Deep Learning Based System for Early Detection and Classification of Melanoma Skin Cancer- A Systematic Review

Abstract

Melanoma is the most dangerous form of skin cancer and its detection at the initial stage is very important to increase patient's survival. This paper compares the use of deep learning systems in the early identification and categorization of melanoma skin cancer. The research used a melanoma cancer dataset obtained from Kaggle which had 9600 training images, and 1000 test images divided equally between the benign and malignant classes. The pre-processed datasets are employed to train transfer learning models, including VGG16, Mobile Net, Dense Net, ResNet50 and InceptionV3. Model performance is assessed by accuracy, loss, Cohen’s kappa, precision, recall, f1-score and confusion matrix. The best performing model is shown to be deployable using Streamlit, which provides a user interface through an input image, on-the-fly pre-processing, and real-time classification. The proposed system is better than the other methods and can be a good and practical approach for melanoma detection..