AI Predicts Breast Cancer Recurrence from Mammograms: New Study 2026
- A new study suggests that artificial intelligence (AI) can accurately predict the likelihood of breast cancer recurrence after initial treatment, using only pre-operative mammograms.
- The research focuses on ductal carcinoma in situ (DCIS), a non-invasive form of breast cancer where abnormal cells are contained within the milk ducts.
- Accurately assessing the risk of recurrence after treatment for DCIS is crucial for guiding follow-up care and treatment decisions.
A new study suggests that artificial intelligence (AI) can accurately predict the likelihood of breast cancer recurrence after initial treatment, using only pre-operative mammograms. The findings, published in the American Journal of Roentgenology in , demonstrate that an AI system’s predictions are comparable to those made using established clinical risk models.
Understanding DCIS and the Need for Improved Risk Assessment
The research focuses on ductal carcinoma in situ (DCIS), a non-invasive form of breast cancer where abnormal cells are contained within the milk ducts. While DCIS is often highly treatable, it carries a risk of developing into invasive breast cancer. Approximately one in five new breast cancer diagnoses made through screening are DCIS. According to data from NHS England, around nine out of every 1,000 women diagnosed with DCIS will develop invasive breast cancer each year.
Accurately assessing the risk of recurrence after treatment for DCIS is crucial for guiding follow-up care and treatment decisions. Current clinical models, such as the Van Nuys Prognostic Index (VNPI) and the MSKCC nomogram, are used to estimate this risk, but there is ongoing interest in exploring whether AI can offer improved or complementary predictive capabilities.
How the Study Was Conducted
Researchers conducted a retrospective analysis of data from over 1,700 women with an average age of 55 who underwent surgery for DCIS between and . All patients had at least one year of postoperative follow-up. The study involved analyzing medical records to identify instances of second breast cancers, either in the same breast (ipsilateral recurrence) or in the opposite breast (contralateral breast cancer).
Preoperative mammograms were processed using a commercially available AI system designed for breast cancer detection and diagnosis. The AI system generated a risk score for each patient, which was then correlated with the observed rates of recurrence.
Key Findings: AI Performance and Comparison to Existing Models
The study identified 28 cases of ipsilateral recurrence after breast-conserving surgery (BCS), seven cases of ipsilateral recurrence after mastectomy, and 25 cases of contralateral breast cancer. A key finding was that an AI score of 73.5% or higher was significantly associated with an increased risk of ipsilateral recurrence at both five and ten years following BCS.
Importantly, the study found that the AI system’s ability to predict ipsilateral recurrence was not significantly different from that of the established clinical models (VNPI and MSKCC nomogram) at both five and ten years. This suggests that the AI tool offers comparable predictive power to existing methods.
The Potential of AI in DCIS Management
Researchers emphasize that the AI scores, derived non-invasively from preoperative mammography, could be a valuable tool for guiding treatment and surveillance strategies for DCIS. By identifying patients at higher risk of recurrence, clinicians may be able to tailor follow-up schedules, imaging protocols, and potentially treatment plans to optimize outcomes.
The non-invasive nature of the AI assessment is particularly noteworthy. It utilizes standard imaging techniques, meaning it doesn’t require additional procedures or expose patients to extra risks. This could make it a practical and accessible tool for widespread use.
Context and Considerations
While this study demonstrates promising results, it’s important to note that the field of AI in breast radiology is still evolving. A study highlighted that current AI tools may overlook nearly one in three breast cancers, particularly in women with dense breast tissue or smaller tumors. However, research also indicates that AI can improve diagnostic accuracy in breast MRI.
The current study focuses specifically on predicting recurrence after DCIS treatment. Further research is needed to determine whether the AI system can be effectively applied to other types of breast cancer or to predict the initial development of breast cancer in screening settings.
The findings underscore the potential of AI to enhance breast cancer care, but also highlight the importance of ongoing evaluation and refinement of these technologies to ensure their accuracy and reliability. As AI continues to develop, it is likely to play an increasingly important role in personalized risk assessment and treatment planning for breast cancer patients.
Reference
Yoon JH et al. Commercially available artificial intelligence score on preoperative mammography for prediction of future breast cancer after DCIS treatment. AJR. ;DOI:10.2214/AJR.25.34364.
National Health Service (NHS) England. Ductal carcinoma in situ (DCIS) data story. . Available at: https://digital.nhs.uk/ndrs/data/data-stories/ductal-carcinoma-in-situ. Last accessed .