Gastric Cancer Detection from Endoscopic Images Using Synthesis by GAN

摘要

Datasets for training gastric cancer detection models are usually imbalanced, because the number of available images showing lesions is limited. This imbalance can be a serious obstacle to realizing a high-performance automatic gastric cancer detection system. In this paper, we propose a method that lessens this dataset bias by generating new images using a generative model. The generative model synthesizes an image from two images in a dataset. The synthesis network can produce realistic images, even if the dataset of lesion images is small. In our experiment, we trained gastric cancer detection models using the synthesized images. The results show that the performance of the system was improved.