Doctors in Wuhan, capital of Hubei province, have developed an AI-enabled model that can diagnose Covid-19 patients based on the sound of their lungs.
Zeng Hesong, professor at the department of cardiology of Tongji Hospital of Tongji Medical College, Huazhong University of Science and Technology and leader of the research team, said by submitting data on the sound of patients' hearts and lungs, the model has a more than 95 per cent success rate of determining whether the patients have Covid-19 and a more than 95 per cent success rate in differentiating the seriousness of their disease.
The model is based on the sound of 172 Covid-19 patients' lungs collected at the hospital from April 1 to 5 and that of 50 non-Covid patients as the control group.
After the team submitted the data to the model and told the model what the different data represented, it used deep learning and AI to distinguish whether any new patients have Covid-19 and how serious their symptoms were, he said.
The model has great significance in fast diagnosis and early intervention of Covid-19 patients, he said, adding that the research team is applying for a patent for the model.
Compared with blood tests and nucleic acid tests, listening to the sounds of a patient's lungs is easier, quicker and noninvasive, he said.
"We plan to build the model into wearable devices that can record and analyse the patients' lung sounds automatically," he said. "However, we still lack enough patient data to make the model more accurate."
It currently takes the model from 20 to 30 minutes to diagnose a patient. The team aims to half the time required to make the diagnosis, he said.
It is working to obtain more data from Covid-19 patients to improve the model, he said.
"When enough lung and heart sounds of asymptomatic patients are collected, we believe the model can also be used to find asymptomatic patients," he added.
For the latest updates on the coronavirus, visit here.