Subject-speciﬁc quantitative susceptibility mapping using patch based deep image priors Arvind Balachandrasekaran Davood Karimi Camilo Jaimes Ali Gholipour

2025-04-26 0 0 482.26KB 5 页 10玖币

侵权投诉

Subject-speciﬁc quantitative susceptibility mapping

using patch based deep image priors

Arvind Balachandrasekaran, Davood Karimi, Camilo Jaimes, Ali Gholipour

Computational Radiology lab

Harvard Medical School

Boston, MA 02115

arvind.balachandrasekaran, davood.karimi@childrens.harvard.edu

camilo.jaimescobos, ali.gholipour@childrens.harvard.edu

Abstract

Quantitative Susceptibility Mapping is a parametric imaging technique to estimate

the magnetic susceptibilities of biological tissues from MRI phase measurements.

This problem of estimating the susceptibility map is ill posed. Regularized recovery

approaches exploiting signal properties such as smoothness and sparsity improve

reconstructions, but suffer from over-smoothing artifacts. Deep learning approaches

have shown great potential and generate maps with reduced artifacts. However,

for reasonable reconstructions and network generalization, they require numerous

training datasets resulting in increased data acquisition time. To overcome this

issue, we proposed a subject-speciﬁc, patch-based, unsupervised learning algorithm

to estimate the susceptibility map. We make the problem well-posed by exploiting

the redundancies across the patches of the map using a deep convolutional neural

network. We formulated the recovery of the susceptibility map as a regularized

optimization problem and adopted an alternating minimization strategy to solve it.

We tested the algorithm on a 3D invivo dataset and, qualitatively and quantitatively,

demonstrated improved reconstructions over competing methods.

1 Introduction

Quantitative Susceptibility Mapping (QSM) is a parametric imaging technique to estimate the

magnetic susceptibilities of biological tissues. It enables quantiﬁcation of iron concentration in gray

matter [

], myelin content in white matter [

], deoxyhemoglobin in veins [

] etc. Changes in the

concentrations reﬂect an underlying pathology, which include micro bleeds [

], hemorrhages [

] and

neurodegenerative diseases [6, 7], and can potentially be used as a bio-marker [8].

The susceptibility map is related to phase measurements through a 3D convolution with a dipole

kernel. However, the problem of estimating the map through direct inversion is ill posed and results

in streaking artifacts. To improve the quality of reconstructions, regularized recovery approaches

enforcing signal sparsity and smoothness have been introduced [

]. These approaches reduce the

streaking artifacts but introduce blurring in the recovered maps. Recently, supervised deep learning

methods have shown great potential and have generated maps with reduced artifacts [

]. However,

to achieve reasonable reconstruction quality and for the network to generalize well to unseen data,

these approaches require a lot of training data, which result in increased data acquisition time.

We propose a subject-speciﬁc, patch-based, unsupervised deep learning algorithm to estimate the

susceptibility map from phase measurements. Patch based priors have been successful in many image

restoration tasks including recovery of MR images from undersampled measurements [

]. To make

the problem well-posed, we exploit the redundancies across the different patches in the map using a

deep convolutional neural network (CNN). The recovery of the map is then posed as a regularized

Preprint. Under review.

arXiv:2210.06471v1 [eess.IV] 10 Oct 2022

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Subject-specicquantitativesusceptibilitymappingusingpatchbaseddeepimagepriorsArvindBalachandrasekaran,DavoodKarimi,CamiloJaimes,AliGholipourComputationalRadiologylabHarvardMedicalSchoolBoston,MA02115arvind.balachandrasekaran,davood.karimi@childrens.harvard.educamilo.jaimescobos,ali.gholipour@childr...

展开>> 收起<<

Subject-speciﬁc quantitative susceptibility mapping using patch based deep image priors Arvind Balachandrasekaran Davood Karimi Camilo Jaimes Ali Gholipour.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Subject-speciﬁc quantitative susceptibility mapping using patch based deep image priors Arvind Balachandrasekaran Davood Karimi Camilo Jaimes Ali Gholipour

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: