Related Experiment Video
Updated: Sep 22, 2025
07:29
Author Spotlight: Advanced Techniques for Characterizing Tissue Mineralization in Bone Regeneration Research
Published on: September 27, 2024
897
Resolving complex cartilage structures in developmental biology via deep learning-based automatic segmentation of
Jan Matula1, Veronika Polakova1, Jakub Salplachta1
1Central European Institute of Technology, Brno University of Technology, Purkynova 123, Brno, 61200, Czech Republic.
Scientific Reports
|May 24, 2022
Summary
We developed a convolutional neural network (CNN) to automatically segment embryonic nasal cartilage from micro-CT scans. This AI model significantly reduces analysis time and improves phenotyping of skeletal development.
Area of Science:
- Developmental Biology
- Medical Imaging
- Bioinformatics
Background:
- Embryonic cartilage phenotyping is challenging due to complex 3D structures.
- Manual segmentation of micro-CT data is time-consuming and labor-intensive.
- Automated methods are needed to accelerate skeletal development research.
Purpose of the Study:
- To develop and validate a convolutional neural network (CNN) for automated segmentation of the embryonic nasal capsule.
- To address challenges posed by large image data and limited training sets in developmental studies.
- To improve the efficiency of analyzing cartilaginous skeletal elements in mouse models.
Main Methods:
- Utilized a convolutional neural network (CNN) for image segmentation.
- Trained the CNN model on a unique, manually annotated database of embryonic nasal capsules.
- Optimized the CNN for large image data sizes typical of micro-CT scans.
- Included data from genetically modified mouse embryos with altered phenotypes.
Main Results:
- Achieved a median segmentation accuracy of 84.44% (Dice coefficient) for the cartilaginous nasal capsule.
- Reduced segmentation time from approximately 8 hours (manual) to 130 seconds per sample.
- Demonstrated successful segmentation despite large image dimensions and variations in embryonic phenotypes.
Conclusions:
- The developed CNN model provides an efficient and accurate method for segmenting embryonic cartilage.
- This automated approach significantly accelerates the analysis of micro-CT data for skeletal development research.
- The tool will be valuable for studying developmental diseases and genetic modifications affecting cartilage formation.
