AI Predicts Breast Cancer Recurrence from Preoperative Mammograms

A recent study published in the European Medical Journal (EMJ) highlights a significant advancement in breast cancer management: artificial intelligence (AI) can effectively predict the recurrence of breast cancer using preoperative mammograms. This retrospective study, dated February 18, 2026, reveals that a commercial AI system for mammographic breast cancer detection and diagnosis demonstrates predictive performance on par with existing clinical risk models in assessing recurrence risk after ductal carcinoma in situ (DCIS) treatment. The core finding indicates that an AI score of 73.5% or higher on preoperative mammography images was significantly associated with a higher cumulative incidence rate of ipsilateral recurrence (cancer returning in the same breast) at both five and ten years post-breast-conserving surgery (BCS). Specifically, women with an AI score above 73.5% had a cumulative incidence rate of 4.13% at five years, compared to 0.86% for those with a lower score. At ten years, these rates were 7.26% and 3.72%, respectively. The research, which involved over 1,700 patients who underwent surgery for DCIS between 2012 and 2017, suggests that these AI scores, readily obtained non-invasively, could be instrumental in informing DCIS treatment and surveillance strategies. The AI model's discriminative ability (AUC 70%) was found to be comparable to established clinical models like the Van Nuys Prognostic Index (73%) and superior to the MSKCC nomogram (63%). This indicates that AI-based imaging biomarkers can offer risk prediction on par with traditional clinicopathologic tools, potentially offering more objective assessment compared to subjective imaging features like calcification morphology or breast density, which this study found not to be associated with recurrence risk. This development holds profound global implications, particularly for countries like India, where breast cancer is the most common malignancy among women. India reported 192,020 new cases of breast cancer in 2022, accounting for 26.6% of all cancer cases, with a concerning 98,337 deaths, making it the highest mortality from any cancer type. A significant challenge in India is that over 60% of breast cancer cases are diagnosed at locally advanced or metastatic stages, contributing to a lower survival rate of around 60% compared to 80% in the U.S. This delay in diagnosis is often attributed to limited population-level screening, uneven awareness, and disparities in access to trained breast radiologists, especially between urban and rural areas. AI-supported mammography has the potential to directly address some of these structural gaps in India. Experts like Dr. TPS Bhandari, a surgical oncologist at Apollo Cancer Centre, Hyderabad, emphasize that AI can act as a powerful support tool for clinicians, helping to standardize screening quality, mitigate the shortage of trained manpower, and enable earlier diagnosis. Initiatives in India are already exploring AI for breast cancer screening, such as Niramai's Thermalytix technology, a non-invasive, low-cost screening method based on thermal imaging, which has gained US FDA authorization for its first device. AI-powered mobile clinics are also being deployed to deliver breast cancer screening to rural communities in India, significantly improving accessibility for underserved populations. While the current study focuses on recurrence prediction, other research underscores the broader utility of AI in breast cancer. For instance, a landmark trial published in *The Lancet* (January 2026) demonstrated that AI-supported mammography screening led to better early detection of clinically relevant cancers and a reduction in later diagnoses by 12%. This also translated to fewer aggressive sub-type cancers in the AI group. Such findings suggest that AI can improve screening outcomes and potentially alleviate workload pressures on radiologists, which is a crucial benefit in regions with healthcare resource constraints. However, it is important to acknowledge the limitations and ongoing discussions surrounding AI in medical imaging. A study published in EMJ in December 2025 noted that nearly one in three cancers might be overlooked by current AI tools, particularly in dense breast tissue and for smaller tumors. This highlights that AI should be viewed as an assistive technology rather than a complete replacement for human expertise, and its integration into clinical practice requires careful validation and appropriate safeguards. Overall, the ability of AI to predict breast cancer recurrence from preoperative mammograms represents a promising step towards more personalized and effective cancer management. For India, with its substantial breast cancer burden and unique healthcare challenges, the thoughtful and validated implementation of such AI technologies could significantly improve early detection rates, guide treatment decisions, and ultimately enhance patient survival and quality of life. The focus remains on leveraging AI to augment human capabilities, bridge healthcare gaps, and ensure that advanced diagnostic tools are accessible to all who need them.

Frequently Asked Questions

Read Full Story on Quick Digest