How can AI help doctors fight coronavirus?

It’s a global pandemic.
We need all the help we can get
for efficient and effective control of the pandemic.
Even from AI.
Icon arrow down
virus objectvirus object

An AI tool that supports coronavirus management

Online and Free of Charge

At Lunit, we develop AI software that can assist radiologists in the interpretation of chest x-rays. The software, Lunit INSIGHT CXR, can discover multiple radiologic findings including lung consolidation, which indicates possible coronavirus (COVID-19) infected pneumonia.

In light of the current outbreak, we have released a special version of Lunit INSIGHT CXR that can support healthcare providers better manage coronavirus. The solution is currently being used to manage coronavirus at hospitals in South Korea and Brazil, providing assistance in patient triage and monitoring.

A medical worker at coronavirus care center located near Daegu, South Korea, is examining a patient’s chest x-ray image with Lunit INSIGHT CXR.

How We Help

Keep Hospitals Running With AI

The current coronavirus pandemic is burdening health care providers around the world with massive demands for tests, diagnosis, and treatment. While PCR is the test for definitive diagnosis of coronavirus infection, it can take upto 2 days, if not more, delaying key decision-making for patients. This can be a big problem as delayed diagnosis may lead to further spread of disease.
Chest x-ray can be a fast, effective, and affordable test to evaluate coronavirus-related pneumonia. We suggest the use of AI as a supportive tool for the interpretation of chest x-ray exams, as below.

Suggested Use of Lunit INSIGHT CXR
in Coronavirus Response

  • Use as a supportive tool during case overload, which can likely lead to low reading quality
  • Use with portable x-ray devices to fast-track test results and decision-making in advance to the PCR results
  • Use when regular monitoring is required among patients showing mild symptoms, in order to identify and/or triage patients by the progression of symptoms

While ACR and CDC currently do not recommend the use of CT or chest x-ray taken in an enclosed space for the fear of infection via air and surface, ACR states that when medically necessary, facilities may consider deploying portable radiography units since it relatively has lower risk of infection.

Publications

AI Performance in COVID-19 Settings Validated in Peer-Reviewed Journals

Implementation of a Deep Learning-Based Computer-Aided Detection System for the Interpretation of Chest Radiographs in Patients Suspected for COVID-19

Published in Korean Journal of Radiology, Jul 17, 2020
Eui Jin Hwang, MD, PhD, Hyungjin Kim, MD, PhD, Soon Ho Yoon, MD, PhD, Jin Mo Goo, MD, PhD and Chang Min Park, MD, PhD
Department of Radiology, Seoul National University College of Medicine, Seoul, Korea.

According to the study, radiologists with the assistance of Lunit INSIGHT CXR identified patients with rRT-PCR-positive COVID-19 or pneumonia on CXR with a reasonably acceptable performance. Sensitivity and specificity of AI-aided CXR interpretation for rRT-PCR-positive COVID-19 cases were 68.8% and 66.7%, and for CT abnormalities suggesting pneumonia was 81.5% and 72.3%, respectively.

In addition, the turnaround times of AI-aided CXR reports were significantly shorter than those of rRT-PCR results, with 51 and 507 minutes, respectively.

The findings of the study demonstrate that AI-aided CXR interpretation may assist with timely clinical decision-making and management of patients suspected for COVID-19. Patients with positive radiographs could be subject to enhanced isolation, which would minimize the transmission of COVID-19 while waiting for rRT-PCR results.

Read the full version

Deep-Learning Algorithms for the Interpretation of Chest Radiographs to Aid in the Triage of COVID-19 Patients: A Multicenter Retrospective Study

Published in PLOS ONE, Nov 24, 2020
Se Bum Jang, Suk Hee Lee, Dong Eun Lee, Sin-Youl Park, Jong Kun Kim, Jae Wan Cho, Jaekyung Cho, Ki Beom Kim, Byunggeon Park, Jongmin Park, Jae-Kwang Lim

According to the study, Lunit INSIGHT CXR showed a satisfactory performance comparable with that of radiologists in the CR-based diagnosis of pneumonia in COVID-19 patients. The sensitivity and specificity of Lunit AI algorithm were 95.6%, and 88.7%, respectively, while those of radiology reports were 91.2% and 96.9%, respectively.

