Intensity-Based Image Registration Using Robust Correlation Coefficients

摘要

The ordinary sample correlation coefficient is apopular similarity measure for aligning images from the sameor similar modalities. However, this measure can be sensitive tothe presence of “outlier” objects that appear in one image butnot the other, such as surgical instruments, the patient table, etc.,which can lead to biased registrations. This paper describes anintensity-based image registration technique that uses a robustcorrelation coefficient as a similarity measure. Relative to theordinary sample correlation coefficient, the proposed similaritymeasure reduces the influence of outliers. We also compared theperformance of the proposed method with the mutual information-based method. The robust correlation-based method shouldbe useful for image registration in radiotherapy (KeV to MeVX-ray images) and image-guided surgery applications. We haveinvestigated the properties of the proposed method by theoreticalanalysis, computer simulations, a phantom experiment, and withfunctional magnetic resonance imaging data.