Classification of Skin Lesions using Squeeze and Excitation Attention based Hybrid Model of DenseNet and EfficientNet

Skin cancer, like all other types of cancer, is a health problem where early diagnosis saves the patient's life. Successful detection of skin lesions in dermoscopic images enables this process to be managed more quickly. Deep learning architectures have advanced significantly in recent years, particularly in the classification of medical images. For the multi-class classification of skin lesions, a hybrid model was developed in this study utilizing the convolutional neural network architectures of DenseNet and EfficientNet. In addition, SE (Squeeze and Excitation) attention mechanism is integrated into both models. Experimental results show that the inclusion of SE attention mechanism provides a noticeable improvement in classification accuracy. This study not only contributes to the development of effective skin lesion classification systems, but also It also provides valuable insights into the synergistic effects of combining neural network architectures with attention mechanisms.