New AI system diagnoses diseases like flu better than doctors

The rapid spread of artificial intelligence has already been taking over the medical field. In another new development, scientists have now created an AI that can diagnose illnesses in children such as flu better than actual doctors.

A group of researchers in the US and China trained an AI on medical health records from 1.3 million visits at a major medical center in Guangzhou, China. They built an AI system that automatically diagnoses common childhood diseases as accurately as a doctor.

For training the AI, the team input health records of patients, covering appointments of children and teenagers from January 2016 to July 2017. Specific words and phrases were also fed into the system through which it was able to identify keywords like ‘vomiting’ and ‘abdominal pain’ as symptoms of diseases.

Artificial intelligence can ‘outperform doctors’ in future

When the AI was tested in previously unseen cases, it diagnosed glandular fever, influenza, chicken pox, roseola, and hand-foot-mouth disease with between 90% and 97% accuracy. When it was tested against 20 doctors, the AI system made more accurate diagnoses than junior medics, reported ABC News.

The results as published in the journal Nature Medicine, showed that the AI was over 90% accurate at diagnosing asthma, as compared to the 80%-94% accuracy range of physicians in the study. In diagnosing gastrointestinal disease, the system was 87% accurate in contrast to the physicians’ 82%-90% accuracy.

Daily Mail reported, the researchers believe that their AI system could further help sort hospital patients according to their severity and also improve the diagnosis of complex, rare conditions. The team is not working to diagnose adult diseases as well.

“Our study provides a proof of concept for implementing an AI-based system as a means to aid physicians in tackling large amounts of data, augmenting diagnostic evaluations, and to provide clinical decision support in cases of diagnostic uncertainty or complexity,” said lead author Kang Zhang.