Binary Classification of Melanoma Skin Cancer using SVM and CNN

Abstract

Skin cancer is seen as one of the most hazardous form of cancers found in humans. Malignant Melanoma is a deadly and a dangerous type of skin cancer. Most skin cancers either spread to other parts of the body and are fatal unless identified and treated early. Medical technology has shown advancement in computer aided diagnosis systems which can classify dermoscopic images. In this paper, we propose two methods for the detection of Skin Cancers particularly with image data taken for melanoma cancerous cells. One is using Convolutional Neural Networks with three layers and the second one is simple model of Support Vector Machines with the default RBF kernel. After applying the image processing techniques, the extracted feature parameters are used to classify the image as Benign or Malignant. The calculation metrics are accuracy, ROC curve and the AUC and confusion matrix. The classification accuracy obtained using SVM classifier is 79.39% and AUC is 0.81. CNN is computed for 100 epochs and the accuracy obtained is 84.39%. The CNN model is bought to deployment in form of a web app with the help of Streamlit.