Will AI Identify Breast Cancer Better Than Radiologists in Actual Clinical Screening?
- A research team at the Karolinska Institutet in Sweden will assess the performance of AI startup Lunit’s breast cancer AI software as an independent mammography reader in screening
- The prospective study marks first of its kind, aiming to screen 55,000 women with mammography at Capio Sankt Gorans Hospital in Stockholm, Sweden
Lunit, a leading medical AI startup, today announced that it will participate in the new study conducted by the research team of Karolinska Institutet in Sweden. According to Lunit, this research is the first prospective study in Europe to validate the performance of AI in an actual clinical trial of screening mammography.
Dr. Fredrik Strand and his research team at Karolinska Institutet will assess the performance of Lunit INSIGHT MMG combined with radiologists in a true screening population, compared to the “double-reading” standard-of-care in Europe, in which two radiologists assess screening mammograms. Lunit INSIGHT MMG is an AI software developed by Lunit to detect breast cancer in mammography images. It is CE marked and currently commercially available in Europe.
“There have been multiple retrospective studies to validate AI performance, but there has not been any prospective studies to assess AI’s clinical utility in an actual clinical setting,” said Brandon Suh, CEO of Lunit. “Documentational processes such as licensing and regulatory approvals for an AI solution are mandatory, but assessing the actual effectiveness of its performance in clinical trials is of vital importance. We look forward to finding a meaningful result to bring deeper understanding of our AI solution.”
The target volume of the study includes 55,000 patients, in collaboration with Capio Sankt Görans Hospital of Sweden. It is one of the largest screening centers in Sweden with a high cancer detection rate. The expected timeline for the study is two years. Anders Byström, head of Radiology at the hospital, is intrigued: “We are truly looking forward to understanding whether AI can make our screening process even more efficient”
Fredrik Strand, principal investigator at the Department of Oncology-Pathology at Karolinska Institutet and a radiologist at Karolinska University Hospital explained “We were impressed by the results of Lunit INSIGHT MMG in our retrospective trials. It will be interesting to see how it plays out in a clinical setting.”
Last year, Dr. Strand’s research team conducted the first independent comparison of three commercialized AI to identify breast cancer. Among them, Lunit’s algorithm showed the best accuracy with around 15 percentage points higher sensitivity compared to the other two algorithms, and also surpassed the sensitivity of one of two radiologists. The study was published in JAMA Oncology.
In April, Dr. Strand will hold an online webinar featuring the findings on the previous studies with Lunit INSIGHT MMG and introduction of the newly initiated prospective study. Details and pre-registration can be foundhere.