Diagnosis and analysis of repeatable diseases based on deep intelligent model driven by big data of medical images

Abstract Repetitive medical examinations often occur when patients transfer hospitals in the process of illness diagnoses, which not only lead to excessive medical expenses, but also affect people's lives. Considering this problem, this paper aims to provide an algorithm basis for the medical image standardization platforms of regional medical centers that may be established in the future. With the help of Deep Learning (DL) algorithms such as convolutional neural networks and generative adversarial networks, a medical image standardization algorithm is constructed, which can standardize the medical imaging data taken in different regions, realize mutual recognition of medical imaging data among different medical institutions, decrease the amount of unnecessary medical imaging examinations, reduce the patients' economic burden of medical examination and alleviate the tension between doctors and patients. To verify the ability of the standardization algorithm, we design two experiments. A variety of learning algorithms are used to reconstruct images under different noises in the experiments. The results show that the deep model has a higher reconstruction quality. In addition, the deep model selected for diagnosing a variety of diseases is also significantly higher in accuracy than the results of oncologists. Both experiments prove that the deep adversarial network has a higher reconstruction and diagnostic accuracy, which can be used to assist diagnosis and reduce repetitive examinations.