An unsupervised medical image registration network for intelligent medical education

Intelligent education, which applies artificial intelligence technologies to the education field, has attracted increasing attention in recent years. Intelligent medical education is an effective learning way to assist medical practitioners in mastering complex medical knowledge through artificial intelligence technologies like big data analysis and machine learning. With the development of medical imaging technology, the types of medical images are becoming abundant, such as computed tomography image and magnetic resonance imaging image. Medical experts with long clinical experience can analyze the patient’s condition from different types of medical images to achieve more accurate diagnosis. Nevertheless, cultivating an expert who can accurately analyze different types of medical images is time-consuming and laborious. Therefore, it is necessary to utilize the intelligent medical education way to design medical image processing method that can guide medical practitioners in simultaneously accurately analyzing different types of medical images. The medical image registration, which can automatically align different types of medical images and guide medical practitioners in diagnosing, has become popular. However, existing methods lack labeled training data and ignore interactive information between registered images. To solve these problems, we propose a cross-view interactive network (CVIN). First, CVIN utilizes alternate stages with information interaction modules to extract data features, which can use unlabeled medical data to train the model. Second, we introduce an InfoMax module to learn the mutual information between registered images. The experiment results show that CVIN is promising and can use to guide the medical practitioners to conduct accurate disease diagnosis.