Automatic segmentation of histological zones in rat distraction osteogenesis using a self-configuring deep learning pipeline
Abstract
Distraction osteogenesis regenerates bone by gradually separating bone segments and inducing new tissue formation within the distraction gap. Although histologic assessment of regenerate composition is essential for evaluating tissue maturation, manual delineation of tissue compartments is labor-intensive and subject to observer variability. This study evaluated a deep learning-based workflow for the automated segmentation of three histologic zones in a rat distraction osteogenesis model: the fibrous interzone, the microcolumn formation zone, and the remodeling zone. High-resolution hematoxylin and eosin- and pentachrome-stained sections were used to train two-dimensional U-Net models in multi-class and single-class settings with five-fold cross-validation and a held-out test set. Performance was assessed using the Dice score, precision, and recall, and was supplemented by a qualitative review of the prediction overlays. On the test set, segmentation was most reliable for the fibrous interzone (Dice 0.76 in H&E; 0.75 in pentachrome), followed by the microcolumn formation zone (Dice 0.63; 0.57). The remodeling zone (Dice 0.29; 0.44) was the most challenging compartment due to its variable and late-stage appearance. Due to incomplete and partly ambiguous ground truth, quantitative metrics were interpreted alongside qualitative histologic assessment. Under this combined evaluation, the pentachrome multi-class configuration yielded the most coherent results. These findings suggest that deep learning-assisted segmentation is a reproducible approach for histologic compartment analysis in distraction osteogenesis. Beyond this application, the evaluated pipeline provides a practical workflow for related histologic segmentation tasks. Implementation in 3D Slicer enables inference, visualization, and area-based quantification in an accessible format for translational research.
Related
- Outcomes Following Lower Limb Corticotomy Using Multiple Drill-Hole and Gigli Saw Techniques During Distraction Osteogenesis: A Prospective Observational Comparative Study.
- Fully automatic airway segmentation in CT images via dual-stage neural network with topology-constrained centerline completion
- Automated segmentation and quantification of histological liver features for MASH/MASLD scoring
- Segmentation and morphological analysis of recanalized coiled aneurysms: Towards an automatic tool for clipping planning in microsurgery
- Semi-automatic segmentation of pelvic bone tumors: Usability testing
BibTeX
@article{dittrich2026automatic,
title = {{Automatic segmentation of histological zones in rat distraction osteogenesis using a self-configuring deep learning pipeline}},
author = {Leonie Dittrich and Jan-Moritz Ramge and Claudia Neunaber and Dominik Wolff},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-76895-x},
url = {https://www.nature.com/articles/s41598-026-76895-x},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Automatic segmentation of histological zones in rat distraction osteogenesis using a self-configuring deep learning pipeline
AU - Leonie Dittrich
AU - Jan-Moritz Ramge
AU - Claudia Neunaber
AU - Dominik Wolff
PY - 2026
DA - 2026/10/21
JO - Scientific Reports
DO - 10.1038/s41598-026-76895-x
UR - https://www.nature.com/articles/s41598-026-76895-x
AB - Distraction osteogenesis regenerates bone by gradually separating bone segments and inducing new tissue formation within the distraction gap. Although histologic assessment of regenerate composition is essential for evaluating tissue maturation, manual delineation of tissue compartments is labor-intensive and subject to observer variability. This study evaluated a deep learning-based workflow for the automated segmentation of three histologic zones in a rat distraction osteogenesis model: the fibrous interzone, the microcolumn formation zone, and the remodeling zone. High-resolution hematoxylin and eosin- and pentachrome-stained sections were used to train two-dimensional U-Net models in multi-class and single-class settings with five-fold cross-validation and a held-out test set. Performance was assessed using the Dice score, precision, and recall, and was supplemented by a qualitative review of the prediction overlays. On the test set, segmentation was most reliable for the fibrous interzone (Dice 0.76 in H&E; 0.75 in pentachrome), followed by the microcolumn formation zone (Dice 0.63; 0.57). The remodeling zone (Dice 0.29; 0.44) was the most challenging compartment due to its variable and late-stage appearance. Due to incomplete and partly ambiguous ground truth, quantitative metrics were interpreted alongside qualitative histologic assessment. Under this combined evaluation, the pentachrome multi-class configuration yielded the most coherent results. These findings suggest that deep learning-assisted segmentation is a reproducible approach for histologic compartment analysis in distraction osteogenesis. Beyond this application, the evaluated pipeline provides a practical workflow for related histologic segmentation tasks. Implementation in 3D Slicer enables inference, visualization, and area-based quantification in an accessible format for translational research.
SN - 2045-2322
ER -