Application

Uses AI analysis of three-dimensional gland architecture in pathology images to predict cancer recurrence risk and support treatment decisions.

Key Benefits

• Captures clinically relevant 3D gland architecture that may be missed in conventional 2D pathology evaluation.
• Provides objective, quantitative prognostic information to imrpove patient risk stratification.
• May offer a more accessible and cost-effective alternative or complement to genomic testing.

Market Summary

Accurately identifying cancer patients at risk of recurrence remains a significant challenge, particularly in prostate cancer where treatment decisions can have long-term impacts on quality of life. Current prognostic methods often rely on pathology grading and clinical assessments that may be subject to inter-observer variability, while genomic tests can be expensive and unavailable in many healthcare settings. Traditional pathology workflows also rely primarily on two-dimensional tissue sections, which may not fully capture the complex three-dimensional structure of glandular networks. There is a growing need for objective, image-based biomarkers that improve risk prediction and help guide personalized treatment strategies.

Technical Summary

Researchers at Emory University have developed an artificial intelligence-based approach that analyzes the three-dimensional architecture of glandular networks from digital pathology images. The technology uses non-destructive 3D pathology to build a representation of glandular structure and extracts quantitative features describing gland shape, branching patterns, spatial organization, curvature, and network topology. These features are evaluated using machine learning models to predict outcomes such as biochemical recurrence risk following radical prostatectomy. Unlike conventional pathology assessments based on 2D tissue sections, the proposed approach leverages 3D tissue architecture to provide a more comprehensive and objective assessment of tumor biology.

Development Stage

Proof-of-concept demonstrated using retrospective prostate cancer pathology datasets.

Publication

Salguero, J., Medina, S., Serafin, R., Wang, R., Abdeltawab, H., Manandhar, S., Mutha, P., Dhamdhere, R., Chow, S., Bishop, K. W., Daniel, R. E., Tokuyama, N., Pathak, T., True, L. D., Corredor, G., Romero, E., Lal, P., Liu, J. T. C., & Madabhushi, A. “AI-Informed Architectural Insights of Three-Dimensional Glandular Networks Identify Prostate Cancer Patients at a Higher Risk of Biochemical Recurrence.” Modern Pathology. 2026 May 25:101018. doi: 10.1016/j.modpat.2026.101018. Online ahead of print. (PMID: 42190816).