AI software successfully predicts patients susceptible to severe COVID-19

A computer programme was recently trained to predict the severity of COVID-19 infections among patients by analysing chest X-rays. The artificial intelligence software analysed thousands of X-rays to predict which Covid patients would end up developing severe complications in the next four days. These predictions were 80 per cent accurate!

The programme was developed by New York University’s Grossman School of Medicine. It analyses hundreds of GB of data from over 5,224 X-rays of 2,943 severely sick Covid patients. The study based on the programme was published on Digital Medicine and cited a “pressing need” to devise ways that may assist in predicting severe cases of Covid among patients.

Doctors still don’t understand why the health of many patients suddenly deteriorates after catching COVID-19. They are then put on intensive life support, increasing the chances of fatality.

Also read: Covid antibodies last 8 months after infection: Study 

The NYU team fed information corresponding to X-ray data to the software which included the age, race, gender of patients. In addition, it included information like weight, body temperature, and blood immune cell levels.

From this they developed mathematical models which learned from examples to deduce who would require a mechanical ventilator and how many patients survived (2,405) the disease as opposed to those (538) who succumbed.

Also read: Risk for erectile dysfunction six-fold higher in men with COVID-19

From March 3 to June 28, 2020 the researchers tested the ability of the software to correctly predict the severity of the virus on 770 chest X-rays from patients admitted in the emergency room at NYU hospitals. The software was able to correct predicted four out of five patients who required intensive care and ventilator assistance and died within four days of admission.

ANI quoted study’s co-lead researcher Farah Shamout as saying as that emergency room doctors are in dire need of such tools that may help them predict who is likely to fall sick in the aftermath of developing COVID-19.

This way, “health care providers can monitor them more closely and intervene earlier," Shamout said.

(With inputs from agencies)