Enhanced Magnetic Resonance Image Synthesis with Contrast-Aware Generative Adversarial Networks.

Jonas Denck,Jens Guehring,Andreas Maier,Eva Rothgang

doi:10.3390/jimaging7080133

Abstract

A magnetic resonance imaging (MRI) exam typically consists of the acquisition of multiple MR pulse sequences, which are required for a reliable diagnosis. With the rise of generative deep learning models, approaches for the synthesis of MR images are developed to either synthesize additional MR contrasts, generate synthetic data, or augment existing data for AI training. While current generative approaches allow only the synthesis of specific sets of MR contrasts, we developed a method to generate synthetic MR images with adjustable image contrast. Therefore, we trained a generative adversarial network (GAN) with a separate auxiliary classifier (AC) network to generate synthetic MR knee images conditioned on various acquisition parameters (repetition time, echo time, and image orientation). The AC determined the repetition time with a mean absolute error (MAE) of 239.6 ms, the echo time with an MAE of 1.6 ms, and the image orientation with an accuracy of 100%. Therefore, it can properly condition the generator network during training. Moreover, in a visual Turing test, two experts mislabeled 40.5% of real and synthetic MR images, demonstrating that the image quality of the generated synthetic and real MR images is comparable. This work can support radiologists and technologists during the parameterization of MR sequences by previewing the yielded MR contrast, can serve as a valuable tool for radiology training, and can be used for customized data generation to support AI training.

Highlights

In magnetic resonance imaging (MRI), multiple contrasts are usually acquired within a single exam that are required to make a reliable diagnosis
To mitigate the problem of the availability of MR images for varying contrast settings, we developed an approach for MR image synthesis that can be parameterized with acquisition parameters
The rest of our paper is structured as follows: first, we shortly review the current literature of generative adversarial networks with the focus on MR image contrast synthesis (Section 2)

Summary

Introduction

In magnetic resonance imaging (MRI), multiple contrasts are usually acquired within a single exam that are required to make a reliable diagnosis. The acquisition parameters for an MR sequence affect image contrast, image resolution, signal-to-noise ratio, and scan time. Important acquisition parameters that affect the image contrast are the repetition time (TR) and echo time (TE). The guidelines do not specify exact acquisition parameter settings as they will depend on the field strength and desired contrast weighting. GANs have been successfully applied to the field of medical imaging (e.g., [15,16]), with applications that can mainly be divided into seven categories: synthesis, segmentation, reconstruction, detection, denoising, registration, and classification, whereas the majority of publications address synthesis applications [17]. A detailed general overview of GANs in the field of medical imaging is given in [18]

Methods

Results

Discussion

Conclusion