In a pandemic, time is of the essence. Quick and accurate testing for Covid-19 can save thousands of lives and curb the spread of infection across borders. But the shortage of testing equipment and lack of rapid testing systems have resulted in lockdowns, paralyzing economies in many countries worldwide. Beijing-based medtech Infervision has adapted its AI-powered screening system for lung disease detection to identify Covid-19 infection within 10 seconds, by analyzing a patient's chest CT scans.
“The beauty of AI is that we can train the AI [solution] and deploy it to other places even faster than the spread of the virus itself, especially in some rural and remote areas where local healthcare workers have never seen any Covid-19 cases before,” said Infervision CEO and founder Chen Kuan in an interview with CNN. “But AI can help prepare them for the battle."
Founded in 2015, Infervision uses deep learning AI to generate a semi-structured report of abnormal areas in the CT scans to be tagged for closer review by radiologists and other medics. The system was mainly used to detect early symptoms of lung cancer like tubercles, a small nodular lesion in the lungs or other tissues.
In mid-January, Infervision staff noticed a surge in usage for the less frequently used function of pneumonia detection. The discovery spurred the medtech to quickly adapt its AI-screening capabilities to detect Covid-19 infections using chest CT scans.
AI assistant for frontline doctors
One of Infervision's first clients to trial the new screening system was a frontline hospital in Wuhan, the epicenter of the Covid-19 outbreak in China and where the crisis started. From Wuhan Tongji Hospital, Infervision quickly introduced the new Covid-19 screening solution to other hospitals in Wuhan, the first city to lockdown in mid-January to slow down the spread of the virus across China.
The rapid spread of the virus took everyone by surprise and became more acute with death tolls rising in hospitals as health workers struggled to cope with the exponential rise in infections. The use of personal protective equipment (PPE), needed to protect medics in close contact with seriously ill patients, created another problem for doctors. “The protective goggles constantly became misty, and sweat would blur vision too,“ said Zhou Hongbin, Vice-President of Marketing at Infervision.
Infervision's AI platform can analyze a patient's chest CT scan to produce a diagnosis within 10 seconds, and even faster if more servers were deployed. The solution also enables doctors to compare a patient's CT scans taken at different dates to keep track of each patient’s progress more quickly and accurately.
By March 12, doctors were using Infervision to analyze over 80,000 chest CT scans to identify over 6,000 infected patients
The startup launched its Covid-19 screening solution on January 31, the use of which was rolled out across China. The medtech teams had had to speed up working to upgrade the platform during the Chinese Spring Festival, sending out staff to set up servers at hospitals across the country in February in spite of lockdowns of cities and airports, paralyzing transport links. Since then, many medics have been trained to use and customize the new screening system for their own hospitals and patients.
By March 12, doctors were using Infervision to analyze over 80,000 chest CT scans to identify over 6,000 infected patients with accuracy rates hitting 98%. Although the first-hand medical database originated from Covid-19 patients in Wuhan, the company has actively deployed its proprietary software to help patients in other parts of China and overseas. By April 7, over 190,000 cases had been diagnosed in China and other countries.
From Wuhan to the US, Italy
The analysis of chest X-rays or CT scans can help doctors to diagnose Covid-19 cases faster than the more widely used polymerase chain reaction (PCR) test. It can also provide more information, especially for patients showing only mild symptoms. Covid-19 patients often exhibit several unique features in the CT scan results compared to normal pneumonia. During its early stages, the scans will show generally patchy infiltrates and changes in lung periphery. Only experienced doctors are able to notice the difference.
But less-experienced medics, dealing with Covid-19 for the first time, can use Infervision to help them to quickly identify the unusual lung changes in the CT scans of their patients. Each chest CT scan also generates hundreds of digital images from different angles. Generally, the doctor would have to spend a long time to go through them one by one.
“If I see each image for three seconds, I will still need 10–30 minutes to go through one patient’s result,” said Xia Liming, a radiologist at Wuhan Tongji Hospital. The pressure of saving lives also makes doctors want to check the results more carefully.
One of the soldiers spoke to us in English: Thanks for helping Italy in this dangerous time!
In early March, Infervision started working in Yokohama with its first overseas partner Doctor Net, a Japanese enterprise with the largest team of certified radiology specialists in the country. Japan has the highest number of CT machines in the world but it had only 6,000 registered radiologists, below the average of other developed countries. Infervision was able to solve the shortage issue by deploying its AI-screening solution in the Japanese hospitals.
On March 23, Infervision’s office in the US was contacted by Thomas Jefferson University Hospital. Within a week, the AI-screening system was set up and used by medics to test and monitor Covid-19 patients at the hospital.
In Europe, the company also quickly responded to a request from Unicampus Bio-medicine Rome Hospital in Italy. Due to the sudden lockdown and travel restrictions imposed by the authorities, Infervision had to send a team via Germany, driving 12 hours from Frankfurt to the Italian hospital on March 16.
“We arrived in Italy at 2 am. At that time, Italy had closed its borders,” recalled Sun Yipeng, head of Infervision's operations in Europe. "We were pulled over by Italian soldiers at the foot of the Alps. We showed them the necessary documents and explained that we were going to help the hospital in Italy.
"One of the soldiers spoke to us in English: Thanks for helping Italy in this dangerous time! We were so moved. Infervison is fulfilling its mission of helping human beings with advanced technology. This is exactly where our value lies.”