Related Experiment Video
Updated: Sep 19, 2026
Platform for Quantitative Detection of Endometrial Immune Cells Based on Immunohistochemistry and Digital Image Analysis
Published on: October 13, 2023
Plasma Cell Counting Methodology Determines Chronic Endometritis Prevalence
Jan Gruszczyński1,2, Kacper Trębacz1,3, Beata Kubiaczyk-Paluch1
1Center for Diagnosis and Treatment of Infertility MedART Poznan Poland.
Purpose:
Histopathological assessment of plasma cells (PCs) via CD138 immunohistochemistry, the diagnostic gold standard for chronic endometritis (CE), is methodologically heterogeneous. Does reported variation in CE prevalence reflect underlying pathobiology, or is it primarily an artifact of counting methodology? We quantified how HPF selection strategy, field count, and magnification influence observed prevalence.
Methods:
We retrospectively studied 886 treatment-naïve patients suspected of CE. CD138-stained biopsies were digitized; a three-pathologist consensus reference validated an AI pipeline for PC detection. Four counting methods were applied across two magnifications (200× vs. 400×) and multiple thresholds.
Results:
Inter-observer reliability was higher for digital (ICC = 0.99) than microscopy-based assessment (ICC = 0.73). The AI pipeline (F1: 0.89-0.97) outperformed manual microscopy (F1: 0.67-0.89) and matched digital assessment (F1: 0.81-1.00). At a fixed threshold and field count (≥ 5 PCs, ×400, 10 HPFs), field-selection strategy alone shifted prevalence from 3.1% (Random Sampling) to 69.5% (Independent Hotspot), p < 0.001. Magnification also significantly altered CE classification (p < 0.001).
Conclusions:
Diagnostic threshold alone does not fully specify a CE criterion; field-selection strategy, field count, and magnification must also be reported for prevalence estimates to be comparable. Hotspot-based and AI-assisted counting are recommended over random field selection for reproducibility and sensitivity.
