Lunit Demonstrates Medical AI Effectiveness Through RWD

The AI imaging analysis for mammography by Lunit, 'Lunit Insight MMG,' being used in actual medical institutions. (Photo by Lunit)

[Asia Economy Reporter Chunhee Lee] Medical artificial intelligence (AI) company Lunit has released real-world data (RWD) of medical AI for the first time in history.

Lunit announced on the 28th that from the 27th to December 1st (local time), at the '2022 North American Radiology Society Annual Meeting (RSNA 2022)' held in Chicago, USA, it disclosed RWD demonstrating for the first time worldwide and in the history of medical AI research that AI can be used safely and meaningfully in actual medical settings. This study was also selected as an 'Oral Presentation,' one of the major research achievements of RSNA 2022 this year.

Now in its 108th year, RSNA is the largest radiology conference in North America and globally, attended by more than 50,000 experts from academia, medical fields, and biotechnology companies worldwide. Lunit submitted a total of 12 research abstracts, of which 8 were presented orally and 4 as posters.

RWD refers to 'real-use clinical data' obtained in medical practice. Unlike clinical data collected under controlled environments with limited conditions to achieve evaluation metrics, RWD is data acquired in actual medical settings with various variables, thus earning higher trust from medical professionals.

In particular, while most existing AI research related to breast cancer diagnosis has been retrospective studies using data from completed treatments, this study is the first prospective study in medical AI research history that is statistically and clinically meaningful through large-scale clinical RWD.

This study was led by Dr. Fredrik Strand of the Karolinska Institute in Sweden, a globally recognized authority in breast cancer research. The research team analyzed the feasibility of AI implementation in real medical environments by using the breast imaging AI analysis solution 'Lunit Insight MMG' on 55,579 Swedish women who underwent breast cancer screening over a period of more than one year from April last year to June this year. Since most European countries have legislated 'double reading' by two radiology specialists during breast cancer screening, the study divided the diagnosis into cases read by two specialists, cases read by Lunit AI combined with one specialist, and cases read by Lunit AI alone, comparing and analyzing each result.

As a result, in terms of cancer detection rate (CDR), the combination of Lunit AI and one specialist detected more cancers than the two specialists reading together. Also, even when Lunit AI read alone, the cancer detection rate was not inferior to that of two specialists. The CDR per 1,000 examinees was 4.3 for AI combined with one specialist, 4.1 for two specialists, and 4.1 for AI alone.

Additionally, in terms of recall rate (RR) for cancer re-examination, the combination of Lunit AI and one specialist showed a lower recall rate than two specialists reading together. Lunit AI alone had a significantly lower recall rate compared to two specialists. The RR per 1,000 examinees was 28.0 for AI combined with one specialist, 29.3 for two specialists, and 15.5 for AI alone. Since unnecessary recalls impose burdens on both patients and medical staff due to additional tests, a lower recall rate while maintaining the usual cancer detection rate is considered a desirable outcome.

Dr. Strand said, "In many countries such as Europe and Australia, two doctors make the final judgment in breast cancer diagnosis, but there is great difficulty due to a shortage of radiology specialists." He added, "This study shows that AI can replace the role of one doctor, reducing total medical costs and laying the foundation for AI to be widely used in breast cancer screening and insurance reimbursement."

Seobeom Seok, CEO of Lunit, also said, "This study demonstrates that AI will become the standard for breast cancer screening," and "The key point is that the effectiveness of AI was proven for the first time through large-scale prospective clinical trials in actual medical settings."

This content was produced with the assistance of AI translation services.