Kendra Batchelder is the founder and CEO of Madera, a company focused on predicting cancer risk using artificial intelligence (AI). She holds a PhD in computational medicine focused on signal processing. She recently shared her insights on the way AI is transforming mammography with Susan G. Komen.
Komen: How are you seeing AI applied to patient care to improve outcomes?
Batchelder: Artificial intelligence is becoming an increasingly important tool in health care, helping clinicians make more informed and personalized decisions. In breast imaging, AI has been used for decades to support radiologists in detecting abnormalities on mammograms. The first FDA-approved computer-aided detection technology for mammography was cleared in 1998. Since then, these technologies have evolved from simply flagging areas that may need a closer look to helping radiologists assess whether an abnormality is more likely to be cancerous. While AI is not perfect and can sometimes be wrong, advances in technology are making these tools increasingly accurate and useful as an additional source of information for clinicians—not a replacement for their expertise.
Komen: How are you seeing AI applied to patient care to improve outcomes?
Batchelder: Artificial intelligence is becoming an increasingly important tool in health care, helping clinicians make more informed and personalized decisions. In breast imaging, AI has been used for decades to support radiologists in detecting abnormalities on mammograms. The first FDA-approved computer-aided detection technology for mammography was cleared in 1998. Since then, these technologies have evolved from simply flagging areas that may need a closer look to helping radiologists assess whether an abnormality is more likely to be cancerous. While AI is not perfect and can sometimes be wrong, advances in technology are making these tools increasingly accurate and useful as an additional source of information for clinicians—not a replacement for their expertise.
Komen: For those of us who don’t fully understand AI, what is it and how does it work?
Batchelder: At its simplest, artificial intelligence is technology that learns patterns from large amounts of data and uses those patterns to make predictions or support decisions. For example, an AI system can be trained using thousands or millions of medical images along with information about what happened to those patients over time. The system learns patterns that may be associated with future health outcomes and can then analyze information from new patients to provide insights for clinicians.
The quality and diversity of the data used to train AI are extremely important. If a model is trained primarily on people from a particular age group, racial or ethnic background, imaging system, or breast composition, it may not perform as well for people who are different from those represented in the training data. That is why researchers need to evaluate AI across diverse populations and be transparent about how models are developed, validated, and perform across different groups.
Komen: In your work, you chose to apply AI to breast cancer first. What drew you to breast cancer?
Batchelder: Breast cancer became a natural area of focus for me because early detection and risk assessment can make such a meaningful difference in outcomes. While mammography has saved countless lives, we know that breast cancer risk varies from person to person, and cancers do not all develop in the same way. I was drawn to the opportunity to use AI to better understand information already contained in a patient’s imaging history and clinical data, with the goal of identifying women who may be at increased risk earlier and helping make screening more personalized.
My interest in research began during college, when I interned at NASA. After earning my degree in mathematics, I became increasingly interested in applying quantitative methods to health care. Someone close to me had been affected by breast cancer, which naturally drew me toward the field. During my studies, I became particularly interested in complex systems and in applying mathematical methods developed in one field to solve problems in another. I began using a mathematical measure originally developed in astrophysics to study how structures change over time in mammograms. The underlying mathematics can remain the same—the difference is how we apply it to a new problem.
As colleagues and family members were affected by different cancers during my graduate studies, I became increasingly interested in translational research: taking scientific discoveries and turning them into tools that can ultimately make a difference for patients.
Komen: How is AI being applied to breast cancer right now? Can you share some examples that illustrate its uses?
Batchelder: Today, AI is being used in several areas of breast cancer care. Some AI tools help radiologists detect suspicious findings on mammograms, helping identify cancers that may be difficult to see. Other tools help prioritize cases, improve workflow efficiency, and support image interpretation. A newer area of research is using AI not just to look for cancer that is already present, but to estimate future breast cancer risk. This could help identify women who may benefit from additional screening or closer monitoring before cancer develops.
