When medical images meet generative adversarial network: recent development and research opportunities

Medical Research
DOI: 10.1007/s44163-021-00006-0 Publication Date: 2021-09-22T06:03:02Z
ABSTRACT
Abstract Deep learning techniques have promoted the rise of artificial intelligence (AI) and performed well in computer vision. Medical image analysis is an important application deep learning, which expected to greatly reduce workload doctors, contributing more sustainable health systems. However, most current AI methods for medical are based on supervised requires a lot annotated data. The number images available usually small acquisition annotations expensive process. Generative adversarial network (GAN), unsupervised method that has become very popular recent years, can simulate distribution real data reconstruct approximate GAN opens some exciting new ways generation, expanding methods. Generated solve problem insufficient or imbalanced categories. Adversarial training another contribution imaging been applied many tasks, such as classification, segmentation, detection. This paper investigates research status analyzes several commonly this area. study addresses both synthesis other tasks. open challenges future directions also discussed.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (107)
CITATIONS (58)