Inception-Based U-Net in Magnetic Resonance Image Reconstruction

Date

2022-08-27

Authors

Vafay Eslahi, Elmira

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Magnetic resonance imaging (MRI) is one of the best imaging techniques that produce high-quality images of objects. The long scan time is one of the biggest challenges in MRI acquisitions. To address this challenge, many researchers have aimed at finding methods to speed up the process. Faster MRI can reduce patient discomfort and motion artifacts. One method to speed up MRI scans is skipping some signals in the k-space. Although the incomplete k-space or sub-sampling causes undersampling artifacts due to missing signals, reconstruction techniques can solve the problem by recovering the missing data. Many reconstruction methods are used in this matter, like deep learning-based MRI reconstruction, parallel MRI, and compressive sensing. Among these techniques, the convolutional neural network (CNN) generates high-quality images with faster scan and reconstruction procedures compared to the other techniques. However, CNN architecture has been an area under study in image reconstruction and needs more investigations. In this study, we propose a new deep learning algorithm for MRI reconstructions. The Inception module proposed by Google inspires this algorithm. The proposed architecture in this work creates a better-reconstructed image than the standard U-Net. The U-Net architecture consists of encoding and decoding sections, which are considered to decrease complexity. Decreasing the complexity prevents overfitting in the network and, consequently, improves image quality. In other words, we introduce a new MRI U-Net modification by using the Inception module, which is more flexible and robust compared to the standard U-Net. Mean square error (MSE), normalized mean square error (NMSE), structural similarity index measure (SSIM), and peak signal-to-noise ratio (PSNR) are the metrics that we use to evaluate the model. In this study, we improve scan speed by 3.2 times. We show that our method with a new architecture performs better than the standard U-Net to decrease the error and increase the image quality.

Description

Keywords

fastMRI, MRI, Image Reconstruction, MRI, Magnetic Resonance Imaging, Magnetic Resonance Image Reconstruction, Image Reconstruction in MRI, MRI Acceleration

Citation