Original article was published on Artificial Intelligence on Medium
Using AI to better manage COVID-19
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.
Back on March 31st, we released a special version of Lunit INSIGHT CXR, online and free of charge, designed with enhanced sensitivity for pneumonia. (You can try the solution here: https://www.lunit.io/en/covid19/)
We had hoped that this solution would help healthcare providers and patients who battle against COVID-19, and we saw an increase in traffic up to four times upon release, which signaled the urgent need and interest of a helping hand, even from AI.
Together with our partners, we took a step further to actively provide the AI in the actual COVID-19 environment within hospitals. Now after two months since the initial release of the special version of Lunit INSIGHT CXR, we now have deployed our software in more than 10 locations throughout different countries, specifically to manage COVID-19.
Here below, we are sharing what we have learned from the past few months, working with different hospitals in the frontline of the battle. Among the many hospitals we have been in touch with, some of them were experiencing frustrations from limited PCR tests and/or a lack of manpower and resources. In hopes of sharing the burden, we invite you to read how our AI solution, Lunit INSIGHT CXR, CE marked, can make a contribution to the healthcare system.
Stay healthy, stay safe, and for those of you interested in using our solution, please reach out at firstname.lastname@example.org. Thank you.
Chief Business Officer and Co-founder, Lunit
How it Works
- Analyzes consolidation (including ground-glass opacity, GGO) within seconds
- Localizes consolidation via heatmap
- Provides the abnormality score, reflecting the probability of the existence of consolidation
- AI results help the user make clinical actions according to the interpretation.
AI can support during case overload, including COVID-19, which can likely lead to low reading quality due to pressure and strain in the healthcare system.
- Assist in Triaging Patients
- Target: Symptomatic patients visiting screening centers, Emergency Room, and etc.
- Method: AI-aided chest x-ray interpretation during patient overflow can help earlier detection of coronavirus-suspicious pneumonia cases, subsequently leading to prompt isolation and treatment while awaiting RT-PCR results.
2. Assist in Monitoring Patients
- Target: Patients with COVID-19 who are admitted to hospitals, community isolation facilities, and etc.
- Method: AI-aided chest x-ray exams through mobile x-ray devices, conducted on a regular basis, can help monitor the progression and/or regression of pneumonia on lung images.
Global Use Cases of Lunit INSIGHT CXR for COVID-19
- Sao Paulo, Brazil — PreventSenior
- Jakarta, Indonesia — Dr. Cipto Mangunkusumo National Public Hospital (RSCM)
- West Java, Indonesia — Rumah Sakit Universitas Indonesia (RSUI)
- Daegu/Seoul, South Korea — Seoul National University Hospital
- Vimercate, Italy — Vimercate Hospital
- Meylan, France — Vizyon
In Brazil, Lunit AI has been installed at more than 11 hospitals including 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. In May, the institution has deployed Lunit INSIGHT for the analysis of more than 20,000 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.”
In Indonesia, Lunit’s AI is currently in demonstration at Dr. Cipto Mangunkusumo National Public Hospital (RSCM) and Rumah Sakit Universitas Indonesia (RSUI), located in Jakarta and West Java, respectively. Both hospitals have COVID-19 patients hospitalized, with primary screening done with clinical exam, and chest x-ray used as a supporting tool for decision-making during triage of suspected patients. PCR tests are conducted in a limited capacity, only for patients with definite symptoms and positive imaging signs prioritized due to lack of resources. From early stages during admission to in-patient, a routine chest x-ray is conducted in one to three days of frequency, depending on the symptoms. The hospitals use a mobile x-ray in the isolation room and dedicated wards. AI analysis results are accessible directly from the PACS worklist and viewer, integrated into the radiologist’s daily diagnostic workflow. The radiologists can easily refer to the AI results when reading studies. For example, they can adjust the reading order according to the complexity and urgency of studies determined based on the AI findings and scores. A radiologist from the hospital says that Lunit’s AI is expected to provide assistance mostly on triage, as the AI is able to detect subtle changes on chest x-ray images. “We still rely on clinical findings and PCR results for treatment decisions. But we are expecting the AI to be useful in triage and also comparing the consolidation findings during follow-up x-rays, which can be used in radiology reports.”
In South Korea, Lunit INSIGHT CXR has been deployed to assist the diagnosis of COVID-19 patients from Daegu, the hardest-hit city where 62% of the entire domestic cases have been diagnosed. Seoul National University Hospital has built a healthcare center in Daegu, where physicians remotely examine patients via video call. The patients’ chest x-ray images from Daegu are sent directly to radiologists in Seoul, where Lunit INSIGHT CXR is used for the instant detection and diagnosis of COVID-19. The system enables prompt and early discovery of severe patients who are then sent directly to larger hospitals for proper treatment.
In Italy, Lunit AI is being used in Vimercate hospital, through a partnership with Fujifilm. The hospital is based in Lombardia, Italy, and takes around 160,000 radiology exams in a year with 17 radiologists and 19 technicians. The COVID-19 crisis has brought a need to improve workflow by easing the burden of reading a mass amount of exams. AI findings are integrated into the Fujifilm Synapse platform which the hospital uses for reading radiologic images. According to the hospital, Lunit INSIGHT is used for every chest x-ray analysis, including ER patients with respiratory symptoms, patients with COVID-19, before dismissing a patient with disappeared symptoms, and before hospitalizing a patient for a surgical operation to evaluate the feasibility of anesthesia. Users can quickly identify severe cases; the doctors at Vimercate says that prioritizing the worklist with AI score and results helps radiologists report immediately for patients with findings. The hospital has a protocol for COVID-19 that all patients are screened with chest x-ray. Screening frequency is defined by the clinical course of each patient.The hospital has been using the AI-integrated system on a daily basis since the end of March. Currently, Lunit INSIGHT analyzes approximately 45 chest x-rays images per day in Vimercate. Lunit is also in discussion with another health facility in Milan, for installation of the AI software.
In France, Lunit is in partnership with Vizyon, a teleradiology firm that specializes in providing a platform for AI software by connecting the technology to hospitals. Vizyon is using Lunit AI in analyzing the images they receive from the clients. Lunit is also providing a demo to one of the hospitals in France for assistance in on-site triage.Dr. Pierre Durand from Vizyon says that during patient triage, Lunit INSIGHT CXR for COVID-19 can accelerate the decision-making of patients with symptoms while awaiting PCR results. Upon showing a symptom, the patient is tested with both PCR and AI-assisted chest x-ray, with the latter providing an analysis result in less than 5 minutes. The doctor, upon reviewing the analysis made by AI, can promptly decide whether to send the patient back home or to the Emergency Hospital, depending on the severity. PCR results generally take six to 24 hours, risking a longer window of contamination while awaiting results, a longer gap of treatment, and a possible hospital overload.