Enhancing Early Detection of Skin Cancer in Clinical Practice with Hybrid Deep Learning Models

Skin cancer is a significant global health issue where early detection is essential to improve outcomes. This study evaluates hybrid deep learning models that combine CNN architectures (MobileNetV2, ResNet-18, EfficientNet-B0, and others) with metadata (age, lesion localization) for classification using the SLICE-3D subset of the ISIC 2024 dataset. MobileNetV2 achieved a recall of 99.2% and an accuracy of 97.7%, while EfficientNet-B0 demonstrated a recall of 98.5% and an accuracy of 97.2%, making them ideal for telemedicine in resource-limited settings due to their low computational demands. ResNet-18 and DenseNet-121, with recalls of 99.0% and 98.7%, respectively, excelled in clinical applications but required greater computational resources. These hybrid models show great potential as accessible and accurate tools for improving skin cancer detection. Future work should validate these findings on diverse datasets and optimize preprocessing to further enhance sensitivity and early diagnostic accuracy.