| Summary: | Many approaches to the task of computing medical image registration have been presented, but there is little knowledge of how to evaluate the quality of the models of transformation these approaches use, or accuracy of the computed parameters. There is the well known technique of comparing the reference image to a transformed secondary image, but evidence to support or deny accuracy of the transformation is still hard to obtain.Many approaches to the task of computing medical image registration have been presented, but there is little knowledge of how to evaluate the quality of the models of transformation these approaches use, or accuracy of the computed parameters. There is the well known technique of comparing the reference image to a transformed secondary image, but evidence to support or deny accuracy of the transformation is still hard to obtain.This paper presents a technique to give researchers and clinicians clear visual evidence to validate the accuracy of 2D and 3D registration of medical images. It works by borrowing the computer vision technique of optical flow to compute disparities between the reference image and the transformed secondary image. The resulting disparity information may be presented as a needle diagram to assist with communication of results via paper, or used in subsequent steps of a registration algorithm.
|
|---|