A deep learning framework for efficient pathology image analysis

Clinical Snapshot

Key Findings

• P Value: Not reported in abstract

• Effect Size: EAGLE outperforms patch aggregation methods by up to 23% in classification performance and achieves the highest overall classification performance across all benchmarked tasks

• Primary Outcome: Overall classification performance across 43 tasks (morphology, biomarker prediction, treatment response, prognosis) spanning nine cancer types, benchmarked against leading slide-level and tile-level foundation models

• Nnt Or Sensitivity: Processing time: 2.27 seconds per slide (>99% reduction compared with existing models); specific sensitivity/specificity or AUC values per task not reported in abstract

• Confidence Interval: Not reported in abstract; variability measures across 43 tasks not provided

Clinical Application

The dramatic reduction in computational requirements (>99% time saving, reduced dependence on high-performance computing) substantially lowers the infrastructure barrier for adoption in standard hospital laboratory settings. Auditability features align with clinical governance requirements. However, prospective clinical validation, regulatory approval, and integration with existing laboratory information systems would be required before routine deployment. Australia's national digital pathology infrastructure is developing, with increasing adoption of whole-slide imaging in major tertiary centres and the Royal College of Pathologists of Australasia (RCPA) providing guidance on digital pathology implementation. EAGLE's reduced computational requirements are particularly relevant for regional and rural Australian laboratories with limited high-performance computing access. The TGA would require prospective clinical validation data and a conformity assessment before EAGLE could be deployed as a Class IIb or III Software as a Medical Device (SaMD) under the TGA's Digital Health regulatory framework. PBS implications are indirect at this stage — the tool does not directly correspond to a PBS-listed test, but could support companion diagnostic workflows for targeted therapies with PBS-listed indications (e.g., MSI/MMR testing for pembrolizumab in colorectal cancer). RACGP and RACGP-affiliated oncology guidelines do not yet address AI-assisted digital pathology at the primary care interface, but specialist oncology and pathology colleges would be the relevant bodies for guideline development. Patients with solid tumours undergoing routine histopathological assessment where digital pathology infrastructure is available; most directly applicable to colorectal, breast, and other common solid tumour types represented in the benchmarking cohorts

Abstract

Artificial intelligence has transformed digital pathology by enabling biomarker prediction from high-resolution whole-slide images. However, current methods are computationally inefficient, processing thousands of redundant tiles per slide and requiring complex aggregation models. We introduce EAGLE (Efficient Approach for Guided Local Examination), a deep learning framework that emulates pathologists by selectively analyzing informative regions. EAGLE combines task-agnostic tile selection with detailed feature extraction and is benchmarked against leading slide- and tile-level foundation models across 43 tasks from nine cancer types spanning morphology, biomarker prediction, treatment response and prognosis. EAGLE outperforms patch aggregation methods by up to 23% and achieves the highest overall classification performance. It processes one slide in 2.27 s, reducing computational time by more than 99% compared with existing models. This efficiency supports rapid and auditable workflows by enabling review of the exact tiles used for each prediction and reducing dependence on high-performance computing. By reliably identifying informative regions and minimizing artifacts, EAGLE provides robust and auditable outputs, supported by systematic negative controls and attention concentration analyses. Its unified embedding enables rapid slide search, integration into multi-omics pipelines and emerging clinical foundation models.

References

1. 1.Neidlinger, P., Lenz, T., Foersch, S., Loeffler, C. M. L., Clusmann, J., Gustav, M., Shaktah, L. A., Langer, R., Dislich, B., Boardman, L. A., French, A. J., Goode, E. L., Gsur, A., Brezina, S., Gunter, M. J., Steinfelder, R., Behrens, H.-M., Röcken, C., Harrison, T., Peters, U., Phipps, A. I., Curigliano, G., Fusco, N., Marra, A., Hoffmeister, M., Brenner, H., & Kather, J. N. (2026). A deep learning framework for efficient pathology image analysis. Nature Communications. https://doi.org/10.1111/his.14059