Related Experiment Video
Updated: Apr 25, 2026
07:32
Standardized Histomorphometric Evaluation of Osteoarthritis in a Surgical Mouse Model
Published on: May 6, 2020
12.3K
Automated Objective Scoring of Osteoarthritis Severity in Mouse Medial Tibial Cartilage Using Deep Learning
Ka Hyon Park1, Young-Gwon Kim1, Gyuseok Lee1,2
1Department of Pharmacology and Dental Therapeutics, School of Dentistry, Chonnam National University, Gwangju, Republic of Korea.
Cartilage
|April 24, 2026
Summary
This study developed a VGG16 deep learning model to automate osteoarthritis (OA) grading from mouse cartilage histology. The model achieved high accuracy, showing feasibility for large-scale animal OA research.
Area of Science:
- Biomedical Engineering
- Computational Biology
- Histopathology
Background:
- Osteoarthritis (OA) diagnosis in animal models often relies on manual histologic scoring, which is time-consuming and subjective.
- Automating the assessment of cartilage degradation is crucial for large-scale studies investigating OA.
- Developing objective and efficient methods for OA grading is a significant challenge in preclinical research.
Purpose of the Study:
- To develop and optimize a deep learning model for automated histologic scoring of mouse medial tibial cartilage based on Osteoarthritis Research Society International (OARSI) criteria.
- To improve the accuracy and efficiency of OA grading in histological images from mouse models.
- To evaluate the performance of different deep learning models for this specific task.
Main Methods:
- A dataset of 2,788 Safranin-O-stained cartilage images from 1,000 knees of 520 mice with surgically induced OA was utilized.
- A VGG16-based regression model was employed for horizontal cartilage alignment, and YOLO-v7 was used for tibial cartilage region detection.
- Rotation- and crop-adjusted images were used to train and evaluate three Convolutional Neural Networks (CNNs), with VGG16 demonstrating superior performance.
Main Results:
- Initial low accuracy prompted an expansion of the dataset and implementation of image alignment and cropping algorithms, significantly reducing misclassification.
- The VGG16 model achieved the best performance with a Mean Absolute Error (MAE) of 0.33.
- Performance metrics for VGG16 included precision of 0.680, recall of 0.645, F1-score of 0.653, and accuracy of 0.648.
Conclusions:
- The VGG16 deep learning model demonstrated high concordance with expert assessments for OARSI-based histologic scoring of mouse medial tibial cartilage.
- The developed model shows significant potential for automating OA grading in histological images, facilitating large-scale animal studies.
- This automated approach offers a feasible and accurate method for objective OA assessment in preclinical research.
