Spatial characterization of prostate cancer microenvironment with DRAQ5&Eosin fluorescence imaging and quantitative feature identification

Papachristos, Michail Georgios, Spezi, Emiliano ORCID: https://orcid.org/0000-0002-1452-8813, Fuentes, Carolina ORCID: https://orcid.org/0000-0002-0871-939X, Errington, Rachel J. ORCID: https://orcid.org/0000-0002-8016-4376 and Parthimos, Dimitris ORCID: https://orcid.org/0000-0003-3852-323X 2026. Spatial characterization of prostate cancer microenvironment with DRAQ5&Eosin fluorescence imaging and quantitative feature identification. Journal of Pathology Informatics , 100746. 10.1016/j.jpi.2026.100746

PDF - Accepted Post-Print Version Available under License Creative Commons Attribution. Download (5MB)

Abstract

Prostate cancer progression is routinely quantified by expert pathologists on hematoxylin and eosin (H&E)-stained tissue using Gleason grading. However, tissue processing for H&E staining is very time-consuming, whereas Gleason grading is often subject to inter- and intraobserver variability. This diagnostic variability can be reduced through utilization of mathematically quantitative and standardized metrics to analyze the prostatic tissue microenvironment. In this study, we identify an optimal set of metrics that could potentially be used as a diagnostic biomarker signature, through a combination of tissue architectures and gradient boosting algorithms. We employed fluorescence microscopy imaged prostatic tissue, stained with DRAQ5&Eosin, which allows for rapid real-time staining and imaging of the tissue, as opposed to H&E, and also provides high-resolution nuclear and cytoplasmic images. We then used engineered algorithmic pipelines to extract intensity-based, morphological, and textural features from four different tissue components: epithelial nuclei, epithelial cytoplasm, stromal nuclei, and stromal cytoplasm. We developed classification models to distinguish healthy from cancerous prostatic regions based on the extracted features and identified the most predictive features. Our epithelial-based models achieved area under the curves (AUCs) in the range 0.646–0.896, whereas our stromal-based models attained AUCs in the range 0.742–0.840. Our DRAQ5&Eosin-based models were thus able to identify important tissue specific spatial biomarkers, allowing for the creation of nuclear and cytoplasmic spatial profiles for healthy and cancerous prostatic tissue respectively.

Item Type:	Article
Date Type:	Published Online
Status:	In Press
Schools:	Schools > Engineering Schools > Medicine
Publisher:	Medknow Publications
ISSN:	2229-5089
Date of First Compliant Deposit:	13 October 2026
Date of Acceptance:	28 September 2026
Last Modified:	13 Oct 2026 10:01
URI:	https://orca.cardiff.ac.uk/id/eprint/190193

Actions (repository staff only)

Edit Item