Simultaneous Super-Resolution and Distortion Correction for Single-shot EPI DWI using Deep Learning

Abstract

Single-shot echo planer imaging (SS-EPI) is widely used for clinical Diffusion-weighted magnetic resonance imaging (DWI) acquisitions. However, due to the limited bandwidth along the phase encoding direction, the obtained images suffer from distortion and blurring, which limits its clinical value for diagnosis. Here we proposed a deep learning-based image-quality-transfer method with a novel loss function with improved network structure to simultaneously increase the resolution and correct distortions for SS-EPI. We proposed a modified network structure based on Generative Adversarial Networks (GAN). A dense net with gradient map guidance and a multi-level fusion block was employed as the generator to suppress the over-smoothing effect. We also proposed a fractional anisotropy (FA) loss to exploit the intrinsic signal relations in DWI. In-vivo brain DWI data were used to test the proposed method. The results showed that the distortion-corrected high-resolution DWI images with restored anatomical details can be obtained from low-resolution SS-EPI images by taking the advantage of high-resolution anatomical images. Additionally, the proposed FA loss can improve the image quality and quantitative accuracy of diffusion metrics by utilizing the intrinsic relations among different diffusion directions.

Distortion-aware super-resolution for planetary exploration images

“…Ye. at al [16] also solved the super-resolution and distortion issues in a Single-shot echo planer imaging (SS-EPI) in the clinical domain. Their proposed network consists of GAN with a gradient map and fusion block to overcome the smoothing effect in the images.…”

Differentiation of white matter histopathology using b-tensor encoding and machine learning

“…In biomedical imaging, there are multiple examples of ML applications like automatic image segmentation, data processing, MRI reconstruction, etc. In DW-MRI, ML techniques have been used for data pre-processing [13,14], estimation of diffusion parameters [15][16][17][18][19], automatic white matter bundle segmentation [20], among other applications (for a review, see [21]). There are still, however, several opportunities for clinical applications and improvements in the detection of histopathology.…”