Unsupervised MRI Reconstruction with Generative Adversarial Networks

Abstract

Deep learning-based image reconstruction methods have achieved promising results across multiple MRI applications. However, most approaches require large-scale fully-sampled ground truth data for supervised training. Acquiring fully-sampled data is often either difficult or impossible, particularly for dynamic contrast enhancement (DCE), 3D cardiac cine, and 4D flow. We present a deep learning framework for MRI reconstruction without any fully-sampled data using generative adversarial networks. We test the proposed method in two scenarios: retrospectively undersampled fast spin echo knee exams and prospectively undersampled abdominal DCE. The method recovers more anatomical structure compared to conventional methods.

References in corpus (4)

• RARE: Image Reconstruction using Deep Priors Learned without Ground Truth
• Highly Scalable Image Reconstruction using Deep Neural Networks with Bandpass Filtering
• Unsupervised Deep Basis Pursuit: Learning inverse problems without ground-truth data
• Analysis of Deep Complex-Valued Convolutional Neural Networks for MRI Reconstruction

Cited by in corpus (5)

• Adaptive Diffusion Priors for Accelerated MRI Reconstruction
• Unsupervised Deep Learning Methods for Biological Image Reconstruction and Enhancement
• Assessment of Data Consistency through Cascades of Independently Recurrent Inference Machines for fast and robust accelerated MRI reconstruction
• Semi-Supervised Learning of Mutually Accelerated MRI Synthesis without Fully-Sampled Ground Truths
• Transfer Learning Enhanced Generative Adversarial Networks for Multi-Channel MRI Reconstruction