Neutro Swin-GNN: a neutrosophic set-based hybrid swin transformer and graph neural network model for monkeypox diagnosis

Monkeypox is a zoonotic viral disease caused by the Monkeypox virus, belonging to the Orthopoxvirus genus. Since the infection primarily manifests as viral dermatological lesions, visual assessment is a crucial part of clinical diagnosis. Therefore, the development of image processing-based machine learning systems for the automated detection of monkeypox-related skin lesions is highly important, not only for medical image analysis but also for computational approaches to diagnosing viral diseases. In this study, a hybrid deep learning model, Neutro-Swin-GNN, enhanced with the Neutrosophic Set (NS) approach, is proposed for the discrimination of Mpox from similar skin lesions (chickenpox, measles, and normal skin). The model addresses the problems of low contrast and similarity in medical images by combining the uncertainty modeling capability of NS with the hierarchical feature extraction of Swin Transformer and the relational feature learning capability of Graph Neural Networks (GNN). In the experimental study conducted on the Monkeypox Skin Images Dataset (MSID), 99.14% accuracy, 99% macro avg, and 98% weighted avg metrics were obtained with our proposed approach. In the comparative analyses conducted in this study, our model was compared with MobileNetV2, ResNet50, Swin-V2, MobileNetV3, EfficientNetV2, and ShuffleNetV2 models, demonstrating superior performance over these models and being optimized with Focal Loss, particularly for class imbalance. This approach presents a unique contribution to the literature by integrating NS, Swin Transformer, and GNN for the first time, and a very competitive result is obtained compared to the studies in the literature using this MSID dataset. KEYWORDS: Monkey Pox, Neutrosophic Set, Swin Transformer, Graph Neural Networks, Deep Learning