Status Report of Practical Melanoma Diagnostic System on the Internet

Abstract

In the past few decades, the incidence of malignant melanoma has increased gradually in most parts of the world. Although advanced malignant melanoma is often intractable, early-stage melanoma is curable in many cases. Therefore, early detection and diagnosis of melanoma is essential to the reduction of melanoma-related deaths. We have established a practical internet-based melanoma screening and data collection system. Key components of the system are a new dermatologist-like tumor area extraction algorithm and a neural network classifier. The dermatologist-like tumor area extraction algorithm achieved superior extraction performance and the neural network classifier achieved classification accuracy of 97.3% in sensitivity and 86.1% in specificity with leave-one-out cross-validation test of 319 dermoscopy images. Furthermore, our system supports SSL transaction and requires only several seconds to complete a procedure. Now we are preparing to make our system into commercial viability. Our current issues are advancement of the system generality and development of trade edition of expensive dermoscopy. In this report, we report the current status of our diagnostic system.