Bangladesh's Daffodil University using open-source AI for COVID-19 test with x-ray images

Researchers at Daffodil International University in Bangladesh are using an open-source Artificial Intelligence technology that can diagnose COVID-19 by using chest x-ray images.

The university’s Department of Public Health, AI Unit, and Daffodil Group’s Cardio-Care Specialized and General Hospital have jointly launched the system with a 96 percent success rate, according to the researchers.

The Directorate General of Health Services has cautiously welcomed the initiative saying that more analysis is needed before the technology can be put to use.

The researchers started working on the technology two and a half months ago after the novel coronavirus emerged in China and a lack of testing kits began straining the public heathcare system the world over, Assistant Professor Sheikh Muhammad Allayar, head of the university’s Department of Multimedia and Creative Technology, told bdnews24.com.

They analysed digital copies of chest x-ray images of COVID-19 patients from the World Health Organization and the Health Directorate “to train the software”, he said.

As the AI software has the chest x-ray data of healthy people, or COVID-19 patients or those with other diseases, it can easily detect coronavirus infection, he explained.

The university has opened a website (http://covid19.helpus.ai/covid_19_public) for the people to use the facility. But they will need approval of doctors of government-designated hospitals for uploading chest x-ray report for the final test.

Allayar said they showed the system to State Minister for ICT Zunaid Ahmed Palak and the Health Directorate.

“We will start contacting the hospitals once we get the approval. The hospitals will also need clearance,” he said.

Habibur Rahman, the line director of DGHS’ information division who saw the university’s presentation, said the system was still lengthy.

“A patient needs to go to the hospital first for the x-ray test. Then a doctor will upload the results. But those at high risk need the results fast,” he said.

“However, we welcome such initiatives. We will need to analyse it further,” Habibur said.