Building a scalable fracture detection system with incremental anatomical region integration using ensembled federated fusion learning

A computer vision-based fracture detection system is essential in the modern healthcare system, especially when identifying the precise fracture location is critical. However, most existing deep learning-based systems rely on centralized learning and struggle to generalize across anatomical regions due to data variability. These models also suffer from scalability issues, where adding new classes in a Convolutional Neural Network(CNN) demands retraining of the entire model. To address these limitations, the proposed fracture detection system employs homogeneous transfer learning, heterogeneous federated learning, and a modified federated approach with ensemble strategies, enabling the system to scale seamlessly to different anatomical regions without retraining the entire model. We employ the FracAtlas dataset to perform homogeneous transfer learning and the Bone Fracture Detection: Computer Vision Project dataset with shoulder, forearm, finger, elbow, and humerus anatomical regions to design the proposed fracture detection system. This research proposes two new ensembled federated techniques: the Global Model with Confidence Score-rule-based Ensemble Weights approach (GM-CSEW) and the Global Model with Class-based Ensemble Weights approach (GM-CEW), which use confidence in predictions and anatomical region classes, respectively. GM-CSEW outperformed GM-CEW with ∼3% and ∼4% gains in mAP@50 and mAP@50:95, respectively. It also surpassed the centralized-based learning model with improvements of ∼5% and ∼6% in mAP@50 and mAP@50:95. With its scalable design, the proposed system is flexible and adaptable, allowing it to support new anatomical regions while still maintaining high diagnostic accuracy for use in clinical applications.

Authors

Institutions

Publication Details

Journal
Frontiers in Digital Health
Published
2026-11-16
DOI
https://doi.org/10.3389/fdgth.2026.1927159
Primary Topic
Medical Imaging and Analysis
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Building a scalable fracture detection system with incremental anatomical region integration using ensembled federated fusion learning

Shaik Mullapathi Farooq, Arumuga Arun R, Aditya Anand, Rampriya R. S et al.
6 links
Frontiers in Digital Health
Medical Imaging and Analysis
article

Building a scalable fracture detection system with incremental anatomical region integration using ensembled federated fusion learning

Shaik Mullapathi Farooq, Arumuga Arun R, Aditya Anand, Rampriya R. S, Aprupshri S. R
article en

Abstract

A computer vision-based fracture detection system is essential in the modern healthcare system, especially when identifying the precise fracture location is critical. However, most existing deep learning-based systems rely on centralized learning and struggle to generalize across anatomical regions due to data variability. These models also suffer from scalability issues, where adding new classes in a Convolutional Neural Network(CNN) demands retraining of the entire model. To address these limitations, the proposed fracture detection system employs homogeneous transfer learning, heterogeneous federated learning, and a modified federated approach with ensemble strategies, enabling the system to scale seamlessly to different anatomical regions without retraining the entire model. We employ the FracAtlas dataset to perform homogeneous transfer learning and the Bone Fracture Detection: Computer Vision Project dataset with shoulder, forearm, finger, elbow, and humerus anatomical regions to design the proposed fracture detection system. This research proposes two new ensembled federated techniques: the Global Model with Confidence Score-rule-based Ensemble Weights approach (GM-CSEW) and the Global Model with Class-based Ensemble Weights approach (GM-CEW), which use confidence in predictions and anatomical region classes, respectively. GM-CSEW outperformed GM-CEW with ∼3% and ∼4% gains in mAP@50 and mAP@50:95, respectively. It also surpassed the centralized-based learning model with improvements of ∼5% and ∼6% in mAP@50 and mAP@50:95. With its scalable design, the proposed system is flexible and adaptable, allowing it to support new anatomical regions while still maintaining high diagnostic accuracy for use in clinical applications.

Frontiers in Digital HealthVol. 8
Vellore Institute of Technology University (IN)
Good health and well-being, Industry, innovation and infrastructure
Medical Imaging and Analysis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon