AI Algorithm Matches Human Accuracy in Breast Cancer Detection: Study
A recent study conducted in the UK has found that an artificial intelligence (AI) algorithm can match the accuracy of human readers in detecting breast cancer. The findings, published in the journal Radiology, highlight the potential for AI to improve the sensitivity and specificity of breast cancer screening.
Mammographic screening, the current standard for breast cancer detection, is not foolproof. False-positive interpretations can lead to unnecessary imaging and biopsies for women without cancer. To address this issue, the study suggests having two readers interpret every mammogram, as it increases the cancer detection rate and keeps recall rates low. However, this approach is labor-intensive and challenging to implement during reader shortages.
In light of these challenges, researchers are exploring the use of AI algorithms in breast cancer screening. The study compared the performance of a commercial AI algorithm with human readers using test sets from the UK’s National Health Service Breast Screening Program. The test sets included 60 challenging exams with abnormal, benign, and normal findings.
The results of the study showed that there was no significant difference in performance between the AI algorithm and human readers in detecting breast cancer. Both had similar sensitivity and specificity rates, with the AI algorithm achieving 91% sensitivity and 77% specificity. These findings suggest that AI has the potential to perform at the same level as human readers in breast cancer screening.
However, the researchers emphasized the need for more research before incorporating AI as a second reader in clinical practice. Large prospective clinical trials are currently underway to further evaluate the performance of AI in real-world settings. Additionally, ongoing performance monitoring of AI will be crucial to ensure its long-term success and to address any potential drift in performance over time.
The study’s lead researcher, Prof. Yan Chen, stated that while it is too early to determine the exact role of AI in breast screening, the study’s results provide strong evidence that AI can perform as well as human readers. The findings also highlight the importance of continuous monitoring of AI performance and the need for imaging centers to have processes in place to ensure its effectiveness in clinical practice.
Overall, this study paves the way for further exploration of AI in breast cancer screening. While more research is needed, the potential benefits of using AI to enhance the accuracy and efficiency of mammographic screening are promising. As technology continues to advance, AI algorithms may play an increasingly important role in detecting breast cancer and improving women’s health.