Komen: Mammography started as 2-D imaging, but 3-D imaging is increasingly common, and now we’re exploring the possibilities of AI-assisted mammography. What does AI-assisted mammography mean?
Batchelder: AI-assisted mammography means using artificial intelligence alongside mammography to provide additional insights to health care providers. Instead of replacing the mammogram, AI analyzes the images and other available information to help radiologists better understand a patient’s risk or identify areas that may require additional attention. For patients, the experience may look very similar to what they already know: they receive their mammogram as usual. Behind the scenes, AI can help provide clinicians with additional information that supports a more personalized screening plan.
Komen: What excites you about incorporating AI into mammography?
Batchelder: What excites me most is the possibility of making breast cancer screening more personalized. Today, many screening decisions are based largely on age and broad risk factors, but every patient has a unique history. AI could help bring together information from mammograms, clinical history, and other risk factors to provide a more complete picture of an individual’s risk. The goal is to help identify patients who may benefit from additional attention while ensuring that everyone receives appropriate care.
Most breast cancers are currently diagnosed when they are invasive. If we can improve our ability to predict future breast cancer risk and make risk assessment more consistent and equitable, we may have an opportunity to identify cancers earlier, including at precancerous (DCIS) or early stages. Because treatment is generally more effective when cancer is found early, improving our ability to understand risk could ultimately make a meaningful difference for patients.
Komen: Will we reach a point where machines are exclusively reading mammograms? What challenges do you see in the path to get there?
Batchelder: I believe the future of AI in breast imaging is one where AI and radiologists work together. Mammography involves much more than analyzing an image—it requires clinical judgment, communication with patients and care teams, and an understanding of each patient’s broader medical history. There is also a deeply human and emotional side to breast cancer screening, particularly for women at higher risk or those who have experienced a previous false-positive result.
AI may help radiologists work more efficiently and provide additional information, potentially giving them more time to focus on patients. But I believe there will always be an important role for physicians, particularly when cases are complex, involve multiple types of imaging or clinical information, or require procedures such as biopsies. AI should support clinical expertise, not replace the human connection at the heart of breast cancer care.
Komen: From your perspective, is AI detecting breast cancers earlier or capable of detecting breast cancers earlier?
Batchelder: AI has shown promise in helping radiologists detect breast cancers that may be subtle or difficult to see. But I believe one of the most exciting opportunities is to use AI not only to detect cancer earlier, but to identify a woman’s risk of developing breast cancer earlier. By analyzing patterns in mammograms and, increasingly, combining imaging with clinical information, AI may help identify women who are at higher risk before cancer develops. This could give patients and their care teams an opportunity to personalize screening and, when appropriate, consider earlier or additional screening.
Komen: How might AI drive changes in screening guidelines or screening frequency?
Batchelder: As we learn more about individual risk, screening guidelines have the opportunity to evolve from a primarily age-based approach to a more personalized approach. Instead of asking, “When should every woman start screening?” we may increasingly ask, “What is the right screening strategy for this individual woman based on her unique risk profile?” For some women, that could mean additional screening or earlier surveillance. For others, it could provide reassurance that their current screening approach is appropriate.
Komen: How does incorporating new technologies into health care affect cost? Will it mean savings to patients? And how?
Batchelder: New technologies can involve upfront costs, but the goal is to create value by improving outcomes and making care more efficient. The use of some breast imaging software is included in reimbursement for a mammogram, but some of the newer AI-approaches may not be. However, in breast cancer, earlier detection and better risk assessment have the potential to reduce costs associated with diagnosing cancers at later stages, when treatment is often more complex and expensive. AI may also help health care systems better allocate resources by identifying patients who need additional attention while streamlining workflows for clinicians. Ultimately, the goal is a health care system that delivers the right care to the right patient at the right time.
Statements and opinions expressed are that of the individual and do not express the views or opinions of Susan G. Komen. This information is being provided for educational purposes only and is not to be construed as medical advice. Persons with breast cancer should consult their healthcare provider with specific questions or concerns about their treatment.
