Published: 08th May 2020
LPU develops cloud-based, AI-powered interface to detect COVID-19 from X-rays, CT scans
It can be accessed by an authorised healthcare professional anywhere in the world via a basic mobile device with internet connectivity and a web browser
Researchers have developed an easy-to-use cloud-based web interface powered by artificial intelligence (AI) that they say can detect COVID-19 quickly from radiology reports such as CT scan or X-Rays.
The system may enable doctors and medical staff even in remotest villages of the country to get quick results on the COVID-19 status of a patient, the researchers said.
AI technologies that study CT Scans and X-Rays are being deployed as part of global efforts to tackle the coronavirus pandemic, according to the researchers from Lovely Professional University (LPU) in Punjab.
However, one of the biggest roadblocks in their adoption has been the need for a high-tech device or computer to support the software, and a qualified technician for operating it, they said.
"The unavailability of resources in our healthcare system is a big hindrance in COVID-19 detection and treatment, especially in the remote areas of the country," said Prabin Kumar Das, a B.Tech student at LPU.
"We wanted to build a cloud-based system which could be accessed at minimal costs even from the remotest possible locations in the country while ensuring high quality results," Das said.
According to the World Health Organization (WHO), the novel coronavirus causes respiratory illness resembling viral pneumonia, resulting in fever, cough, and shortness of breath. A study published in the journal Radiology found that chest X-ray scan features can aid in the early detection and diagnosis of the virus.
The interface developed by researchers at LPU is cloud-based and does not need any high-tech device or technicians to operate. It can be accessed by an authorised healthcare professional anywhere in the world via a basic mobile device with internet connectivity and a web browser, they said.
The researchers noted that the user will have to upload a patient's scan on the website and it will return the diagnosis result "within a few seconds".
The interface is connected to an AI model at a centralised server which processes the image received from the web portal and sends the results back to it, they said. The researchers said no extra technical skills are required to use the portal, except for uploading the images.
The portal is capable of hosting any AI model that is preferred by a hospital, and will accordingly be able to detect the infection as well as the extent to which it has affected the patient, they said.
"Several experts from around the world have strongly recommended the use of CT scans for quick diagnosis of COVID 19. We are proud to have come up with a solution that can rapidly scale these tests," said Lovi Raj Gupta, Executive Dean of Science and Technology at LPU.
He said the team is in touch with the Indian Council of Medical Research (ICMR) for further evaluation, and is also in talks with industry partners to collaborate by providing the most promising medically-certified AI models which can be used for diagnosis of COVID19.
This, Gupta said, will help doctors to segregate patients and help them decide whether a patient has to go to triage, isolation or if they need a ventilator. The available resources can be allocated efficiently.
The images uploaded on the interface will be AES encrypted ensuring complete patient privacy, according to the researchers. It took the team, which includes B.Tech student Biswajyoti Roy, and professors Rajesh Singh and Anita Gehlot, around one month to develop the framework, they said.
The AI model that the team built for testing the framework was trained and tested over more than 13,000 chest radiography images, according to the researchers.