A Hybrid Convolutional Neural Network and Graph Convolutional Networks framework for Effective Skin Spot Classification

Accurate classification of skin spots as benign or malignant is crucial for early diagnosis and treatment of skin cancer. In this paper, we propose a hybrid approach that combines Convolutional Neural Networks (CNNs) for feature extraction with Graph Convolutional Networks (GCNs) for classification. The CNN is used to extract high-level features from skin spot images, while the GCN is employed to classify the images by leveraging the relationships between the extracted features. We evaluate the performance of the proposed CNN-GCN model on a publicly available skin spot dataset, comparing it with a traditional CNN model used for direct classification. The results show that the CNN-GCN model outperforms the basic CNN approach in terms of accuracy, precision, recall, and F1-score, achieving a validation accuracy of 91.3%. This demonstrates the effectiveness of combining CNNs and GCNs for skin spot classification. Our work highlights the potential of graph-based methods in medical image classification and opens further avenues for improving model performance through optimization, present finetuning, feature selection, and more elaborated data augmentation techniques for better classification.