Convolutional Neural Network Based Diagnosis System on Skin and Breast Cancers

Abstract

Use of artificial intelligence in medicine makes a difference in diagnosis methods. A diagnosis system based on deep neural network can efficiently make predictions for many known diseases. Our study is to construct a cancer diagnosis system using CNN models. The cancer diagnosis system is capable of giving predictions on skin cancer and breast cancer with input images. The diagnosis model for skin cancer is AlexNet, and the model for breast cancer is VGGnet. Based on the two pre-trained CNN models, we use PyQt5 to develop the user interface and construct the diagnosis system. According to the test result, the skin cancer diagnosis model achieves about 80% accuracy, and the breast cancer model achieves about 85% accuracy. As for the diagnosis system, users can upload at most three images, select cancer type, and view the analysis results on the interface. In conclusion, our diagnosis system can accurately and efficiently present skin and breast cancer diagnosis results.