Evaluation of a multimodality image (CT, MRI and PET) coregistration procedure on phantom and head and neck cancer patients:: accuracy, reproducibility and consistency

Background and purpose: In radiotherapy planning of head and neck tumors, coregistration of various anatomical (e.g. CT and MRI) and functional (e.g. PET) images is a promising way to improve the delineation of the target volumes. In this paper, we report data on accuracy, reproducibility and consistency of an interactive coregistration procedure in a specifically designed phantom and in a group of patients with primary head and neck tumors. Patients and methods: A phantom and a group of four patients with pharyngo-laryngeal tumors were imaged by CT scan (taken as the reference image), MRI (T1- and T2-weighed sequences) and PET (transmission and FDG emission). Sets of images were coregistered using an interactive rigid coregistration method based on interactive surface segmentation. Translational and rotational displacements relative to the reference CT scan were measured and expressed in terms of accuracy, reproducibility (inter- and intra-observer variation) and consistency (between T1- and T2-weighed MRI). Results: Coregistration accuracy was in the range of 0.8-6.2 mm and 1.2-4.6 mm for the phantom and the patients, respectively. Accuracy was slightly worse in the z direction, and was significantly correlated with the spatial resolution of the imaging modalities, at least for the phantom. Inter- and intra-observer variations were very small and always far below the residual variance. Consistency was perfect except in the y direction. Conclusion: Providing adequate set-up is chosen, accurate coregistration of CT, MR and FDG-PET images can be obtained in the head and neck area. Coregistration was consistent and highly reproducible among observers. (C) 2003 Elsevier Ireland Ltd. All rights reserved.