Skin cancer classification: A comparative analysis of ResNet 50 and Hybrid model(Xception+Efficient Net+GCN)

Skin cancer, characterized by the abnormal proliferation of skin cells due to excessive UV light exposure, includes common types such as squamous cell carcinoma, basal cell carcinoma, and melanoma. Early detection and accurate classification are essential for effective diagnosis and treatment. This study presents a comparative analysis between a newly developed hybrid model and various pre-trained models for efficient skin cancer image classification. The analysis includes popular pre-trained convolutional neural network (CNN) architectures such as ResNet-50 and a hybrid model combining Xception, EfficientNet, and Graph Convolutional Networks (GCN) for enhanced feature extraction and classification. These models were fine-tuned using the Skin Cancer ISIC 2020 dataset from Kaggle, comprising 11,396 images, with 9,200 images used for training (5,200 benign and 4,000 malignant) and 2,196 images for testing (1,100 benign and 1,096 malignant). The models' performances were evaluated based on runtime, accuracy, precision, recall, and F1 score. The ResNet-50 model achieved an accuracy of 83.3% with a runtime of 43 seconds, while the hybrid model demonstrated notable advancements in classification performance. Additionally, a user-friendly interface was developed using the Flask framework, allowing users to easily test and validate the model's performance.