Optimizing breast core needle biopsy biomarker throughput using an AI-based workflow.

AIMS: To evaluate the feasibility, appropriate relevance and impact on turnaround time (TAT) of an AI-supported workflow in which AI, integrated in the Laboratory Information System (LIS) as well as in the daily workflow, autonomously triggers breast biomarker testing (ER, PR, HER2, Ki-67) for breast core needle biopsies with a high AI-based likelihood of invasive breast cancer. METHODS AND RESULTS: This retrospective, unicentric implementation study utilized an AI system (IBEX™ Breast) to stratify H&E slides. For cases classified as 'high likelihood for invasive carcinoma', the LIS automatically ordered the standard biomarker panel (ER, PR, HER2 and Ki67), bypassing initial pathologist slide reading. The TAT was measured from the moment of the case assignment to both the availability of IHC staining (TATtech) and the moment of reporting of the biomarkers (TATclin). A pre-implementation phase (December 2024-May 2025; June 2024-November 2025) was compared with a post-implementation phase (June 2025-November 2025). The appropriate relevance of the automated ordering was assessed by evaluating the AI-initiated testing and the relevance of the ordered stains for the final diagnosis. Implementation of the AI-enhanced workflow resulted in a statistically significant reduction in TATtech from approximately 32 h to 23 h, (-30%, P < 0.01) and TATclin reduction of approximately 1 working day (P ≤ 0.03) demonstrating that technical gains translate directly into faster biomarker reporting. Diagnostic accuracy was maintained throughout, with no unnecessary stains triggered by the system. CONCLUSIONS: Integrated, LIS-embedded AI in the daily workflow significantly optimizes breast cancer biomarker turnaround time without compromising the diagnostic process of a pathology laboratory. This demonstrates the transformative potential of AI to enhance laboratory efficiency and support more rapid patient-centred oncologic care.