Skin Cancer Detection (SCAM-D)

Abstract

Melanoma is defined as a cancer that is incurable in its advanced stages, emphasizing the need of early detection and treatment. Various procedures and tools have been employed to diagnose this type of cancer early, practically all of which needed a visit to the doctor and were not available to the general public. This work presents an automated and accurate approach for distinguishing between benign skin pigmented lesions and malignant melanoma that may be used by the general population and does not require special imaging equipment or conditions. After preprocessing the input photos, the region of interest is segmented using the Otsu method in this study. Then, on the segmented image, a new feature extraction is used to mine the advantageous qualities. The method is then completed by categorizing the data using an optimized Deep Believe Network (DBN) into two classes: normal and melanoma cases. To achieve improved efficacy in different terms, the optimization procedure in DBN was carried out by a developed version of the recently presented Defense Test and evaluation Organization (DTEO) method. The method's performance is compared to 7 distinct procedures from the literature to demonstrate its superiority.