Adversarial and Perceptual Refinement for Compressed Sensing MRI Reconstruction

TL;DR

This paper introduces a hybrid deep learning method combining adversarial, perceptual, and MSE losses for improved MRI reconstruction, along with a new interpretability score to objectively evaluate image quality for clinical use.

Contribution

It proposes a novel hybrid loss function and a semantic interpretability score for better MRI image reconstruction and evaluation.

Findings

Significant improvement in image quality over state-of-the-art methods

Enhanced visual sharpness and detail preservation in reconstructed images

Objective quantification of image usefulness for clinical analysis

Abstract

Deep learning approaches have shown promising performance for compressed sensing-based Magnetic Resonance Imaging. While deep neural networks trained with mean squared error (MSE) loss functions can achieve high peak signal to noise ratio, the reconstructed images are often blurry and lack sharp details, especially for higher undersampling rates. Recently, adversarial and perceptual loss functions have been shown to achieve more visually appealing results. However, it remains an open question how to (1) optimally combine these loss functions with the MSE loss function and (2) evaluate such a perceptual enhancement. In this work, we propose a hybrid method, in which a visual refinement component is learnt on top of an MSE loss-based reconstruction network. In addition, we introduce a semantic interpretability score, measuring the visibility of the region of interest in both ground truth and reconstructed images, which allows us to objectively quantify the usefulness of the image quality for image post-processing and analysis. Applied on a large cardiac MRI dataset simulated with 8-fold undersampling, we demonstrate significant improvements ($p<0.01$) over the state-of-the-art in both a human observer study and the semantic interpretability score.