Unsupervised Domain Adaptation with Adversarial Learning for Liver Tumors Detection in Multi-phase CT Images

Automatic and accurate liver tumor detection in multi-phase CT images plays an important role in computer-aided diagnosis. A deep learning-based AI system requires a large amount of training data, but sufficient training data with high quality annotations is a major issue in the medical field. The generalization of a label-rich training domain to a new test domain causes a domain shift problem in deep learning models. Here, we present a domain adaption-based approach using adversarial learning for liver tumor detection in multi-phase CT images. We adopt knowledge for model learning from PV phase to ART and NC phase. Our method avoids the need for separate annotations for each phase. To address the domain gap between these different phases of CT images, we employ adversarial learning scheme using an anchorless object detector. The results of the experiments show that our proposed model trained with adversarial learning scheme perform much better than those trained in normal setting.