Early Melanoma Detection and Classification Using CNN and Confusion Matrix Analysis

This article brings up a new method for detecting and classifying skin cancers using Convolutional Neural Networks (CNN) and Confusion Matrix analysis. The main work of this work evolves around detection of Melanoma and Non Melanoma cells. The large dataset on skin cancers is used to teach the CNN model to accurately classify different types of skin lesions into different malignant and non-cancerous groups. The integration of Confusion Matrix, in this case, allows for accurate identification of the classification errors made by the model as well as possible areas for improvement hence making it possible to comprehensively evaluate how well the model performs. Therefore, early recognition is indispensable to successful treatment as skin cancer is a typical yet fatal condition. Results shows that the suggested proposed model achieves 95.5% accuracy and performance in comparison with other methods found in literature. Additionally, Modified DenseNet201 model has a sensitivity of 95.96% as well as a specificity of 98.03%.