Generative adversarial networks (GANs) are an elegant deep learning approach to generating artificial data that is indistinguishable from real data. Two neural networks are paired off against one another (adversaries). The first network generates artificial data to reproduce real data. The second, discriminative network, is tasked with trying to decide which is real and which is artificial data. The process is repeated over and over, requiring both the generation of artificial data and the detection of artificial data to continuously improve. The importance of being able to produce realistic artificial data, synthetic data, allows for the generative network to learn what features are important to pass as real data.
Generative adversarial networks in radiology
A recent publication demonstrated the use of GANs in the detection of congestive cardiac failure on chest radiographs with an overview provided in the AI's Black Box: Cracking Open The Chest X-ray Radiopaedia blog post 3. GANs are also being used for several image quality improvement algorithms, eg. denoising algorithms 4.
Unable to process the form. Check for errors and try again.