AI can check coronavirus CT scans in 20 seconds with 96% accuracy

A CT scan can hold the answer as to whether a patient is infected with the novel coronavirus. Whitening, unusual spots or hazy increases in lung density are all cues of COVID-19.

The signs are pretty clear, but here's the problem – A doctor spends about 15 minutes analyzing one case. Luckily, with the help of AI, CT scans can be interpreted in 20 seconds with an accuracy rate of 96 percent.

So how does AI work?

Convolutional neural network is an important branch of AI. The algorithm can automatically recognize clinical telltale signs by combing through over 5,000 CT scans.

Simply put, the algorithm learns new input just like children, except it is much faster and can be trained in a week.

First, the image is automatically analyzed using a deep learning classifier to detect abnormalities. After that, lung images are segmented and finally, a massive extraction of texture features is applied.

With that, AI can instantly distinguish between the lungs of a patient with common viral pneumonia and that of a COVID-19 patient.

AI can also directly calculate the number and size of lesions and determine how severe the case is.

This move can boost the efficiency of the double checks on diagnosis.

It was first put into use in central China's Henan Province, and nearly 100 more hospitals in China are currently eyeing this technology.

If you want to know more about the mind-blowing technology being used in the fight against the coronavirus, stay tuned and subscribe to CGTN accounts.