Skin Melanoma Assessment with Machine-Learning Approach—A Study

Abstract

Recently, a number of biomedical image examination techniques are proposed and employed to assess a variety of clinical images. This work examines the RGB scale skin melanoma (SM) images obtained from the benchmark skin cancer dataset. The difficulty in SM picture is it should be examined in its RGB case and the complexity will be more due to the R, G, and B histogram. The aim of the proposed study is to suggest all the possible procedures to implement a machine-learning methodology to examine the dermoscopy pictures with greater accuracy. The proposed work discusses various procedures to be followed to attain better evaluation of the SM images using the machine-learning techniques. The need and the assessment of the ABCD rule are also demonstrated with relevant results. In this work, the infected skin section is initially preprocessed and extracted by implementing a thresholding and the segmentation section. Later, the segmentation section is examined with the ABCD rule. Further, suggestion to implement the possible feature extraction and classifier training is also given to classify the existing dermoscopy pictures as melanoma and non-melanoma cases. The results confirm that the proposed suggestions can be adopted to examine the clinical dermoscopy pictures.