‘AI analysis can reduce unnecessary CT scans by 30%’

Radiologists can make more effective analyses of chest X-rays by using artificial intelligence, according to a joint study by the Korean AI company Lunit and the Massachusetts General Hospital (MGH) of Harvard Medical School.

The research result was published in European Radiology, Europe’s most renowned journal of radiology, last month, Lunit said on Thursday.

MGH, in affiliation with Harvard Medical School, is one of the oldest hospitals in the U.S. and a global research institution, and has maintained a research partnership with Lunit since 2020.

Last year, the two partners proved that Lunit Insight CXR, the software that supports lung disease screening, could precisely detect malignant solitary pulmonary nodules that can cause cancer from chest R-rays, publishing the result in JAMA Network Open, a world-famous academic journal.

The focus of the research was the accuracy and efficiency in analyzing chest X-rays. Eight specialists -- three radiology medical residents and five expert radiologists -- participated in the study to confirm whether Lunit Insight CXR provides efficient help in actual X-ray interpretation.

By analyzing the chest X-ray data of 519 participants in the National Lung Screening Trial (NLST), they found that analysis using AI can become efficient means for both doctors and patients.

The participating doctors recommended 28 percent more chest CT screening for patients with the risk of lung cancer while advising 30 percent fewer unnecessary CT scans for cancer-negative patients.

“The use of AI detects solitary pulmonary nodule with accuracy on X-rays and omits unnecessary CT screening for some patients,” said Mannudeep K. Kalra, a radiologist at the MGH who led the research. “Through the result, patients can save unnecessary radiation exposure and reduce medical expenses.”

Lunit CEO Beomseok Brandon Suh also said, “Chest X-rays are the primary tool in making lung cancer diagnosis. However, it has limits in portraying a three-dimensional body structure in the form of a two-dimensional image.”

Through accurate analysis by Lunit Insight CXR, medical experts can make more precise and efficient diagnoses. In addition to preventing potential dangers of cancer in its early stages, it can save time and money for patients not needing additional inspections, Suh added.

Through machine learning of more than 3.5 million pages of medical data, Lunit Insight CXR detects nine major chest diseases, including solitary pulmonary nodule, induration, and pneumothorax, with high accuracy. It has won approval from the Ministry of Food and Drug Safety and CE mark from Europe and is used in more than 300 medical institutions in about 30 countries.

Experts analyze chest X-rays using Lunit Insight CXR.