A study investigating the use of artificial intelligence (AI) to read mammograms has found that AI can help improve breast screening outcomes. The BRAIx project demonstrated that while AI did not outperform the standard multi-human reader system currently used in Australia, it offered significant benefits when used in conjunction with human readers. This approach improved cancer detection and reduced unnecessary recalls.
Detailed simulations of five AI-integrated pathways highlighted AI as an effective tool for supporting human readers, improving both accuracy and efficiency. In the current program, every mammogram is read independently by two breast radiologists, with a third consulted in cases of disagreement. The study found that using AI as the second reader provided a beneficial approach.
Associate Professor Helen Frazer, the Project Lead and Clinical Director of St Vincent’s BreastScreen, emphasized that AI is not ready for standalone performance but is effective when used alongside human readers. The BRAIx study, recently published in Nature Communications, utilized Australian AI capability developed by a team of clinicians, data scientists, and engineers from various institutions.
The algorithm was trained on a dataset of 3.3 million breast screen images collected from women who attended the BreastScreen Victoria program between 2016-2019. This local development ensures a high degree of transparency and trust for Australian use.
Key findings from the study include a reduction of up to 10.9% in unnecessary recalls for assessment and a nearly 50% reduction in the workload for breast radiologists when supported by AI. This is particularly significant given the shortage of breast radiologists to meet increasing screening demands.
Additionally, the AI reader showed potential in predicting future breast cancer risk, suggesting a move towards a more personalized screening approach. A randomized controlled trial is planned to start later this year, aiming to ensure the safe and beneficial use of AI in breast cancer screening. This trial is expected to involve an estimated 200,000 Australian women.
Breast cancer remains the leading cause of cancer death in women worldwide, with close to one million women screened annually through the national BreastScreen Australia program. The development of a state-of-the-art AI reader in Australia not only showcases the skills and capabilities of local researchers and institutions but also underscores the importance of investing in AI applications in healthcare.
