CNNs and Melanoma: The Perfect Pair for Accurate and Timely Cancer Staging

According to international World Cancer Research Fund, in the year 2020, 324635 patients were diagnosed with melanoma which led to 57043 deaths. This huge 18% death rate could be significantly reduced if this deadly disease is detected at an early stage. Although being the rarest form of skin cancer, melanoma is the most lethal and is to responsible for 75% of skin cancer fatalities. The main factor affecting melanoma patients' chances of survival is the thickness of the tumour. The most common way to record it is as a millimetre (mm) depth measurement obtained by pathological testing after a biopsy of the suspicious lesion. We suggest a dermoscopic image-based computational image analysis system to do away with the need for an invasive method of determining the thickness of melanoma prior to surgery. Clinical findings that relate particular dermoscopic picture characteristics to tumour depth are the foundation for the proposed feature extraction. In this paper, we examine and analyse various methods for classifying various stages of melanoma. We compare the effectiveness of the CNN and SVM algorithms for melanoma stage classification. By comparing all the approaches, we can select the approach that is the most successful for this issue and carry on our research using that model.