Skin cancer is a significant global health concern. Early and accurate detection is crucial for enhancing patient outcomes. This study conducts an in-depth literature review to identify commonly used Convolutional Neural Network (CNN) variants, datasets, and key evaluation metrics to assess their performance in classifying benign and malignant skin lesions. Widely used CNN architectures, including ResNet, EfficientNet, DenseNet, AlexNet, VGG, GoogleNet, LeNet-5, Xception, and MobileNet were implemented. A comparative analysis is conducted based on metrics such as accuracy, precision, sensitivity, recall, and F1-score, highlighting the strengths and limitations of each algorithm. The results show that VGG-16 outperforms other models with an accuracy of 97%, followed by VGG-19 and Mobilenet-v2 with 88%. Lastly, this paper highlights the trade-offs between various metrics, providing critical insights for deploying AI-based skin cancer detection algorithms in clinical practice.

Skin cancer particularly melanoma is significant global health concern. Early and accurate detection is very important for improving patient outcomes. This study conducts an in-depth literature review to identify commonly used CNN variants, datasets, and key evaluation metrics to assess their performance in classifying benign and malignant skin lesions. Widely used Convolutional Neural Network (CNN) architectures including ResNet, EfficientNet, DenseNet, AlexNet, VGG, GoogleNet, LeNet-5, Xception, and MobileNet were implemented. A comparative analysis is conducted based on metrics such as accuracy, precision, recall, sensitivity, and F1-score, highlighting the strengths and limitations of each algorithm. The results demonstrate that VGG-16 emerged as the best performer with an accuracy of 97%, followed by VGG-19 and Mobilenet-v2 with 88%. Lastly, this paper highlight the trade-offs between various metrics that provides critical insights for deploying AI-based skin cancer detection algorithms in clinical practice.