Computer-Assisted Image Analysis of Bronchioloalveolar Carcinoma

Abstract In clinical trials, response rate is an important endpoint for assessing the efficacy of an anticancer drug. The Response Evaluation Criteria for Solid Tumors (RECIST) has been widely used as a standard method to assess response. The RECIST requires only 1-dimensional measurement of tumor size. However, bronchioloalveolar carcinoma (BAC), which commonly presents as infiltrative or micronodular lesions, is not always readily assessable by RECIST. During the past 2 years, we have been developing computer-based programs to more accurately measure tumor size on chest computed tomography (CT) scans. In a first-generation computer- assisted image analysis (CAIA) system, we were able to capture and quantify lesions on CT scans by linking the software programs of eFilm, HyperSnap, and Scion. We have applied this CAIA approach to measuring BAC response to gefitinib in the Southwest Oncology Group (S0126) trial. However, this first-generation CAIA system involves multiple manual steps and is therefore labor intensive. We are now developing a fully automated CAIA program based on a versatile software platform, ImageJ, created at the National Institutes of Health. Taking theoretical and physical considerations into account, Java plug-in programs for ImageJ are created to automatically analyze CT scans in the Digital Imaging and Communications in Medicine format. We have demonstrated the feasibility of an ImageJ-based automated CAIA program for measuring BAC bidimensionally on CT scans. This automated CAIA system will be applied in a prospective clinical trial of the GVAX vaccine in patients with BAC.