Objective assessment of 3-D medical image registration results using statistical confidence intervals

It is of great interest to provide statistical information of image registration results. The authors successfully proposed a novel automatic approach based on statistical theory to estimate confidence intervals of the parameters in 2-D registrations. Here, the theory has been extended to more sophisticated conditions in the current 3-D phantom study. Extensive Monte Carlo simulations have been conducted and the results are consistent with the calculated confidence intervals when different amounts of displacement and smoothing are presented. The predicted 95% confidence intervals have less then 20% error of their sizes in normal noise conditions. By properly removing the systematic errors, the new theory works well even when large amounts of noise and gray value inconsistency are present in the images. The present results indicate that the use of the statistical confidence intervals developed in this paper can provide an objective assessment for 3-D image registration results