Precision medicine comes to breast cancer screening as Massachusetts General Hospital implements new AI-based tool to double the precision of mammograms. How it’s challenging traditional breast cancer risk models
Breast cancer screening is an important tool for detecting tumors in at-risk patients while their breast cancer is still in an early, treatable stage. Breast cancer screening is recommended every two years for all women ages 50 to 74 years old unless they have a major variable affecting their risk.
Current breast cancer screening guidelines are based on the traditional one-size-fits-all medical approach. Finding the right balance of screening frequency has proven tricky. Some patients with higher risk should be screened more frequently than once every two years while other patients have a lower risk and may incur unnecessary costs from screening that is performed every two years.
To better individualize breast cancer screening and more accurately predict the risk of breast cancer development, researchers from Massachusetts General Hospital (MGH) and Massachusetts Institute of Technology (MIT) teamed together to develop a new precision medicine tool. This tool was designed to identify the specific risk of each patient based on their mammogram findings.
A New Example of AI Now in Clinical Practice: Mammogram Images
Researchers at Mass General used artificial intelligence (AI)-based technology to analyze patients’ mammogram images against a database of over 200,000 other mammograms to identify patients who had higher risk of developing breast cancer. The AI component of this technology allowed for a more complex analysis than a human could perform while simultaneously allowing each mammogram to be almost instantly compared to far more images than would ever be practical using human evaluation.
Also used were other personalized data, such as age and hormonal factors, as part of the algorithm, but the researchers made these other data optional to allow the AI technology to be used in the widest variety of clinical sites. Researchers also optimized the AI tool to work with a variety of imaging machines and different environments so that it could be more widely applied.
While two years of research has gone into developing this AI tool, dubbed “Mirai”, the technology has only recently been integrated into actual clinical practice, with MGH being the first adopter of this new technology. The new AI tool is already showing promising results now in clinical practice.
Mirai AI Tool Performs Differently Than Current Clinical Standard
Researchers have found that Mirai identified almost two times as many future cancer diagnoses for high-risk patients than the current clinical standard. Mirai also provided similar accuracy regardless of a patient’s age, race, breast density, and cancer subtypes. This finding is particularly important because existing models were developed based on Caucasian populations, even though African American women are 43% more likely to die from breast cancer than Caucasian women. A clinical standard that is accurate regardless of race is likely to be more inclusive and have a wider application.
Mirai allows clinicians to predict each specific patient’s unique risk of developing breast cancer each year for the next five years, allowing clinicians to identify the screening needs of each patient based on their unique findings. The precision medicine approach allows personalized care and avoids the traditional one-size-fits-all approach that is currently used.
Challenges Traditional Breast Cancer Risk Models
“Deep learning (DL) models that aid in predicting the risk of breast and lung cancer are leading the way as artificial intelligence (AI) applications in radiology move from the theoretical to the practical,” wrote Richard Dargan in July 2021 for the Radiological Society of North America (RSNA).
A number of breast cancer risk models are used to inform decisions. Wrote the authors of a validation study of four of them, prior to February 2019, and published in Cancer Network in the journal Oncology: “Unlike cardiovascular models, there are few prospective, independent validations of models for estimation of cancer risk.” Therefore, validation studies of breast cancer risk models will continue.
Breast imagers are uniquely positioned to lead the way in AI cancer screening through their extensive experience with computer-assisted diagnosis (CAD), stated Constance Lehman, MD, PhD, a diagnostic radiologist, Professor of Radiology at Harvard Medical School, and Director of Breast Imaging at MGH, for an article published by RSNA. “We’ve learned a lot in breast imaging, from the traditional approaches through computer-aided diagnosis, so we’re trying to leverage that knowledge to ensure the AI tools we’re implementing are really going to help our patients.”
Said Adam Yala, a Computer Science and Artificial Intelligence Laboratory (CSAIL) PhD student and lead author on a paper about Mirai that was published in Science Translational Medicine: “Improved breast cancer risk models enable targeted screening strategies that achieve earlier detection, and less screening harm than existing guidelines. Our goal is to make these advances part of the standard of care. We are partnering with clinicians from Novant Health in North Carolina, Emory in Georgia, Maccabi in Israel, TecSalud in Mexico, Apollo in India, and Barretos in Brazil to further validate the model on diverse populations and study how to best clinically implement it,” Yala told MIT News.
While the Mirai AI tool is demonstrating early successes in the area of breast cancer screening at MGH, researchers are continuing to build and expand its capabilities. The next step, researchers say, is to use Mirai to not only analyze a patient’s mammogram image against a large database of images, but also against that patient’s previous mammogram images. This would provide even more data that would theoretically enhance Mirai’s performance and even better identify each patient’s breast cancer risk.
Once only relevant in the realms of oncology treatments, precision medicine is now expanding into every area of medicine, enabling imaging technologies like Mirai to achieve results that are humanly impossible. As AI technology and precision medicine treatment paradigms grow together, hospitals will continue to have more tools at their disposal to provide individualized patient care.