PARIS: A large Swedish study has indicated that artificial intelligence (AI) could potentially reduce the workload of radiologists by almost half when analyzing routine scans for signs of breast cancer. The interim results of the trial show promising potential, but the authors caution that further research is necessary before AI can be widely implemented for breast cancer screening.
AI has already demonstrated proficiency in reading medical scans, and with a shortage of radiologists in many countries, there is hope that AI can expedite the time-consuming task of analyzing routine scans, making it more accurate and efficient.
Breast cancer screening is crucial for early detection, as more than 2.3 million women were diagnosed with the disease in 2020 alone, resulting in 685,000 deaths, according to the World Health Organization. In Europe, women aged 50 to 69 are advised to undergo a mammogram every two years, with two radiologists analyzing the results.
The Swedish study involved 80,000 women who underwent mammograms at four sites in southwest Sweden between April 2021 and July last year. The scans were randomly divided for analysis, with one group using an AI-supported system, while the control group consisted of two human radiologists.
AI Algorithm Accurately Predicted Breast Cancer Risk
The AI algorithm accurately predicted the risk of cancer out of 10 and spotted 20% more cancers compared to the human radiologists, resulting in an additional case detected for every thousand women screened. Both the AI-supported system and the human radiologists showed the same rate of false positives at 1.5%.
Furthermore, the AI-supported system reduced the workload for radiologists by 44%, as only one person was required to read the scans instead of the usual two.
Kristina Lang, a radiologist at Sweden’s Lund University and the lead author of the study, emphasized that the AI’s potential lies in reducing the burden on radiologists and streamlining the excessive amount of reading.
However, Lang stated that the “promising interim safety results” were insufficient to confirm that AI is ready for widespread use in mammography screening. The researchers caution that it will take two more years before the trial can determine whether AI leads to a reduction in interval cancers, which are detected between routine screenings.
Stephen Duffy, a professor of cancer screening at Queen Mary University of London, praised the study’s high quality while highlighting the AI algorithm’s possible over-diagnosis of certain early breast cancer forms. Nonetheless, he emphasized that reducing the burden on radiologists’ time is of considerable importance in breast screening programs.
