Gradient-based 2D/3D rigid registration of fluoroscopic X-ray to CT

Harel Livyatan

Master of Science Thesis, 2003

School of Engineering and Computer Science
The Hebrew University of Jerusalem

Abstract

Computer-aided intraoperative navigation systems aim to improve the
surgeon's hand/eye coordination and spatial perception, to improve the
accuracy of the surgical gestures and implant placements, to reduce the
cumulative radiation exposure to the surgeon, and to shorten surgery time.
In orthopaedics, preoperative CT and intraoperative fluoroscopic X-ray
images are the imaging modalities of choice for navigation.  To achieve the
desired precision, the fluoroscopic X-ray images must be calibrated and
corrected for distortion.  To incorporate the CT during the surgery and
thus allow true spatial navigation, the CT must be registered to the
intraoperative patient situation.  One of the most sought after
registration methods is anatomy image-based rigid registration which uses
intraoperative fluoroscopic X-ray images to compute the registration in a
non-invasive manner.  This thesis presents practical algorithms for C-arm
calibration and image-based rigid registration.

The first part of the thesis presents a new on-line, automatic X-ray
fluoroscopic C-arm calibration method for intraoperative use. The method
utilizes a custom-designed calibration ring with a two-plane pattern of
fiducials that attaches to the C-arm image intensifier, and an on-line
calibration algorithm.  The algorithm is robust, fully automatic, and works
with images containing anatomy and surgical instruments which cause
fiducial occlusions.  It consists of fiducial localization, distortion
correction, and camera calibration.  Our experimental results show
submillimetric accuracy for calibration and tip localization with occluded
fiducials.

The second part of the thesis presents a gradient-based method for rigid
registration of a patient preoperative CT to its intraoperative situation
with a few fluoroscopic X-ray images obtained with a tracked C-arm. The
method is non-invasive, anatomy-based, requires simple user interaction,
and includes validation.  It is generic and easily customizable for a
variety of routine clinical uses in orthopaedic surgery.

Gradient-based registration consists of three steps: 
1. initial pose estimation; 
2. coarse geometry-based registration on bone contours, and; 
3. fine Gradient Projection Registration on edge pixels.  

It optimizes speed, accuracy, and robustness.  Its novelty resides in using
volume gradients to eliminate outliers and foreign objects in the
fluoroscopic X-ray images, speed up computation, and achieve higher
accuracy.  It overcomes the drawbacks of intensity-based methods, which are
slow and have a limited convergence range, and of geometry-based methods,
which depend on the image segmentation quality.

Our simulated, in-vitro, and cadaver experiments on a human pelvis CT, dry
spine, dry femur, fresh lamb hip and human pelvis under realistic
conditions show a mean 0.5--1.7mm (0.5--2.6mm max) target registration
accuracy.