Deep Ensemble Learning for Skin Lesion Classification from Dermoscopic Images

Abstract

Skin cancer is one of the leading causes of death globally. Early diagnosis of skin lesion significantly increases the prevalence of recovery. Automatic classification of the skin lesion is a challenging task to provide clinicians with the ability to differentiate between different kind of lesion categories and recommend the suitable treatment. Recently, Deep Convolutional Neural Networks have achieved tremendous success in many machine learning applications and have shown an outstanding performance in various computer-assisted diagnosis applications. Our goal is to develop an automated framework that efficiently performs a reliable automatic lesion classification to seven skin lesion types. In this work, we propose a deep neural network-based framework that follows an ensemble approach by combining ResNet-50 and Inception V3 architectures to classify the seven different skin lesion types. Experimental validation results have achieved accurate classification with an assuring validation accuracy up to 0.899.