Multi-shot diffusion-weighted MRI reconstruction using unrolled network with U-net as priors
View Patent ↗A method of magnetic resonance imaging performs a scan by a magnetic resonance imaging system to acquire k-space data; applies the k-space data as input to an unrolled convolutional neural network comprising multiple iterations, and generates reconstructed images from the output of the unrolled convolutional neural network by combining images from different shots. Each iteration of the unrolled network performs a first gradient update, applies the result to a first U-net in k-space, performs a second gradient update, and applies a second U-net in image space. The first gradient update and the second gradient update are based on a theoretical gradient from a physical measurement model.
1. A method of magnetic resonance imaging comprising:
a) performing a scan by a magnetic resonance imaging system to acquire k-space data;
b) applying the k-space data as input to an unrolled convolutional neural network comprising multiple iterations, wherein each iteration comprises:
i) performing a first gradient update,
ii) applying a first U-net in k-space,
iii) performing a second gradient update, and
iv) applying a second U-net in image space;
wherein the first gradient update and the second gradient update are based on a theoretical gradient from a physical measurement model; and
c) generating reconstructed images from the output of the unrolled convolutional neural network by combining images from different shots.
2. The method of claim 1 wherein the physical measurement model comprises a receiving coil sensitivity map and a data sampling pattern.
3. The method of claim 1 wherein the scan is a multi-shot diffusion weighted scan.
4. The method of claim 1 wherein the first U-net and second U-net include skip connections.
5. The method of claim 1 wherein the first U-net and second U-net include res-nets.