Article Preview
Top1. Introduction
Among various forms of skin diseases, skin cancer is the most severe disease in current times. Various studies reveal that skin cancer starts with a change in normal skin. The reason for skin cancer and its associated diseases is the skin cells that grow and metastasise unusually. There are various forms of skin cancers such as Melanoma, Basal, and Squamous – Melanoma. Of these, Squamous-Melanoma is the most dangerous one. As per the American Cancer Society, in 2019, in the USA, approximately 96,480 emerging cases of melanoma were detected, affecting 39,260 women and 57,220 men. Also, a projected death toll of 7,230 due to melanoma was predicted - consisting of 4,740 men and 2,490 women (Siegel, 2019).
Melanoma moles have uneven fringes and develop hues, for example, blue, red, darker shades of red and blue, white and pink, indicating the malady's seriousness. Basal cell carcinoma is a bit silvery knob, translucent, and regularly with surface macules. As the lesion increases in size, it generally ulcerates to make a moving edge and follow outside. Squamous cell carcinoma is typically showed up in perpetual sunlight based on harms scalp, dorsum of the hand, lower lip, lower arm, and ear. It begins with little and crusted plaque and moves toward becoming indurated and modular. Observation at an early stage for change in normal skin can lead to preventing any skin disease. The changes can be size, color, shape, or new occurrence of a mole or feel. There are varieties of prevention or detection systems, starting from manual to automated systems. The biopsy was one of the traditional methods in which a part of the affected skin area is taken for further analysis. However, the process of biopsy is painful and affects the patient psychologically as well. Also, dermoscopy tools are used. These again have some harmful impacts in terms of exposure of skin to laser beams. On the other hand, differently trained clinicians sometimes lack accuracy and analysis. Therefore, Computer-Aided Diagnostic System (CAD) comes into image processing to control these challenges.
A computer-aided diagnostic system is an automated system for the detection of various kinds of skin diseases. There are various powerful approaches to the CAD system, and one of the approaches includes stages like pre-processing, feature extraction, segmentation and finally, classification of a lesion. Segmentation has a powerful impact on the quality of the CAD system as it has to cover all aspects of the lesion area as edges, contour, shape, sharpness, and many more. The significance of the visual representation of the external body for medical importance exists, consistently expanding around the ongoing years, focusing more on developing automated segmentation methods for more accuracy of the lesion (Oliveira, 2016; Sreena, 2019).
The ISIC 2018 challenge that supports participants to develop image processing tools to classify various types of skin diseases from dermoscopic images (Codella, 2019). It also provides many dermoscopic images of both training input data and training ground-truth response data for researchers. It comprises three tasks related to the lesion as boundary segmentation, attributes detection and diagnosis. In the first task of this challenge, the highest performance achieved is 84%, like the Jaccard Index (Qian, 2018).