A Robust Deep Feature Fusion Model for Skin Cancer Classification

Skin cancer ranks among the deadliest forms of cancer worldwide, leading to numerous fatalities annually. Hence, the development of efficient computer-aided diagnosis (CAD) systems has become a pressing need to help medical professionals classify various types of skin cancer. This paper introduces a novel method which combines the outcomes of two deep learning models, leveraging the feature fusion strategy. Our approach utilizes Xception and Google’s Big Transfer (BiT-M) model as base learners, complemented by a squeeze and excitation attention module to enhance the fused feature maps. We also establish a combined loss function to optimize our model. We have evaluated the model on the challenging ISIC 2017 dataset and obtained an impressive accuracy of 0.795, surpassing the performance of many recently published methods within the literature. The source code of the proposed model is available at https://github.com/Cmatermedicalimageanalysis/ISIC2017 .