An artificial intelligence model accurately predicted key biomarkers in ductal carcinoma in situ directly from digitized pathology slides, according to a new study aimed at improving active surveillance selection and reducing overtreatment.
Ductal carcinoma in situ (DCIS) is recognized as a non-essential precursor to invasive breast cancer. However, reliable prognostic markers remain limited, and most women diagnosed with DCIS receive intensive treatment regardless of their individual risk of progression. Researchers developed a deep learning pipeline to identify patients who may be candidates for less aggressive management within the context of the LORD trial.
The study focused on women whose screening tests were positive for estrogen receptors, negative for human epidermal growth factor receptor 2, and had grade 1 or 2 DCIS. The researchers trained and tested their artificial intelligence model using digitized pathology slides stained with hematoxylin and eosin from a Dutch multicenter dataset containing 887 patients. External validation was then performed using a dataset of 259 independent patients from the United Kingdom.
Deep learning model predicts key DCIS biomarkers
This model was designed to predict tumor grade, estrogen receptor status, and human epidermal growth factor receptor 2 status directly from pathological images. In the Dutch dataset, model performance was high across all biomarkers, with mean area under the receiver operating characteristic curve of 0.90 for estrogen receptor status, 0.84 for human epidermal growth factor receptor 2 status, and 0.86 for grade prediction.
External validation showed poor but clinically relevant performance with area under the receiver operating characteristic curve values of 0.80 for estrogen receptor status, 0.74 for human epidermal growth factor receptor 2 status, and 0.75 for grade.
DCIS Surveillance Eligibility Identified Using AI
The researchers combined the model outputs to stratify patients according to active surveillance eligibility criteria. This approach achieved a balanced accuracy of 0.81 for the Dutch cohort and 0.64 for the UK cohort.
Negative predictive values for the Dutch and UK datasets were 0.86 and 0.76, respectively, suggesting that the model may be useful in identifying patients who are less likely to require intensive intervention.
Potential to reduce overtreatment
The findings support the potential use of artificial intelligence-assisted pathology tools to guide treatment decisions in DCIS. The researchers concluded that the model generalized across the cohort and reliably predicted clinically relevant biomarkers associated with eligibility for active surveillance.
The authors suggested that these approaches may ultimately support less aggressive management strategies for selected patients with DCIS, reducing unnecessary treatments while maintaining patient safety.
reference
Doyle S et al. Enabling DCIS subtyping: Leveraging foundational models for robust grading and molecular biomarker scoring. Npj breast cancer. 2026; https://doi.org/10.1038/s41523-026-00957-6.
Featured image: Siam from Adobe Stock.
