Prediction of Skin Cancer Based on Convolutional Neural Network

Skin cancer is one of the most lethal cancers in the world, which is a common type of cancer. And the early diagnosis of it is quite important. Aiming to improve the classification of skin cancer, a modified convolutional neural net- work model is proposed. Firstly, a convolutional layer and a pooling layer is added on the original network, constructed an 8-layer convolutional neural network. And then the network output value and the data label by the backward propagation adjust the network parameters. Lastly, by constantly debugging the network, the best model for the recognition network is selected. The proposed network applied to the dataset consisting of 900 cases of melanoma with the malignancy status pathologically confirmed. The results of the revised network prediction accuracy got 91.92% on training set and 89.5% on test set. The results obtained demonstrate that the proposed method is performant and can be used to predict whether the skin cancer is benign or malignant. With the development of deep learning and the clinical application of medical images, this method is expected to become a diagnostic tool for doctors to diagnose skin cancer.