Deep-Learning-regularized Single-step Quantitative Susceptibility Mapping (QSM) Quantification.

Researchers

Journal

Modalities

Models

Abstract

To develop a deep-learning-regularized single-step QSM quantification, directly generating QSM from the total phase map.A deep-learning-regularized single-step QSM quantification model, named SS-POCSnet, was trained with datasets created using the QSM synthesis approach in QSM reconstruction challenge 2.0. In SS-POCSnet, iteratively applied a data fidelity term based on a single-step model that combined the spherical mean value kernel and dipole model. Meanwhile, SS-POCSnet regularized susceptibility maps, avoiding underestimating susceptibility values. We evaluated the SS-POCSnet on N=10 synthetic datasets; N = 24 clinical datasets with lesions of cerebral microbleed (CMB) and calcification, and N = 10 datasets with multiple sclerosis (MS).On synthetic datasets, SS-POCSnet showed the best performance among the methods evaluated, with NRMSE of 37.3±4.2%, XSIM of 0.823±0.02, HFEN of 37.0±5.7, and pSNR of 42.8±1.1. SS-POCSnet also reduced the underestimations of susceptibility values in deep brain nuclei, compared with those from other models evaluated. Furthermore, SS-POCSnet was sensitive to CMB/calcification and MS lesions, demonstrating its clinical applicability. Our method also supported variable imaging parameters, including matrix size and resolution.Deep-learning-regularized single-step QSM quantification can mitigate underestimating susceptibility values in deep brain nuclei.This article is protected by copyright. All rights reserved.