Reproducible Semi-Automated Quantification of Vascularization in Bone Sections Using CD31 Immunohistochemistry and Trainable Weka Segmentation

Quantitative assessment of vascularization is important in bone regeneration research, but CD31-immunohistochemically stained sections are often evaluated manually or semi-quantitatively, limiting reproducibility and comparability. The aim of this study was to establish and validate a reproducible, open-source workflow for semi-automated quantification of CD31-positive area fraction in bone sections using Fiji/ImageJ and Trainable Weka Segmentation (TWS). CD31-immunohistochemically stained rat bone sections from defect/regenerating tissue, femur, tibia, and spine were analyzed. The workflow combined standardized image acquisition, predefined regions of interest, pixel-based TWS classification, extraction of the CD31-positive class, and CD31-positive area normalized to tissue area (CD31.Ar/T.Ar). Manual reference measurements were performed by two independent observers in repeated runs. Manual CD31.Ar/T.Ar measurements showed good retest reliability, with mean coefficients of variation (CV) of 7.17% and 6.84% for observer 1 and observer 2, respectively, and good interobserver agreement. Independently trained TWS classifiers produced highly stable CD31.Ar/T.Ar values, with an overall mean CV of 1.98%. Manual assessment required 3:17 ± 1:35 min per section, whereas the TWS-based workflow separated an initial classifier training step from rapid repeated analysis of larger image sets. This study provides a transparent, reproducible, and time-efficient open-source workflow for semi-automated quantification of CD31-positive vascular area fraction in bone sections and supports its use as a scalable method for vascular histomorphometry in preclinical bone regeneration research.