A Deep Learning Approach Utilizing Covariance Matrix Analysis for the ISBI Edited MRS Reconstruction Challenge

TL;DR

This paper introduces a deep learning method that uses covariance matrix analysis to improve the speed and robustness of high-quality edited MRS scans, applicable to synthetic and real data.

Contribution

It presents a novel deep learning approach that leverages covariance matrices, invariant to transient count and noise, for enhanced MRS reconstruction.

Findings

Effective acceleration of MRS acquisition

Robust performance with noisy data

Applicable to synthetic and in-vivo scenarios

Abstract

This work proposes a method to accelerate the acquisition of high-quality edited magnetic resonance spectroscopy (MRS) scans using machine learning models taking the sample covariance matrix as input. The method is invariant to the number of transients and robust to noisy input data for both synthetic as well as in-vivo scenarios.