Skin lesion classification using machine learning approach: A survey

Study of skin disease images through human effort, for detection of skin cancer, has always been a very difficult task. Distinguishing and manually analyzing skin lesions for detection of melanoma can be tedious as it is a lengthy process. Computational resources have advanced and are progressing in an innovative way, due to which the analysis of skin diseases, especially skin lesion, has been made easier by the help of innumerable AI strategies. The outcomes of such models, showed after implementation, are quite impressive but the drawbacks of these models include the failure in recognizing some of the skin lesion problems due to the complex skin lesion images. A complete study of procedures for distinguishing skin diseases from a healthy skin is presented in this work. This survey study will help examiners in creating effective models that automatically identify diseased skin from healthy skin images. Firstly, the difficulties in identifying skin tumor from skin images are recognized. Secondly, the pre-processing and segmentation techniques in determining various skin lesions are discussed. Thirdly, latest research comparisons are presented. Fourth, different methods for classification of skin lesion in various categories of skin tumor are examined. Lastly, the segmentation and classification process applying latest machine learning techniques utilized in well-known skin disease images examination are investigated and difficulties of skin disease analysis using ISIC 2018 and 2019 dataset are outlined.