Disocclusion-based 2D-3D registration for aortic interventions

被引：19

作者：

Demirci, Stefanie ^{[1
]}

Baust, Maximilian ^{[1
]}

Kutter, Oliver ^{[1
]}

Manstad-Hulaas, Frode ^{[2
,3
]}

Eckstein, Hans-Henning ^{[4
]}

Navab, Nassir ^{[1
]}

机构：

[1] Tech Univ Munich, D-80290 Munich, Germany

[2] Norwegian Univ Sci & Technol NTNU, Dept Circulat & Med Imaging, Trondheim, Norway

[3] St Olavs Hosp Trondheim, Dept Surg, Trondheim, Norway

[4] Tech Univ Munich, Klinikum Rechts Isar, Dept Vasc Surg, D-80290 Munich, Germany

来源：

COMPUTERS IN BIOLOGY AND MEDICINE | 2013年 / 43卷 / 04期

关键词：

2D-3D registration; Medical image registration; Image dissimilarity; Robust similarity measures; Disocclusion; IMAGE REGISTRATION; 3-D/2-D REGISTRATION; NORMALIZATION; VALIDATION; CT;

D O I：

10.1016/j.compbiomed.2013.01.012

中图分类号：

Q [生物科学];

学科分类号：

07 ; 0710 ; 09 ;

摘要：

Occlusions introduced by medical instruments affect the accuracy and robustness of existing intensity-based medical image registration algorithms. In this paper, we present disocclusion-based 2D-3D registration handling occlusion and dissimilarity during registration. Therefore, we introduce two disocclusion techniques, Spline Interpolation and Stent-editing, and two robust similarity measures, Huber and Tukey Gradient Correlation. Our techniques are validated on synthetic and real interventional data and compared with well-known approaches. Results prove that an integration of disocclusion into the registration procedure yield higher accuracy and robustness. It is also shown that the robust measures have different effects depending on the type of occluding structure. (c) 2013 Elsevier Ltd. All rights reserved.

引用

页码：312 / 322

页数：11

共 37 条

[1] Cost function masking during normalization of brains with focal lesions: Still a necessity? [J].