Enhancing Skin Cancer Detection: A Deep Learning Approach with Augmented CNNs

Skin cancer can be a highly fatal disease, and early detection is crucial to saving lives. Conventional diagnosis involves the need for technicians who may be slow to examine or who might misread notes. Deep learning has its success in medical image analysis due to its accuracy and efficiency. This research proposed a CNN-based deep learning method to enhance the detection accuracy. These were a basic CNN, a model with Data Augmentation and dropout layers and one trained on a class-balanced dataset. The simplest CNN exhibited high variance, where training accuracy was at 86.66% whereas validation accuracy was very low (at $52.80 \%$). With data augmentation and dropout layers, the model achieved $56.36 \%$ train and $54.57 \%$ validation accuracy, slightly improving generalization. At the end of the training, the best model has max result of $93.19 \%$ on training data and $89.50 \%$ on validation data which was trained with using Augmentor on class-balanced data. Their findings indicate that balancing the dataset and data augmentation enhance CNN performance in skin cancer detection. Future explorations may utilize pre-trained deep learning models so as to enhance accuracy and diagnose patients as well as assist in their care.