Dec 24, 2020 — 4 min read
A study conducted by Seoul National University Hospital, one of the most renowned medical institutes in Asia, has revealed that an AI can help radiologists improve their performance and reduce reporting time for critical or urgent cases. The study has recently been published in European Respiratory Journal, one of the leading peer-reviewed journals for respiratory diseases.
According to the study, Lunit INSIGHT CXR, an AI software for chest x-ray, showed significant value in a simulated reading experiment for emergency patients. When assisted by Lunit’s AI, the diagnostic accuracy more than doubled from 29.2% to 70.8% for critical-emergency diseases such as pneumothorax and pneumoperitoneum.
In addition, by analyzing abnormal findings in advance and triaging the images for urgent diagnosis, the waiting time of emergency patients has been drastically reduced by 80%. The overall reading time has also shortened, leading to reading efficiency and dramatic improvement of patient treatment in emergency situations.
“This study actually shows that reading efficiency can be maximized when the radiologists and AI system are organically integrated,” said Professor Chang-min Park, the corresponding author of the study. “It especially shows high value in emergency cases where timely diagnosis and care is required for patients."
Brandon Suh, CEO of Lunit said, “It is meaningful that our AI is recognized and validated for more sophisticated and quick diagnosis of chest diseases. Continuous development of our algorithm shall be implemented to reduce the heavy workloads of radiologists and emergency personnels.”
The study further emphasized that Lunit INSIGHT CXR can detect 10 diseases (lung cancer, pulmonary nodules, atelectasis, calcification, cardiomegaly, consolidation, fibrosis, pleural effusion, pneumoperitoneum, and pneumothorax) on chest x-ray images. In fact, most lung and chest diseases can be diagnosed through this AI solution.
Using two external validation datasets of same-day CT-confirmed and open-source datasets, Lunit INSIGHT CXR showed diagnostic ability comparable to that of a radiologist. It resulted in high AUROCs for all 10 common abnormalities, ranging from 0.9-1.00. AUROC value of 1.00 implies the highest performance.
Professor Joo-gang Nam, the first author of the study stated “Lunit INSIGHT CXR individually marks the location and probability of abnormal findings for each of the 10 lesions. This is a technology that was not implemented in the previous generation of artificial intelligence, which opens the way for automatic detection and diagnosis.”
Lunit INSIGHT CXR clinically analyzed more than 6.5 million images in more than 80 countries. It is CE marked and clinically available in Europe, Middle East, Latin America, South East Asia, Australia, and New Zealand. It is expecting its FDA clearance within early 2021.