BEIJING -- Imsight Technology, a Chinese company that has been developing equipment for digital radiographic examinations of the chest since 2018, has devised an artificial intelligence system to assist computer-aided tomographic diagnostic screening of the lungs amid the ongoing coronavirus pandemic.
The new system is a deep-learning process that reads in a huge amount of data and is capable of reading images as accurately as experienced doctors. During a trial in February, it processed images at more than 99% accuracy.
In the field of AI-based diagnostic imaging, China is leading the world in providing equipment that can read huge amounts of data.
Imsight launched a project on Jan. 20 before the start of the Chinese New Year to develop an AI-based medical imaging system for the lungs. It created a prototype algorithm during the Lunar New Year and began data training after the holiday season ended in February.
The company operated the system on a trial basis and examined its accuracy at the Shenzhen No. 3 People's Hospital in the province of Guangdong on Feb. 24. It officially released it on March 17.
At present, two medial institutions in Shenzhen have adopted the system for computer-aided tomographic screening of the lungs. When the system detects a possible COVID-19 infection, it immediately specifies an affected site and conducts the necessary quantitative data analyses.
The system is capable of processing radiographs for 600 people a day, taking only 10 to 20 seconds to check each image, according to Imsight. In detecting infections related to the new coronavirus, it has proved more than 99% accurate.
Experience pays in clinical medicine. Doctors in the diagnostic imaging field become able to make highly accurate diagnoses after many years of reading large numbers of radiographs.
But when a new pandemic suddenly occurs as in the case of the coronavirus, it is difficult for doctors to share knowledge. The improvement of diagnostic ability is thus delayed.
If AI is used, prompt training based on a vast amount of data is possible. Features in radiographs are extracted and contribute to the rapid accumulation of image-reading experiences. An AI system thus assists doctors, who cannot review many cases in a short period of time, in making diagnoses and compensating for a shortage of shared knowledge when a pandemic suddenly occurs.
AI systems offer solutions to a nationwide shortage of medical workers and imbalanced distribution of resources as they provide results from more efficient and accurate screening.
The spread of COVID-19 has prompted medical technology-related companies to rapidly shift to commercial production of equipment.
Under normal circumstances, the companies focus on activities such as approaching customers. Amid the steep increase in demand for medical examinations, new services such as remote medical care, online hospitals, temperature checks by AI and delivery of drugs via the internet are rapidly moving toward practical use.
AI technology for image processing is growing faster in China than in the rest of the world, according to Xiao Xiang, Imsight's chief operating officer. This is primarily because the Chinese market is huge, and offers a large parameter for the accumulation of image data as well as an ample volume and various data sources.
In addition, China leads the U.S. and Europe in the cost of and number of workers trained in data annotation, the process of labeling data to make it usable for machine learning.
The number of radiologists is increasing at around 4% a year in China, while the market for medical imaging equipment is growing at around 30%. The imbalance in development of medical care has created extremely strong demand in some regions.
Other factors contribute to the rapid growth of AI-based diagnostic imaging in China. Many Chinese companies can operate quickly and flexibly, and they receive policy support from the government, which is promoting AI development as a national strategy.
However, there still is a long way to go before AI is adopted widely.
The development of AI-based products for diagnostic screening has focused on those related to computed tomography to detect nodular shadows in the lungs. Although it has been expanded to include digital radiography of the chest, bone fractures, strokes and mammograms, AI products are most widely used and functionally advanced when it comes to the diagnosis of lung disorders.
There is considerable room to expand their application to more diseases and boost their effectiveness.
To further develope the field, authorities will have to create policy concerning certification systems, rules for clinical trials, insurance issues, and other matters.
36Kr, a Chinese tech news portal founded in Beijing in 2010, has more than 150 million readers worldwide. Nikkei announced a partnership with 36Kr on May 22, 2019.
For the Japanese version of this story, click here.
For the Chinese version, click here.