Intelligence in Image and Signal Processing (CIISP 2007) Parameterization of Dermoscopic Findings for the Internet-based Melanoma Screening System

Abstract — In this paper, we develop an automated method to recognize the dermoscopic criteria used for diagnosing melanomas as defined by two commonly used diagnostic schemes, namely the ABCD rule and the 7-point checklist. We use a database of 105 dermoscopy images and their dermoscopic findings determined by four dermatologists as the gold standard. We extract 356 objective image parameters from the images and build multiple regression models for the 15 clinical findings defined by the aforementioned diagnostic schemes. Our model provides comparable results to dermatologists in recognizing almost all of the findings. The results of our preliminary experiments show that computer-based image analysis has the potential of identifying the dermoscopic criteria. Index Terms — dermoscopy, computer-aided diagnosis (CAD), melanoma, ABCD rule, 7-point checklist, feature extraction I