AI Unveils Uncommon Triumph in Breast Cancer Detection

AI's potential opportunities and risks for medicine are generating increased interest. The results were reported during a time of notable attention on this topic. [Source: Emily Kask/AFP via Getty Images]

Groundbreaking research reveals that artificial intelligence (AI) has accomplished a remarkable feat, surpassing traditional radiologists in detecting breast cancers from mammograms by an impressive 20 percent. The findings, recently disclosed as part of a Swedish trial, mark the first-ever randomized controlled study that explores the integration of AI into breast cancer screening, a significant development in the ever-evolving landscape of technology regulation.

The interim results, published in the esteemed Lancet Oncology on a late Tuesday evening, demonstrate that AI-supported mammogram analysis, in conjunction with one or two radiologists, matched the performance of two radiologists unaided by AI, thereby leading to the identification of 20 percent more cancer cases.

Additionally, radiologists experienced a substantial reduction in their workload, saving 44 percent of their time that would otherwise be spent on mammogram readings.

The ongoing trial, conducted in Sweden, involved over 80,000 women, with half the participants having their mammograms reviewed by two radiologists without AI, while the other half had their mammograms meticulously analyzed by AI in tandem with a radiographer. However, in instances where AI generated the highest risk score, two radiologists assessed the screening.

These findings emerge amid a heightened interest in the potential and risks that AI presents in the field of medicine and beyond. The application of AI in medical settings has been increasingly prevalent, but concerns persist regarding the training and validation of algorithms in these domains, along with the risks of potential bias and over-diagnosis.

Meanwhile, the European Union is gearing up for stringent regulation concerning the utilization of AI, and the European Medicines Agency is diligently assessing the risks and benefits of integrating AI into drug development processes.

Nevertheless, despite the encouraging results, the lead author of the study, Kristina Lång from Lund University in Sweden, urges caution, stating that the interim safety findings alone are insufficient to confirm AI's readiness for implementation in mammography screening. Lång and her team await further results from the trial to determine whether the utilization of AI reduces the number of cancers detected between screenings and whether it is a viable solution to adopt.

Lång asserts that the true potential of AI lies in alleviating the burden on radiologists, enabling them to tackle the excessive amount of reading more effectively. By potentially eliminating the need for a second radiologist to review mammograms, AI could allow radiologists to extend their help to more patients.

Stephen Duffy, a professor of cancer screening at Queen Mary University of London, comments on the results, recognizing the significance of reducing the time burden on radiologists in breast screening programs. However, he also raises concerns about AI potentially over-detecting harmless lesions.

This pioneering study not only stands as the first randomized trial exploring the use of AI in reading mammograms but also represents one of the earliest ventures into AI's application in the realm of radiology. An editorial published in European Radiology in January emphasised the importance of randomized controlled trials to ensure the safety of AI systems, which can produce unpredictable and undetectable errors beyond the scope of human logic.

Summarised from the original article by:

Post a Comment

Previous Post Next Post