AI Boosts Breast Cancer Detection in Mammography Screening: Study
A recent study conducted by researchers at Karolinska Institutet has revealed that artificial intelligence (AI) can significantly improve the detection of breast cancer in screening mammography. The study, known as the ScreenTrustCAD study, found that one radiologist supported by AI detected more cases of breast cancer than two radiologists working together.
For more than three decades, screening mammography has played a crucial role in reducing breast cancer mortality rates. However, challenges such as the lack of radiologists and the fact that not all cancers are detected have hindered its effectiveness. Retrospective studies have shown that AI could potentially address these issues.
According to Karin Dembrower, the lead author of the study, AI and humans perceive images slightly differently, which creates a synergy that improves our chances of detecting cancer.
Traditionally, two radiologists analyze every screening mammogram. However, in this study, the exams were assessed by both two radiologists and AI to determine which women should be recalled for further investigation. The researchers compared the accuracy of different combinations of AI and radiologists with the traditional two-radiologist approach based on the ultimate breast cancer diagnosis.
The study was conducted at Capio St Göran’s Hospital in Stockholm and involved screening over 55,500 women between the ages of 40 and 74. The results showed that the addition of AI to the two-radiologist approach detected the most cases of cancer, with a total of 269. However, it was also found that one radiologist supported by AI detected 261 cases, which was statistically non-inferior to the two-radiologist approach. AI alone detected 246 cases of breast cancer.
Dr. Fredrik Strand, the principal investigator of the study, stated, Compared with the current two-radiologist standard, assessment by one radiologist and AI resulted in a four percent increase in breast cancer detection and halved the radiologists’ image reading time.
In addition to improving cancer detection rates, the study also revealed a reduction in false positives, which refer to the recall rate for healthy women. Compared to the two-radiologist approach, one radiologist supported by AI and AI alone led to a six percent and 55 percent reduction in false positives, respectively. This reduction in false positives can help minimize unnecessary suffering and costs for patients.
Dr. Strand emphasized that AI should be viewed as a complement to radiologists rather than a substitute. He stated, Even if AI takes over much of the initial examination, a radiologist is needed to make the judgment before any patient is recalled for further investigation, and, if necessary, to take biopsies from suspicious breast areas.
The findings of the study suggest that AI is ready to be implemented in breast cancer screening. Capio St Göran’s Hospital has already started using an AI-supported radiologist since June 2023, which has freed up time for radiologists to focus on breast cancer patients. However, it is crucial to choose an AI system that has been thoroughly tested on images from the same type of mammography equipment and to ensure continuous monitoring after clinical implementation.
In conclusion, AI has shown significant potential in improving breast cancer detection rates and reducing false positives in mammography screening. By combining the unique perceptual abilities of humans and AI, the chances of detecting cancer can be greatly enhanced. While AI has the potential to take over most screening mammography assessments in the future, it should be seen as a valuable tool that complements the expertise of radiologists. With further research and implementation, AI can revolutionize breast cancer screening and save more lives.
