Towards Ultrafast MRI via Extreme k-Space Undersampling and Superresolution
Authors:
Aleksandr Belov,
Joel Stadelmann,
Sergey Kastryulin,
Dmitry V. Dylov
Abstract:
We went below the MRI acceleration factors (a.k.a., k-space undersampling) reported by all published papers that reference the original fastMRI challenge, and then considered powerful deep learning based image enhancement methods to compensate for the underresolved images. We thoroughly study the influence of the sampling patterns, the undersampling and the downscaling factors, as well as the reco…
▽ More
We went below the MRI acceleration factors (a.k.a., k-space undersampling) reported by all published papers that reference the original fastMRI challenge, and then considered powerful deep learning based image enhancement methods to compensate for the underresolved images. We thoroughly study the influence of the sampling patterns, the undersampling and the downscaling factors, as well as the recovery models on the final image quality for both the brain and the knee fastMRI benchmarks. The quality of the reconstructed images surpasses that of the other methods, yielding an MSE of 0.00114, a PSNR of 29.6 dB, and an SSIM of 0.956 at x16 acceleration factor. More extreme undersampling factors of x32 and x64 are also investigated, holding promise for certain clinical applications such as computer-assisted surgery or radiation planning. We survey 5 expert radiologists to assess 100 pairs of images and show that the recovered undersampled images statistically preserve their diagnostic value.
△ Less
Submitted 4 March, 2021;
originally announced March 2021.
Synthetic CT Generation from MRI Using Improved DualGAN
Authors:
Denis Prokopenko,
Joël Valentin Stadelmann,
Heinrich Schulz,
Steffen Renisch,
Dmitry V. Dylov
Abstract:
Synthetic CT image generation from MRI scan is necessary to create radiotherapy plans without the need of co-registered MRI and CT scans. The chosen baseline adversarial model with cycle consistency permits unpaired image-to-image translation. Perceptual loss function term and coordinate convolutional layer were added to improve the quality of translated images. The proposed architecture was teste…
▽ More
Synthetic CT image generation from MRI scan is necessary to create radiotherapy plans without the need of co-registered MRI and CT scans. The chosen baseline adversarial model with cycle consistency permits unpaired image-to-image translation. Perceptual loss function term and coordinate convolutional layer were added to improve the quality of translated images. The proposed architecture was tested on paired MRI-CT dataset, where the synthetic CTs were compared to corresponding original CT images. The MAE between the synthetic CT images and the real CT scans is 61 HU computed inside of the true CTs body shape.
△ Less
Submitted 19 September, 2019;
originally announced September 2019.
