Detection of infectious skin diseases in children's nurseries using AI model

الملخص

Infectious skin diseases present a significant health concern in children, particularly in densely populated environments such as nurseries. Early diagnosis is critical to preventing complications and improving healthcare outcomes. This study develops an image-based artificial intelligence (AI) model using the Mpox Skin Lesion Dataset v2.0 to detect infectious conditions such as measles, chickenpox, and hand-foot-mouth disease. Three convolutional neural network (CNN) architectures—ResNet50, VGG19, and a custom CNN—were trained and evaluated. The custom CNN achieved the highest validation accuracy at 88.70%, while VGG19 demonstrated superior generalization performance with an accuracy of 83.33% on unseen real-world clinical images. ResNet50 exhibited overfitting, underscoring the importance of selecting appropriate transfer learning strategies. The proposed system does not aim to replace healthcare professionals but rather to serve as a supportive diagnostic tool in pediatric and nursery settings. These findings suggest AI models can significantly enhance early screening processes and improve children's health outcomes.