The study demonstrates the automatic interpretation of the CR with Lunit INSIGHT CXR can significantly reduce the burden and workload of medical staff and radiologists in a sudden surge of suspected COVID-19 patients.

Furthermore,AI-aided CR can offer fast and reliable examinations, facilitating early screening and isolation of COVID-19 suspected patients, which is the most basic and important response strategy in emergency departments.

Read the full version

Presentation by Prof. Dong Eun Lee from the Chilgok Kyungpook University Hospital at the Korean Congress of Radiology meeting 2020

Use Cases

South Korea and Brazil

In South Korea, Lunit INSIGHT CXR has been deployed to assist the diagnosis of patients in Daegu and nearby region, the hardest-hit area where approximately 8,000 patients—about 85% of the entire domestic cases—has been diagnosed with coronavirus. (as of 30th March, 2020)

One of the major university hospitals in Seoul has built a healthcare center in Daegu, where physicians remotely examine patients and provide consultations via video conference call. The patients’ chest x-ray images are sent directly to radiologists in Seoul, where Lunit INSIGHT CXR is used to support radiologists instantly detect and diagnose coronavirus infection.

The system enables prompt and early discovery of patients, sending them directly to larger hospitals for proper treatment if required.

A healthcare professional in Seoul is examining a patients’ chest x-ray using Lunit INSIGHT CXR.

In Brazil, Lunit INSIGHT CXR has been installed at PreventSenior, one of the largest hospital networks in Brazil with 8 locations throughout the metropolitan region of Sao Paulo. This institution is one of the COVID-19 detection centers that use chest x-ray screening for patients with mild symptoms.

As of late March, the institution has deployed Lunit INSIGHT for the analysis of more than 3,000 chest x-ray images suspected of coronavirus infection. Medical workers here are telling us that Lunit INSIGHT CXR is providing great help “especially for patient triage,” as the hospital is “overflowing with patients while the number of radiologists remains low.”

A medical worker at one of the COVID-19 centers in Brazil is using Lunit INSIGHT CXR to examine a chest x-ray image.

Helping Online

Providing Free COVID-19 Analysis

To support healthcare professionals in this pandemic, we are providing free online and instant AI analysis for the detection of radiologic findings suggestive of coronavirus-related pneumonia on chest x-rays. We hope this free access can work as a supportive tool for healthcare providers in the diagnosis and treatment of coronavirus.

The case upload is limited to 20 cases per day per user, but if you need to analyze more cases to accommodate a large volume of coronavirus-suspected patients, please contact us via email at contact@lunit.io and we will provide you with an assistance.

Preliminary Study

Analyzing COVID-19 Cases With AI

Our preliminary study results show that Lunit INSIGHT CXR is capable of detecting consolidation, which is a major finding that indicates pneumonia, which, in this study, uses data from coronavirus patients. The AI showed performance at a level comparable to that of radiologists.

In our study, we collected six AP (anteroposterior, taking the image from the front side of the chest) x-ray images of six patients with confirmed diagnosis of coronavirus at the time of admission. All patients had pneumonia on chest CT. On x-ray, three patients had clearly visible lesions, subtle in one patient, and not visible in the remaining two patients, when interpreted by a radiologist.

In the retrospective analysis of these images, our goal was to compare the detection performance of human radiologist and AI, to see if our AI would be capable of detecting at human expert-level or more, correctly finding lesions from a chest x-ray of a coronavirus infected patient. The result showed that  AI was able to detect what radiologists detected, demonstrating same-level performance.

Case 01

There is an ill-defined patchy increased opacity in the left mid to lower lung zone, which may be unclear for interpretation. The Lunit INSIGHT CXR has correctly detected the lesion with the abnormality score of 34.0%.

Case 02

The plain CXR taken on the patient in ICU care shows patchy ground-glass opacities in the bilateral lower lung zones. The Lunit INSIGHT CXR has correctly detected the lesions with the abnormality score of 72.0%, even in the setting with many tubes, central line, and external devices overlapped on the image.

Case 03

There are multifocal patch/nodular consolidations and ground-glass opacities around the right mid to lower lung zone. The Lunit INSIGHT CXR has correctly detected the lesions with the abnormality score of 76.0%.
link to facebooklink to twitterlink to linkedin

Address - 27 Teheran-ro 2-gil, Gangnam-gu, Seoul, South Korea © Lunit Inc.