Sci‐YIS Fri ‐ 01: A protocol for the validation of non‐linear image registration systems

Abstract

An important step in the image guided adaptive radiotherapy (IGAR) process is the registration of medical images. Image registration has been used in clinically for a number of years; however registration systems have been restricted to linear or rigid registration, meaning that they cannot take into account soft tissue or organ motion with respect to rigid bony structures. Among its applications, non‐linear or deformable registration will allow for more accurate delineation of tumours and critical structures by correcting for organ motion and patient miss‐alignment from image study to study. Since deformable registration is still in its infancy, a standard protocol for the validation of these systems does not exist. A comprehensive protocol to assess the accuracy of deformable registration systems over a wide range of clinical and research applications has been developed. The protocol has been applied to the Reveal‐MVS Fusion Workstation from Mirada Solutions Ltd. It consists of a preliminary phantom study designed to assess the registration of images with well‐defined objects that have known positions, sizes, and shapes. In addition, a collection of novel and established metrics are used to determine image registration accuracy for both, real and simulated patient images. Results show that the Reveal‐MVS system is well suited for some applications of non‐linear image registration, but not applicable for others. Results will be used to further refine and improve upon existing non‐linear image registration algorithms.