Related Experiment Video
Updated: Aug 7, 2026
09:24
Tissue Engineering: Construction of a Multicellular 3D Scaffold for the Delivery of Layered Cell Sheets
Published on: October 3, 2014
Deep learning-enabled tissue engineering scaffold classification using cell morphology
Fatemeh Razaviamri1, Clémence Jégard1, Amir Rouhollahi1
1Division of Cardiac Surgery, Department of Surgery, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA.
Acta Biomaterialia
|August 5, 2026
Summary
Analyzing cell 3D morphology offers a non-destructive way to identify biomaterial scaffold types. Computational morphotyping using radiomics and ray-tracing pipelines can accelerate scaffold design and quality control in tissue engineering.
Area of Science:
- Biomaterials Science
- Computational Biology
- Tissue Engineering
Background:
- Biomaterial scaffold properties influence cell behavior, but current evaluation methods are slow and destructive.
- Scalable design and optimization of biomaterials are hindered by limitations in characterization techniques.
- Developing high-throughput, non-destructive methods for scaffold assessment is crucial for regenerative medicine.
Purpose of the Study:
- To investigate if quantitative analysis of single-cell 3D morphology can serve as a non-destructive readout of biomaterial scaffold type.
- To develop and compare computational pipelines for classifying scaffold architecture based on cell morphology.
- To assess the potential of morphology-based computational frameworks for accelerating biomaterial development and quality control.
Main Methods:
- Analyzed 969 high-resolution 3D reconstructions of human bone marrow stromal cells (hBMSCs) cultured on four scaffold types (2D, fibrous, porous, hydrogel).
- Developed a radiomics pipeline using hand-crafted shape descriptors and a feed-forward neural network.
- Developed a ray-tracing pipeline transforming 3D cell structures into 2D distance maps for convolutional neural network (CNN) analysis.
Main Results:
- The radiomics pipeline achieved a best test accuracy of 73.2% for scaffold classification.
- The ray-tracing pipeline achieved a best test accuracy of 72.7% for scaffold classification.
- Both pipelines demonstrated complementary strengths, with radiomics offering interpretability and ray-tracing capturing subtle features.
Conclusions:
- Single-cell 3D morphology quantitatively encodes scaffold type under controlled conditions.
- Computational morphotyping presents a scalable, non-destructive strategy for biomaterial identification and screening.
- This approach could significantly accelerate scaffold design, optimization, and quality control in tissue engineering.
