Non-iterative multi-modal partial view to full view image registration using local phase-based image projections

Authors:
Ilker Hacihaliloglu;David R. Wilson;Michael Gilbart;Michael Hunt;Purang Abolmaesumi
Affiliations:
Departments of Orthopaedics, University of British Columbia, Vancouver, BC, Canada;Departments of Orthopaedics, University of British Columbia, Vancouver, BC, Canada;Departments of Orthopaedics, University of British Columbia, Vancouver, BC, Canada;Physical Therapy, University of British Columbia, Vancouver, BC, Canada;Electrical and Computer Engineering, University of British Columbia, Vancouver, BC, Canada
Venue:
IPCAI'12 Proceedings of the Third international conference on Information Processing in Computer-Assisted Interventions
Year:
2012

Citing 4
Cited 0

A Stochastic Iterative Closest Point Algorithm (stochastICP)

MICCAI '01 Proceedings of the 4th International Conference on Medical Image Computing and Computer-Assisted Intervention
Detection of planar motion objects using Radon transform and one-dimensional phase-only matched filtering

Systems and Computers in Japan
Biomechanically constrained groupwise statistical shape model to ultrasound registration of the lumbar spine

IPCAI'11 Proceedings of the Second international conference on Information processing in computer-assisted interventions
Towards real-time 3D US to CT bone image registration using phase and curvature feature based GMM matching

MICCAI'11 Proceedings of the 14th international conference on Medical image computing and computer-assisted intervention - Volume Part I

Quantified Score

Hi-index	0.00

Visualization

Abstract

Accurate registration of patient anatomy, obtained from intra-operative ultrasound (US) and preoperative computed tomography (CT) images, is an essential step to a successful US-guided computer assisted orthopaedic surgery (CAOS). Most state-of-the-art registration methods in CAOS require either significant manual interaction from the user or are not robust to the typical US artifacts. Furthermore, one of the major stumbling blocks facing existing methods is the requirement of an optimization procedure during the registration, which is time consuming and generally breaks when the initial misalignment between the two registering data sets is large. Finally, due to the limited field of view of US imaging, obtaining scans of the full anatomy is problematic, which causes difficulties during registration. In this paper, we present a new method that registers local phase-based bone features in frequency domain using image projections calculated from three-dimensional (3D) radon transform. The method is fully automatic, non-iterative, and requires no initial alignment between the two registering datasets. We also show the method's capability in registering partial view US data to full view CT data. Experiments, carried out on a phantom and six clinical pelvis scans, show an average 0.8 mm root-mean-square registration error.