Robust Image Registration with Absent Correspondences in Pre-operative and Follow-up Brain MRI Scans of Diuse Glioma Patients

2025-05-03 0 0 5.6MB 10 页 10玖币

侵权投诉

Robust Image Registration with Absent

Correspondences in Pre-operative and Follow-up

Brain MRI Scans of Diﬀuse Glioma Patients

Tony C. W. Mok and Albert C. S. Chung

The Hong Kong University of Science and Technology, Hong Kong, China

{cwmokab,achung}@cse.ust.hk

Abstract. Registration of pre-operative and follow-up brain MRI scans

is challenging due to the large variation of tissue appearance and missing

correspondences in tumour recurrence regions caused by tumour mass ef-

fect. Although recent deep learning-based deformable registration meth-

ods have achieved remarkable success in various medical applications,

most of them are not capable of registering images with pathologies. In

this paper, we propose a 3-step registration pipeline for pre-operative

and follow-up brain MRI scans that consists of 1) a multi-level aﬃne

registration, 2) a conditional deep Laplacian pyramid image registration

network (cLapIRN) with forward-backward consistency constraint, and

3) a non-linear instance optimization method. We apply the method to

the Brain Tumor Sequence Registration (BraTS-Reg) Challenge. Our

method achieves accurate and robust registration of brain MRI scans

with pathologies, which achieves a median absolute error of 1.64 mm

and 88% of successful registration rate in the validation set of BraTS-Reg

challenge. Our method ranks 1st place in the 2022 MICCAI BraTS-Reg

challenge.

Keywords: Absent correspondences ·Patient-speciﬁc registration ·De-

formable registration

1 Introduction

Registration of pre-operative and follow-up images is crucial in evaluating the

eﬀectiveness of treatment for patients suﬀering from diﬀuse glioma. However,

this registration problem is challenging due to the missing correspondences and

mass eﬀect caused by resected tissue. While many recent deep learning-based

deformable registration algorithms [21,2,3,10,9,13,14,15] are available, only a few

learning-based methods [5] address the missing correspondences problem. In this

paper, we propose a 3-step registration pipeline for pre-operative and follow-

up brain MRI scans that consists of 1) a multi-level aﬃne pre-alignment, 2) a

conditional deep Laplacian pyramid image registration network (cLapIRN) with

forward-backward consistency constraint [17,16,18], and 3) a non-linear instance

optimization with inverse consistency. We validate the method using the pre-

operative and follow-up images brain MRI scans in the Brain Tumor Sequence

Registration Challenge (BraTS-Reg) challenge [1].

arXiv:2210.11045v2 [eess.IV] 12 Nov 2022

2 Tony C. W. Mok and Albert C. S. Chung

2 Related Work

Accurate registration of pre-operative and post-recurrence brain MRI scans is

crucial to the treatment plan and diagnosis of intracranial tumors, especially

brain gliomas [7,20]. To better interpret the location and extent of the tumor

and its biological activity after resection, the dense correspondences between

pre-operative and follow-up structural brain MRI scans of the patient ﬁrst need

to be established. However, deformable registration between the pre-operative

and follow-up scans, including post-resection and post-recurrence, is challenging

due to possible large deformations and absent correspondences caused by tu-

mor’s mass eﬀects [4], resection cavities, tumor recurrence and tissue relaxation

in the follow-up scans. While recent deep learning-based deformable registra-

tion (DLDR) methods [2,10,9,6,13] have achieved remarkable registration perfor-

mance in many medical applications, these registration approaches often ignored

the absent correspondence problem in the pre-operative and post-recurrence im-

ages. To address this issue, we extend our deep learning-based method described

in [17] by introducing aﬃne pre-alignment and non-linear instance optimiza-

tion as post-processing to our method. DIRAC leverages conditional Laplacian

Pyramid Image Registration Networks (cLapIRN) [16] as the backbone net-

work, jointly estimates the bidirectional deformation ﬁelds and explicitly locates

regions with absent correspondence. By excluding the regions with absent corre-

spondence in the similarity measure during training, DIRAC improves the target

registration error of landmarks in pre-operative and follow-up images, especially

for those near the tumour regions.

3 Methods

We propose a 3-step registration pipeline for pre-operative and follow-up brain

MRI scans which consists of 1) a gradient descent-based aﬃne registration

method, 2) a deformable image registration method with absent correspondence

(DIRAC), and 3) a non-linear instance optimization method. Let Band Fbe

the pre-operative (baseline) scan Band post-recurrence (follow-up) scan deﬁned

over a n-D mutual spatial domain Ω⊆Rn. Our goal is to establish a dense non-

linear correspondence between the pre-operative scan and the post-recurrence

scan of the same subject. In this paper, we focus on 3D registration, i.e., n= 3

and Ω⊆R3.

3.1 Aﬃne Registration

Although all MRI scans provided by the challenge are rigidly registered to the

same anatomical template [1], we found that there are large linear misalignments

between the pre-operative and follow-up images in cases suﬀering from serious

tumor mass eﬀect. To factor out the possible linear misalignment between MRI

scans Band F, we register T1-weighted Band Fscans using the iterative aﬃne

registration method.

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：

RobustImageRegistrationwithAbsentCorrespondencesinPre-operativeandFollow-upBrainMRIScansofDiuseGliomaPatientsTonyC.W.MokandAlbertC.S.ChungTheHongKongUniversityofScienceandTechnology,HongKong,Chinafcwmokab,achungg@cse.ust.hkAbstract.Registrationofpre-operativeandfollow-upbrainMRIscansischallengingdu...

展开>> 收起<<

Robust Image Registration with Absent Correspondences in Pre-operative and Follow-up Brain MRI Scans of Diuse Glioma Patients.pdf

共10页,预览2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！

Robust Image Registration with Absent Correspondences in Pre-operative and Follow-up Brain MRI Scans of Diuse Glioma Patients

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: