Poster #91: Lexi Li

Towards intelligent histopathology for enhancing and guiding spatial transcriptomics with ATHENA

C. Li¹*, X. Yu¹, W. Li¹, Z. Jing¹, K. Jin¹, S. Jiang¹, A. Schroeder¹, L. Mao¹, S. Huang¹, M. Loth¹, M. Yuan¹, L.I. Gomez-Bolanos², M. Nasrallah³, L. Wang⁴, G.X. Xu³, A. Huang^5,6, H. Kadara^2,7,8, M. Li¹

Modern spatial-omics assays are powerful but costly and limited in capture area, prohibiting large-scale investigations. Histopathological review of hematoxylin and eosin (H&E) slides provides the opportunity for cheaper, high-resolution, and large-scale tissue characterization. We present ATHENA (Augmented Tissue Histopathology Evaluation and Navigation Assistant), an integrated computational pipeline that transforms routine H&E images into an actionable organizational layer for efficient spatial-omics experimental design and streamlining clinical decision-making. ATHENA contains three major components: a high-efficiency, cellular-resolution segmentation module that overcomes memory and GPU limitations to enable rapid characterization of tissue architecture across diverse cohorts; a human-in-the-loop label propagation module that refines expert annotations and focuses specialist attention on rare cell types or complex transitional states; and an ROI navigation module that reduces manual burden in pathology annotation by identifying most representative regions for expert review. ATHENA further integrates applications such as experimental ROI selection, ensuring that downstream spatial transcriptomics assays capture the full spectrum of cellular heterogeneity; and automated pathologic scoring, thereby minimizing interrater variability and enhancing accuracy. By unifying scalable segmentation, pathologist-in-the-loop label refinement, and intelligent navigation, ATHENA provides the computational infrastructure to link morphology to spatial transcriptomics, facilitating the development of standardized, representative, and clinically relevant maps of human biology.