Full-Stack Application of Skin Cancer Diagnosis Based on CNN Model
Yiyang Huo · 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI) · 2021
Convolutional neural network (CNN) is a subset of deep neural networks, and it has commonly applied to analyze images. Skin cancer is a disease that can be observed without the help of expensive and professional instruments. At the same time, consulting a doctor in a hospital is not a cheap thing for many people. To tackle this issue, this paper introduced a skin cancer detection application based on CNN. In this application, the graphical user interface is implemented by Swift UI, and the backend is ExpressJs. It loads a keras CNN model through TensorFlow JS to detect and classify skin cancer. The CNN model used in the application is created through TensorFlow and trained with the HAM10000 skin cancer data set. It also integrated Natural Language Process (NLP) function, so that users can ask some questions and consult doctors online. The current version of this application can successfully operate in the IOS environment, and fully achieve those functions above. Also, the experimental result demonstrated that the accuracy of CNN model is around 75% for HAM10000 test set.