Head and neck multimodality volumes visualization methods

This paper describes design strategies and algorithms which have been developed for visualization. registration, Volume of Interest (VoI) definition and measurements on multimodality single subject head and neck CT, MR, PET and surgical tumor specimen images. This work was part of a project that measured the usefulness of MR and PET images in three-dimensional radiotherapy treatment planning which was originally based only on CT images. The VoIs. which could be defined on any of the registered volume. were exported with the CT images to the treatment planning system, using DICOM. The major problem was the handling of large datasets with big differences in FOV and voxel sizes. We used object-oriented design and C++ programming language to develop object classes and methods for efficient data management. We used the eXtensible Markup Language (XMEL) standard to save the registration transformers and the VoIs. The registered volumes were resliced on the fly for visualization or VoI statistical information computing. Therefore, we avoided creation of large resliced dataset files. We used the new classes to extend two existing applications: interactive registration and RoI drawing. The paper describes: (i) the data handling problems.. (ii) the developed objects relationship and methods and (iii) the two applications.