[ad_1]
A new Artificial Intelligence Artificial Intelligence (AI) system helps researchers identify rare medical conditions in medical images.
Researchers at the University of Toronto have developed a new system of artificial intelligence. generated X-rays that increase AI training sets, which could improve the speed and accuracy of medical diagnoses.
"In a sense, we use machine learning to learn by machine," Shahrokh Valaee, professor at the Edward S. Rogers Sr. Department of Electrical and Computer Engineering (ECE) at 39. University of Toronto, said in a statement. "We are creating simulated X-rays that reflect some rare conditions so that we can combine them with real X-rays to have a database large enough to train neural networks to identify these conditions in other X-rays. " 1965 AI00 "has the potential to help in many ways in the field of medicine," he added. "But to do that, we need a lot of data – the thousands of labeled images that we need to run these systems simply do not exist for some rare conditions."
Researchers used a convolutive generalized network, a technique that continually generates and improves simulated images.
Adverse Generator Networks (GANs) are a type of algorithm consisting of a network that generates images and another network that attempts to discriminate synthetic images from actual images. The two networks are formed so that the discriminator can not differentiate the real images from the synthesized images.
After developing sufficient amounts of artificial X-rays, they are combined with real x-ray images to form a deep convolutive neural network. The network then clbadifies the images as normal or identifies a number of conditions.
"We have been able to show that artificial data generated by a deep convolutional GAN can be used to augment real datasets," says Valaee. "This provides a greater amount of data for training and improves the performance of these systems in identifying rare conditions."
In the tests, the researchers compared the accuracy of the augmented dataset with the original dataset. The team found that the accuracy of clbadification was improved by 20% for current conditions
They also found that under some of the rarer conditions, the accuracy improved up to 40%.
Another advantage of the new system is the synthesized radiographs do not come from real people, the dataset can be easily accessible to researchers outside the hospital without violating confidentiality concerns. "It's exciting because we have managed to overcome an obstacle in the application of artificial intelligence to medicine. Showing that these augmented data sets help to improve the accuracy of the clbadification", said Valaee. "In-depth learning only works if the volume of learning data is large enough and it's a way to make sure we have neural networks capable of clbadifying images with a great precision. "
[ad_2]
Source link
