This article was originally written for Forbes. View the article here.
Ainsley MacLean, MD, is the Former Chief Medical Information Officer (CMIO) at the Mid-Atlantic Permanente Medical Group, Kaiser Permanente.
Current breast cancer diagnosis practices, such as mammograms and ultrasounds, are usually effective tools. However, they don’t always yield the right results. For instance, according to the NIH’s National Cancer Institute, “screening mammograms miss about 20% of breast cancers that are present at the time of screening.” A lack of early detection can prove fatal.
But today, new technology stands to help women who might have breast cancer that may otherwise go undetected. Doctors can increase accuracy in breast cancer screening and diagnosis by using artificial intelligence to find patterns in vast amounts of data that radiologists and pathologists assisting cases may not notice.
The Data On AI Usage In Breast Cancer Screening And Diagnosis
Research has revealed the crucial role AI can play in breast cancer detection and diagnosis.
A study published in 2020 in Nature, for instance, found that an AI system “was capable of surpassing human experts in breast cancer prediction.” Specifically, researchers found that the AI system decreased false positives in the United States by 5.7% and false positives in the United Kingdom by 1.2%. Additionally, a study published in The Lancet Digital Health found that “combining the strengths of radiologists and AI could result in marked improvements in the sensitivity and specificity of individual radiologists ahead of the consensus conference.”
Granted, more research needs to be conducted. My own organization is preparing to launch a study similar to the one published in Nature so we can evaluate first-hand how artificial intelligence can be added to our radiologists’ workflows and potentially improve breast cancer screening and diagnosis for our patients. However, the results of published studies are promising—they indicate that using AI can significantly improve healthcare outcomes for women, helping them live longer lives.
The Benefits Of Leveraging AI In Breast Cancer Screening
There are several key benefits of leveraging AI in breast cancer screening.
First, even the most skilled radiologists will not detect every single instance of breast cancer within their patient base. Some cancers are not detectable by the human eye. Even if two radiologists read every mammogram (as is standard practice in the U.K.), there is still a likelihood of false negatives. Artificial intelligence can function as an assistant for radiologists, helping them with their decision-making and ultimately increasing breast cancer detection rates. Essentially, a radiologist will review the images at hand, whether from a mammogram, MRI or ultrasound. They’ll go through their standard process and sign off on a report. Then an AI tool can scan those images and either come to the same conclusion as the radiologist—or a different one—triggering further review by the physician. This particular process is how we have set up our study (our radiologists will be reviewing mammograms), and as a safeguarding measure, we will be sending the results to what we call an arbitrator, someone who will take a look at the AI tool’s analysis and evaluate whether or not it actually picked up on cancer that a radiologist didn’t see.
In addition to further standardizing the breast cancer detection process, AI can streamline radiologists’ workflows. The average radiologist screens dozens of images per day. AI tools can help them work more efficiently and effectively as they examine image after image. In Europe, there’s an especially pressing radiologist shortage, which arguably makes the need for AI tools more pressing. In the U.S. healthcare system, we do not have that drastic of a shortage, but there are parts of our country that do.
As for the pathology side of breast cancer screening and diagnosis, AI can be used as an assistant for pathologists during whole slide imaging. Specifically, AI can highlight areas on slides that look abnormal, areas that are indicative of breast cancer.
AI Should Not Be Approached As A Be-All, End-All Solution In Breast Cancer Diagnosis And Screening
While there are benefits of leveraging AI in the breast cancer screening and diagnosis process, there are drawbacks as well. For one, AI should not replace the work of radiologists and other physicians. AI has come a long way, but it’s far from being able to act autonomously. And in my view, when it comes to medicine, even if AI did reach a point where it could one day act autonomously, we should not put aside the expertise of medical practitioners. Rather than being approached as a be-all, end-all solution for breast cancer diagnosis and screening, AI should act like a co-pilot for radiologists and pathologists.
Another drawback is that AI tools can hallucinate and produce results that are biased and inaccurate. When it comes to AI algorithms, particularly with cancer, it’s important for physicians and other leaders to carefully examine and account for certain variables in data sets, such as racial groups and age so they can minimize the chances of biased and inaccurate results.
Then there’s the privacy and security element. With AI tools, some patient data may need to enter additional servers. Physicians and leaders of medical institutions need to make sure that they are operating within the bounds of all regulations and that patients give consent for their medical data to be used in this manner. They also need to make sure they have the proper cybersecurity protocols and guidelines in place to minimize the chances of sensitive patient information getting into the wrong hands.
Finally, AI systems can be immensely expensive to implement, especially in smaller medical facilities. Due to its costs, not all physicians will be able to leverage AI in breast cancer screening and diagnosis.
However, the medical institutions that have the resources to start implementing AI systems now for breast cancer screening and diagnosis should consider doing so. Artificial intelligence, I believe, is driving us to higher quality, more personalized diagnosis processes not just for breast cancer, but for other cancers as well. By taking steps now to implement AI, physicians and leaders of medical institutions can pave the path for more patients to survive cancer—and go on to live fulfilling lives.
