EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Federated Deep Learning for Privacy-Preserving Musculoskeletal Image Analysis and Collaborative Orthopedic Decision Support

R. Pandiyarajan, Dr. Dama Kondaiah Naidu, Dr. Jishnu Jonnalagadda, Ajit Pal Singh, Ashish Kumar Shukla

Computer and Decision Making: An International Journal · 04.11.2026 · 3 (6S) · 3127-3138

DOI: 10.1922/comdem.v3i6s.19058 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 04.11.2026. Verfasst von R. Pandiyarajan und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Musculoskeletal fracture detection requires accurate radiographic interpretation to support timely orthopedic diagnosis and clinical decision-making. This study evaluated the effectiveness of federated transfer learning for musculoskeletal fracture classification and its potential application in collaborative orthopedic decision support. A total of 4,024 radiographs from the FracAtlas dataset were analyzed using pretrained ResNet50 feature representations. Centralized and federated classification models were developed, with three simulated healthcare clients participating in ten Federated Averaging (FedAvg) communication rounds. Diagnostic performance was assessed using accuracy, precision, sensitivity, specificity, F1-score, and classification errors. The centralized model achieved higher accuracy (85.43%), precision (57.69%), specificity (88.91%), and F1-score (0.6303). Conversely, the federated model demonstrated greater sensitivity (82.41% versus 69.44%), reducing missed fractures from 33 to 19 cases, although false-positive predictions increased from 55 to 127. These findings demonstrate distinct diagnostic trade-offs between centralized and federated optimization, particularly concerning fracture detection sensitivity and classification reliability. The proposed framework demonstrated the feasibility of simulated collaborative classifier training using shared model parameters. Further multicenter evaluation and optimization of diagnostic thresholds may strengthen its potential application in orthopedic screening. Overall, the study provides evidence supporting further investigation of federated learning for collaborative musculoskeletal image analysis and clinical decision support.

Zitieren

APA

Pandiyarajan, R., Naidu, D. D. K., Jonnalagadda, D. J., Singh, A. P., & Shukla, A. K. (2026). Federated Deep Learning for Privacy-Preserving Musculoskeletal Image Analysis and Collaborative Orthopedic Decision Support. Computer and Decision Making: An International Journal, 3(6S), 3127-3138. https://doi.org/10.1922/comdem.v3i6s.19058

Vancouver

Pandiyarajan R, Naidu DDK, Jonnalagadda DJ, Singh AP, Shukla AK. Federated Deep Learning for Privacy-Preserving Musculoskeletal Image Analysis and Collaborative Orthopedic Decision Support. Computer and Decision Making: An International Journal. 2026;3(6S):3127-3138. doi:10.1922/comdem.v3i6s.19058

BibTeX
@article{Pandiyarajan2026,
  author = {Pandiyarajan, R. and Naidu, Dr. Dama Kondaiah and Jonnalagadda, Dr. Jishnu and Singh, Ajit Pal and Shukla, Ashish Kumar},
  title = {Federated Deep Learning for Privacy-Preserving Musculoskeletal Image Analysis and Collaborative Orthopedic Decision Support},
  journal = {Computer and Decision Making: An International Journal},
  year = {2026},
  volume = {3},
  number = {6S},
  pages = {3127--3138},
  doi = {10.1922/comdem.v3i6s.19058}
}
Kostenlos starten →