Pathomic Texture Composites from H&E, and Immunohistochemistry of Prostatectomy Specimens for Exploratory Modeling of Time to Biochemical Recurrence

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Background

Pathomic analysis, the extraction of quantitative features from digitized histology, can capture tissue patterns that may complement conventional prognostic factors. We explored whether stain-specific texture composites derived from hematoxylin and eosin (H&E), CD8, and CD163 immunohistochemistry add information to models of time to biochemical recurrence (BCR) after radical prostatectomy.

Methods

This retrospective exploratory study included 15 patients with digitized prostatectomy specimens and linked clinicopathologic data. Handcrafted pathomic features were extracted from image tiles with PyRadiomics, averaged within patients, and standardized. Three stain-specific, equal-weight texture composites developed for this study were integrated with clinicopathologic variables in Cox models with an L2 penalty of 0.1. Four predictor configurations were compared using apparent concordance indices (C-indices). Fitted survival functions were used to explore recurrence probabilities over six-month intervals.

Results

Apparent C-indices were 0.66 (95% CI, 0.38–0.85) for clinicopathologic variables alone, 0.68 (0.38– 0.86) after adding H&E, 0.88 (0.75–0.95) after adding CD8 and CD163, and 0.89 (0.76–0.96) with all three composites. The increase was confined to the immune-stain composites. Composite log-hazard coefficients were approximately 1.06 for CD8, 0.51 for CD163, and −0.10 for H&E.

Conclusion

Stain-specific pathomic composites are a compact way to integrate immunohistochemical and H&E texture with clinicopathologic predictors. The gain in apparent discrimination with CD8 and CD163 texture is hypothesis-generating; larger, representative cohorts with independent validation are needed to establish incremental prognostic value and reliable absolute recurrence probabilities.

Article activity feed