A Two-stage Model for Skin Cancer Detection Using Convolutional Neural Networks

Early detection of skin cancer is a critical issue in the medical field, as the success of treatment depends heavily on the time of disease detection. Traditional methods often lack the necessary accuracy and speed, failing to properly exploit the fine details of medical images, which play a crucial role in the correct and timely diagnosis of diseases. This research presents an intelligent method for the accurate analysis of skin images using convolutional neural networks. In this approach, the U۲Net  model is first used to extract skin-related areas, and then the ResNet-۱۰۱ model identifies and classifies the images with high accuracy. Also, preprocessing techniques such as hair removal and contrast enhancement are used to improve the quality of input images and increase the accuracy of the models. The proposed model was evaluated on the HAM۱۰۰۰۰ database, and the results showed that this system achieved ۹۴% accuracy in the classification section with the ResNet-۱۰۱ network and ۹۶% accuracy in the segmentation section with U۲Net. The results show that this method can be an effective tool for dermatologists and, by focusing on image details, allows for more accurate diagnosis of skin cancer. This achievement will also pave the way for further research in the field of artificial intelligence in the diagnosis of skin diseases and other diseases.

کلیدواژه ها:

Convolutional Neural Networks ، Skin cancer detection ، image processing ، U۲Net ، ResNet ۱۰۱