2D-3D image registration in diagnostic and interventional X-Ray imaging

Abstract

Clinical procedures that are conventionally guided by 2D x-ray imaging, may benefit from the additional spatial information provided by 3D image data. For instance, guidance of minimally invasive procedures with CT or MRI data provides 3D spatial information and visualization of structures that are not visible with x-ray. Since 3D imaging modalities may not be available during the procedure or require increased patient dose, it is desirable to use pre-interventional/preoperative 3D patient data for guidance that was obtained for diagnosis and treatment planning. To accomplish this, a relationship between patient and image data has to be realized. This relationship can be obtained by 2D-3D image registration. The aim of the research presented in this thesis is to evaluate the performance and limitations of 2D-3D image registration for diagnostic and interventional x-ray imaging, and to develop new methods to overcome these limitations. In Chapter 2 we have compared the registration performance of seven optimization methods in combination with three similarity measures: gradient difference, gradient correlation, and pattern intensity. Optimization methods included in this study were: regular step gradient descent, Nelder-Mead, Powell-Brent, Quasi-Newton, nonlinear conjugate gradient, simultaneous perturbation stochastic approximation, and evolution strategy.

Registration experiments were performed on patient data sets that were obtained during cerebral interventions. Various component combinations were evaluated on registration accuracy, capture range, and registration time. The results showed that for the same similarity measure, different registration accuracies and capture ranges are obtained when different optimization methods are used. Overall, it could be concluded that the Powell-Brent is a reliable optimization method for intensity-based 2D-3D registration of x-ray images to CBCT, in terms of accuracy, capture range, and computation time.

In Chapter 3, a relatively computationally inexpensive and robust estimation method is proposed with the objective to enlarge the capture range of 2D-3D registration methods. The method uses the projection-slice theorem in combination with phase correlation in order to estimate the transform parameters, which provides an initialization of the subsequent registration procedure. The method was evaluated as an initialization method for an intensity-based 2D-3D registration method. The uninitialized and initialized registration experiments had success rates of 28.8% and 68.6%, respectively. With these results it was shown that the initialization method based on the projection-slice theorem and phase correlation yields adequate initializations for existing registration methods, thereby substantially enlarging the capture range of these methods. Although spatial information provided by CT during image-guided procedures is substantial, the soft tissue contrast this modality accommodates is rather limited. Availability of MRI data during x-ray-guided procedures may provide such soft tissue contrast. However, registering x-ray images to MRI data is not a trivial task because of their fundamental difference in tissue contrast. In Chapter 4 a technique is presented that generates pseudo-CT data from MRI data, which is sufficiently similar to real CT data as to enable registration of MRI to x-ray. The method was evaluated by comparing registration performance of x-ray to pseudo-CT data to the registration performance of x-ray to real CT data. The results showed that pseudo-CT data facilitates registration of x-ray images to MRI, and that the accuracy achieved was comparable to that of registering x-ray images to CT data. In Chapter 5 the effects of perspective imaging on radiographic measurements were investigated.

Radiographic diagnosis and follow-up studies of developmental dysplasia of the hip are commonly done by measuring the acetabular index on radiographs using Hilgenreiner’s method. The outcome of the measurement, however, depends on the orientation of the subject’s pelvis relative to the x-ray source. The influence of combinations of pelvic rotation and tilt on the systematic error in the acetabular index measurement was investigated in a reproducible way. Additionally, two ratios Rrotation and Rtilt, evaluating pelvic rotation and tilt, respectively, were measured. The study was done by using digitally reconstructed radiographs of a highresolution 3D CT data set. The results show that the effects of rotation and tilt accumulate and either amplify or counteract the underestimation or overestimation of the acetabular index. For rotations and tilt up to 12 degrees the average systematic errors in the acetabular index varied from 8.8 degrees◦ underestimation to 4.5 degrees overestimation. In order to limit the systematic error caused bypelvic misalignment, we advise to consider radiographs acquired with ±4 degrees ◦ rotation and ±4 degrees tilt as acceptable. In our research, these pelvic orientations correspond to Rrotation values between 1.0 and 2.0 and Rtilt values between 1.1 and 1.8. In Chapter 6, the feasibility of obtaining ischemic lesion volumes from CBCT-based CBV measurements was investigated. The reliability of CBCT-assessed CBV was evaluated by characterization of image noise of a flat-panel c-arm CBCT system. CBV of seven canines with unilateral stroke was assessed. Ischemic lesion volumes were calculated by using various lesion thresholds. Volume measurements were evaluated by a regression analysis using gold standard data obtained with histology. From the results it could be concluded that in order for this technique to be applicable in clinical practice, correlation between lesion volumes measured with CBCT-based CBV and histology should be improved. The quality of CBV measurements are limited by imaging physics, and can be improved by reducing detector noise and photon scatter.