A recent large-scale and rigorous study conducted in Sweden has shed light on the potential of artificial intelligence (AI) to improve breast cancer AI detection in screenings. The study marks a significant milestone as it is the first trial to test AI in real-time on actual patients, paving the way for further investigations into the technology’s true value and potential risks. While the findings are incredibly promising, additional research is necessary, especially in the United States, to guide doctors in integrating AI into healthcare systems effectively after AI detection.
The primary focus of the initial stage of the study, reported in the Lancet Oncology, was to determine the safety and practicality of incorporating AI into breast cancer screening practices. The results were unequivocally positive, with the AI system aiding human radiologists in identifying approximately 20% more cancers compared to traditional double readings performed by two independent radiologists. Impressively, the AI system achieved this without significantly increasing the rate of false positives.
The study demonstrated that AI could substantially reduce the workload of radiologists, estimating a remarkable 44% reduction in screen reading time. This reduction in burden comes as a significant improvement, particularly in healthcare systems facing strained medical workforces.
The quest to prove AI’s impact on cancer care
While AI’s efficiency in detecting breast cancer has been established, the next critical step is to ascertain whether this technology genuinely improves overall health outcomes for women. Researchers in Sweden plan to continue monitoring the women involved in the trial to answer this crucial question. Apart from validating AI’s performance in cancer detection, they will assess whether the additional cancers detected by the AI system are clinically meaningful – i.e., lesions that would eventually pose harm to women. Furthermore, the study will investigate whether AI can reduce the occurrence of “interval cancers,” which are typically more aggressive and deadlier.
The need for thorough evaluation of AI’s impact is evident from the past experience with computer-aided detection using rudimentary AI versions, which led to increased false positives and unnecessary biopsies for non-dangerous precancerous cells. Thus, it is essential to demonstrate AI’s true value in order to avoid any undue strain on the healthcare system.
Adaptability to regional healthcare systems
Efficiencies achieved through the study are likely to benefit European and Australian healthcare systems where mammograms are commonly reviewed by two radiologists. In contrast, the U.S. standard of care typically involves a single radiologist reviewing 3-D scans, which differs from the 2-D scans used in the Swedish study. Despite the regional disparities, there are valuable lessons to be learned from the study’s algorithm, which effectively stratified women by cancer risk. This suggests the potential for AI to triage exams and prioritize high-risk cases, expediting treatment and improving patient outcomes.
AI detection: A complement, not a replacement
The vision of AI completely replacing doctors in cancer diagnosis has been tempered over time. While some early predictions speculated that AI might replace radiologists in five to ten years, leading AI expert Geoffrey Hinton has since warned of potential dangers and urged cautious implementation. The results from the Swedish study emphasize that AI should be seen as an addition to, rather than a replacement for, the expertise of doctors.
The groundbreaking study in Sweden offers encouraging evidence of AI’s potential to revolutionize breast cancer detection. While the technology has shown remarkable efficiency and safety, it remains critical to evaluate its impact on overall health outcomes before widespread adoption. As AI becomes increasingly integrated into healthcare systems, more comprehensive studies are required, especially in the United States, to guide doctors and ensure AI’s benefits are harnessed most effectively, ultimately leading to improved breast cancer care and patient outcomes.
