Skin cancer classification: A comparative analysis of EfficientNet and Hybrid model(Xception+Efficient Net+GCN)

Skin cancer, primarily caused by abnormal skin cell growth due to excessive exposure to UV radiation, includes common forms such as squamous cell carcinoma, basal cell carcinoma, and melanoma. Early and accurate detection is crucial for effective treatment. This study compares the performance of a newly developed hybrid model with several well-established pretrained convolutional neural networks (CNNs) for skin cancer image classification. The analysis focuses on popular CNN architectures, including EfficientNet, alongside a hybrid model that integrates Xception, EfficientNet, and Graph Convolutional Networks (GCN) to enhance feature extraction and classification accuracy. Both models were fine-tuned using the Skin Cancer ISIC 2020 dataset from Kaggle, comprising 11,396 images. Of these, 9,200 images (5,200 benign and 4,000 malignant) were used for training, while 2,196 images (1,100 benign and 1,096 malignant) were reserved for testing. Evaluation metrics such as runtime, accuracy, precision, recall, and F1 score were employed. The EfficientNet model achieved an accuracy of 77% with a runtime of 43 seconds. However, the hybrid model demonstrated significantly better performance, achieving 90% accuracy and greater efficiency. Additionally, a user-friendly interface was developed using the Flask framework to allow users to easily test and validate the model's performance.