Breast Cancer Detection Rates Can Significantly Be Improved With AI, New Research Shows

Breast cancer can now be detected more accurately as the use of artificial intelligence (AI) has been linked to improved breast cancer detection rates among nearly half a million women. In a groundbreaking research, spearheaded by Alexander Katalinic of the University of Lübeck and a team of 200 certified radiologists, this study spanned from July 2021 to February 2023, covering 12 screening sites.

The findings revealed that with the aid of AI, radiologists could identify an additional breast cancer case per 1000 mammograms, marking a 17.6 percent increase over traditional methods without AI assistance.

The study, which involved the analysis of 461,818 women's mammograms, saw 260,739 of these exams reviewed with the help of AI. This innovative approach resulted in a cancer detection rate of 6.7 per 1000 screens, a noticeable improvement from the 5.7 per 1000 rate for screens examined without AI.

Impressively, biopsies conducted after AI-assisted screenings were more likely to confirm the presence of cancer, suggesting that AI not only enhances detection rates but may also reduce the necessity for unneeded biopsies. This is the first large-scale test of AI in a real-world screening programme, demonstrating that it can improve detection rates while not increasing false positives or unnecessary follow-up tests, as reported by The Guardian.

This approach to breast cancer screening typically involves two radiologists independently examining each mammogram. By integrating AI, which categorizes mammograms as normal, suspicious, or unclassified and provides alerts for potential cancers, the process aims to lessen the immense workload of analyzing millions of images each year. The AI's "safety net" feature ensures a reduction in missed detections, offering a path to a more efficient and accurate screening process that minimizes false positives.

The positive outcomes from this study highlight AI's potential to not only match but surpass traditional screening methods. Stefan Bunk from Vara, the AI company that contributed to the study, noted the AI's superiority in certain aspects of breast cancer screening, opening the door for discussions on how AI could be more seamlessly integrated into radiologist workflows. This could even lead to AI replacing one of the traditional human reviews in the screening process.

Despite initial hesitations around the potential for AI to lead to missed diagnoses or create disparities in treatment, the study's results have been met with optimism. The integration of AI has shown to decrease the time radiologists spend on scans that AI has already identified as normal, while simultaneously enhancing detection rates and accuracy.

Ben Glocker from Imperial College London emphasized the necessity of real-world testing and policy development to facilitate broader adoption of AI in healthcare, underscoring the safety and effectiveness of AI-assisted screenings.

This pivotal study marks a significant milestone in the use of AI within healthcare, particularly in the early detection of breast cancer. It suggests a future where screening programs might increasingly rely on AI to improve accuracy, efficiency, and possibly patient outcomes.

As AI technology progresses, its incorporation into healthcare practices holds the promise of refining diagnostic processes and alleviating the workload on medical professionals, setting a new standard for breast cancer screening and beyond.