Home» News» News» Media»

China Focus: Visualizing what COVID-19 does to lungs

MAY . 14 2020
Peking University, May 14, 2020: Chinese researchers have developed an artificial intelligence (AI) system to speed up the diagnosis of COVID-19 and get a clear picture of its effect on the lungs.

Currently, experienced radiologists need up to 15 minutes to diagnose COVID-19 from about 300 CT images.

Researchers from Visual Perception Center under the Institute of Artificial Intelligence, Peking University have developed a low-cost 3D volume rendering tool to visualize CT scans. It is expected to aid COVID-19 diagnosis and facilitate communication between patients and doctors.

Computer scientists have been trying to visualize modern CT/MR images, making them more intuitive and realistic. The 3D volume rendering tool visualizes large volumes of data generated by CT/MR scanners in three-dimensional space and different aspects of the data set can be interactively explored.

After the outbreak of COVID-19, several Chinese companies such as SenseTime rolled out AI-assisted systems to improve the efficiency and accuracy of the diagnosis.

Ma Lei, of Peking University, said high-fidelity volume rendering focuses on interactive visual realism. The colored 3D images deliver more information than black and white scans, acting as an "amplifier" of organ lesions and giving a vivid demonstration of a patient's lungs.

According to Ma, they optimized the algorithm to adapt to the physiological structure of the lungs and were able to achieve real-time rendering. Patients can have their colored 3D lung images minutes after the scan with no extra equipment. The system is able to clearly show the lung lesions and ground-glass opacity with realistic 3D movies.

Qiu Jianxing, a chief physician at the imaging and radiology department at Peking University First Hospital, said the 3D volume rendering tool would help primary care hospitals that have few experienced radiologists.

Ma noted that Chinese hospitals have relied on foreign rendering tools that are often bundled with CT machines. The rendering tool developed by Peking University is based on a self-developed ray-tracing engine which is expected to reduce the cost as well as dependence on technology imports.

With more data feedback, the system is getting more accurate and sensitive. It is expected to demonstrate the condition of the skeleton and airway, distinguish arteries and veins, and provide more precise analysis on the lesions.

Beyond COVID-19, the tool could be used to locate lung tumors and aid surgery.

Source: Xinhua News Agency