The proposed research aims, in Phase I and Phase II, to develop and clinically evaluate software tools for rapid and reproducible measurement of tumor-related volumes visible in 3-D medical image data. The goal is to provide better accuracy and reproducibility than existing manual and semi- manual methods of measurement, and to do so without penalty of increased user interaction time. Improved accuracy would mean that change in disease status could be measured more sensitively, thereby enabling shorter delays between therapy and followup examinations. Thus, the tools could enable more responsive management of therapy. The proposed approach will employ intensity (including multiparameter MRI intensity), texture, connectivity and morphologic criteria in defining tumor margins. The problem is made tractable by selection of a relatively small volume, the tumor, which the user specifies by depositing a """"""""seed"""""""" using a graphics cursor. In the Phase I research, the ability to accurately place contours in 2-D slice images will be tested. Algorithms will be optimized for agreement of computer generated contours with contours traced by experts. The extension of the technique to three dimensions is planned for Phase II.
