Machine learning tools promises faster deliverance and improved accuracy of medical diagnostics. A team of researchers from Faculty of Applied Science and Engineering at University of Toronto has designed a new and effective training program for AI specifically created for diagnostics purposes,
By using large data set of X-ray images which shows medical conditions, the team trained the machine learning’s neural network to distinguish the ailments on other X-ray images. But producing massive amount of data about rare medical conditions is impossible. To solve this challenge, they generated artificial X-rays through deep convolutional generative adversarial network, also known as DCGAN. The combined organic and synthetic X-rays images promote better and deeper AI understanding of medical conditions.
The training resulted to raise of 20% analysis accuracy on common diseases and 40% analysis accuracy on rare conditions; another medical discovery through technology aiming to improve humanity’s health.