Related Experiment Video
Updated: Jun 5, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
BigOrthoATD.Net: A scalable and adaptable distributed deep learning framework for multi-class orthopedic
Haider A Alwzwazy1, Laith Alzubaidi1, Zehui Zhao2
1School of Mechanical, Medical, and Process Engineering, Queensland University of Technology, Brisbane, 4000, QLD, Australia.
Summary
BigOrthoATD.Net offers a scalable, privacy-preserving framework for decentralized medical image analysis. This approach significantly improves accuracy and efficiency in orthopedic AI, overcoming limitations of current methods.
Area of Science:
- Artificial Intelligence
- Medical Imaging
- Orthopedics
Background:
- Multi-class medical image classification faces challenges in data management, adaptability, and resource constraints.
- Existing AI approaches struggle to address these issues simultaneously, limiting clinical scalability.
Purpose of the Study:
- Introduce BigOrthoATD.Net, a unified, serverless, decentralized learning framework for orthopedic image analysis.
- Redefine scalability, adaptability, and efficiency in AI for healthcare, particularly in resource-limited settings.
Main Methods:
- Developed a decentralized learning framework enabling privacy-preserving knowledge fusion and multimodal integration (X-ray, CT).
- Designed for progressive scalability across distributed clinical nodes and continual learning without retraining.
- Implemented a serverless architecture for efficient operation under resource constraints.
Main Results:
- BigOrthoATD.Net achieved 97.0% accuracy across 50 orthopedic classes on 13 simulated decentralized nodes.
- Outperformed swarm learning (70.8%) and centralized learning (84.8%).
- Federated learning failed to converge at scale under identical conditions.
Conclusions:
- BigOrthoATD.Net sets a new benchmark for decentralized medical imaging, surpassing existing frameworks in accuracy and scalability.
- Demonstrates efficient operation in low-resourced settings while handling diverse orthopedic classes and multimodal data.
- Enables privacy-preserving, adaptable, and scalable AI solutions for clinical healthcare.
