Auto Skin Tumour Classification Using CNN Framework with Tensorflow and Keras

Abstract

Skin tumour classification is emergent and most challenging problem in the medical diagnosis because of its similarity in the patterns of tumour cells with other symptoms & diseases on the victim body. An automatic and robust system is required for the early detection of the diseased providing aid in the field of computer assisted medical diagnosis thus decreasing the mortality rate. Machine learning and Deep Learning play a significant role in transforming the health care sector. Convolutional Neural Network (CNN) is used in healthcare sector for assisting to generate massive amount of health records, analyse thousands of health records and further provide insights on clinical decisions to service providers. In this article, we propose robust solution to detect and classify the skin tumour:  Benign vs. Malignant using CNN architecture, TensorFlow and Keras on publicly available ISIC (International Skin Imaging Collaboration) data set. The results obtained in the experimental study records 0.34% loss with accuracy of 80.72% in the validation set and 82.75% on the test set.