

Abstract: Spiral acquisitions are preferred in real-time MRI because of their time efficiency. A fundamental limitation of spirals is image blurring due to off-resonance, which degrades image quality significantly at air-tissue boundaries. Here, we demonstrate a simple CNN-based deblurring method for spiral real-time MRI of human speech production. We show the CNN-based deblurring is capable of restoring blurred vocal tract tissue boundaries, without a need for exam-specific field maps. Deblurring performance is superior to a current auto-calibrated method, and slightly inferior to ideal reconstruction with perfect knowledge of the field maps.

Track: short paper

Paper Type: well-validated application

Presentation Upload Agreement: I agree that my presentation material (videos and slides) will be made public.