
On the left of each quadrant is a real X-ray image of a patient’s chest and beside it, the syntheisized X-ray formulated by the DCGAN. Under the X-ray images are corresponding heatmaps, which is how the machine learning system sees the images. Credit: Hojjat Salehinejad/MIMLab
A new artificial intelligence (AI) system trained on artificial x-rays is helping researchers identify rare medical conditions in medical images.
Researchers from the University of Toronto have developed a new AI system that enables machine learning to create computer generated X-rays that augment AI training sets, which could improve the speed and accuracy of medical diagnostics.
“In a sense, we are using machine learning to do machine learning,” Shahrokh Valaee, a professor in the Edward S. Rogers Sr. Department of Electrical & Computer Engineering (ECE) at the University of Toronto, said in a statement. “We are creating simulated X-rays that reflect certain rare conditions so that we can combine them with real X-rays to have a sufficiently large database to train the neural networks to identify these conditions in other X-rays.
“AI has the potential to help in a myriad of ways in the field of medicine,” he added. “But to do this we need a lot of data—the thousands of labelled images we need to make these systems work just don’t exist for some rare conditions.”
The researchers used a deep convolutional generative adversarial network (DCGAN), a technique that generates and continually improves simulated images.
Generative adversarial networks (GAN) are a type of algorithm made up of a network that generates the images and another network that tries to discriminate synthetic images from real images. The two networks are trained so that the discriminator cannot differentiate real images from synthesized ones.
After sufficient amounts of artificial X-rays are developed, they are combined with real X-ray images to train a deep convolutional neural network. The network then classifies the images as either normal or identifies a number of conditions.
“We’ve been able to show that artificial data generated by a deep convolutional GANs can be used to augment real datasets,” Valaee said. “This provides a greater quantity of data for training and improves the performance of these systems in identifying rare conditions.”
In testing, the researchers compared the accuracy of the augmented dataset to the original dataset when fed through the new system. The team found that the classification accuracy was improved by 20 percent for common conditions.
They also found that in some of the rarer conditions, the accuracy improved by up to 40 percent.
Another benefit of the new system is because the synthesized X-rays are not from real individuals, the dataset can be readily available to researchers outside the hospital without violating privacy concerns.
“It’s exciting because we’ve been able to overcome a hurdle in applying artificial intelligence to medicine by showing that these augmented datasets help to improve classification accuracy,” Valaee said. “Deep learning only works if the volume of training data is large enough and this is one way to ensure we have neural networks that can classify images with high precision.”