A Hybrid machine learning approach for Melanoma cancer stage detection

Skin cancer that spreads quickly and exceedingly dangerous is melanoma. It is the skin cancer that kills the most people, which makes it the deadliest. When a patient is diagnosed, classifying their cancer is a very significant and time-consuming procedure. The primary factors influencing the cancer diagnosis during the surgical therapy phase are the tumor thickness and cancer stage. To detect melanoma cancer and initiate treatment in its early stages, machine learning algorithms can be employed. It helps lighten the strain on doctors, increase diagnosis accuracy, speed up analysis, and lower clinical costs. Two approaches for melanoma cancer stage classification are given in this research. In this approach, melanoma can be categorized into two stages within the first system and into three stages within the second system. The approach involves the use of Convolutional Neural Network (CNN) techniques, with the SMTP employed as the loss function. The study presents the SMTP loss function and compares it to the outcomes obtained with alternative loss functions. When compared to several existing loss functions tailored for classification tasks, the proposed technique demonstrates superior effectiveness.