A Tumor Image Feature Agumentation Algorithm based on Improved Convolutional Neural Network
Zhengqi Ding, Gang Sun, Huanqing Xu · 2021 IEEE 5th Information Technology,Networking,Electronic and Automation Control Conference (ITNEC) · 2021
Cell image-based disease prediction model is an important research in the field of artificial intelligence, which is based on the assumption of converting tumor images into matrices for classifying and learning matrix features in neural networks. However, the differences in the incidence of different diseases lead to insufficient data from medical samples, making it difficult train efficient and accurate prediction models. Therefore, the small data problem becomes an urgent problem for scientists to solve. In this paper, we propose a feature augmentation algorithm based on an improved visual geometry group (VGG19) neural network, which can augment the size and quality of the medical image dataset and avoid the overfitting problem of disease prediction models due to insufficient medical image dataset. We obtain the feature-augmented dataset for traditional machine learning model. The feature augmentation algorithm set out in the present paper improves in accuracy, f-score, and recall metrics.