Autonomous labeling of surgical resection margins using a foundation model

Abstract Assessing resection margins is central to pathological specimen evaluation and has profound implications for patient outcomes. Failure to achieve clear margins significantly increases the risk of local recurrence and may necessitate adjuvant therapies such as chemotherapy or radiotherapy. However, margin assessment faces multiple procedural challenges, including human error during manual tissue margin labeling (physical inking), tissue shrinkage and distortion from fixation and processing, and thermal damage from cautery that obscures cellular features, all of which can compromise accurate margin identification. We demonstrate a virtual inking network (VIN) that localizes the resection margin on whole-slide images (WSIs), producing a standardized, ink-free margin map that helps mitigate limitations of physical inking and enables accurate and reproducible margin measurements. VIN uses a frozen foundation model as the feature extractor and a compact two-layer multilayer perceptron trained for patch-level classification of cautery-consistent features. The dataset comprised 120 hematoxylin and eosin (H&E) stained slides from 12 human tonsil tissue blocks, resulting in ~ 2 TB of uncompressed raw image data, where a board-certified pathologist provided boundary annotations. In blind testing with 20 slides from previously unseen blocks, VIN produced coherent margin overlays that qualitatively aligned with expert annotations across serial sections. Quantitatively, region-level accuracy was ~ 73.3% across the test set, with errors largely confined to limited areas that did not disrupt continuity of the whole-slide margin map. Our results suggest that VIN is capable of capturing cautery-related histomorphology, offering a potential ink-free margin delineation for digital pathology workflows. These proof-of-concept results warrant further validation of VIN as a tool for margin localization in larger, clinically representative cohorts.