Novel Hybrid Swing Transformer–BiLSTM Model for Accurate Identification of Pelvic Fractures

DOI

10.2991/978-94-6239-674-6_34How to use a DOI?

Keywords

Pelvic fracture; Swin Transformer; BiLSTM; Hybrid deep learning; Medical imaging; Orthopedic diagnostics

Abstract

Pelvic fractures are considered some of the worst orthopedic injuries because of the complex nature of the anatomy, the seriousness of the vascular and visceral injuries that can be caused by the injury, and the fact that they are not easily diagnosed using medical imaging. The traditional radiographic interpretation is usually associated with the lack of inter-observer consistency, de- lays, and lack of sensitivity when it comes to identifying faint fracture lines. In an effort to counter all these, this paper proposes a Novel Hybrid Swin Transformer-BiLSTM Model to identify and classify pelvic fractures in radiographic and computed tomography (CT) images accurately. The Swin Transformer takes advantage of hierarchical shifted window attention to obtain multi-scale spatial representations, which is able to resolve both global and local fracture features. A Bi-directional Long Short-Term Memory (BiLSTM) network further improves these representations with even-outstanding contextual understanding and robustness in fracture recognition, modeling sequential dependencies between the extracted features. To test the model, a curated sample consisting of di-verse pelvic fracture subtypes (such as acetabular, iliac wing, sacral and pubic rami fractures) was used to validate the model. Experimental results showed that the hybrid architecture obtained classification accuracy of 96.4, precision of 95.2, re- call of 94.7 and F1-score of 95.0, which is much better than current deep learning baselines in the form of ResNet, DenseNet and models operates as pure trans- formers. Moreover, Grad-CAM visualizations proved the interpretability of the model since they precisely localized fracture sites. The given hybrid framework has a sub-substantial potential of being a clinical decision-support system and provides radiologists with an effective and trustworthy tool to detect and plan the early treatment of the pelvic fractures.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Aditri Ashish
AU  - Santosh Kumar
AU  - Kumud Dixit
AU  - Arpit Pandey
PY  - 2026
DA  - 2026/05/28
TI  - Novel Hybrid Swing Transformer–BiLSTM Model for Accurate Identification of Pelvic Fractures
BT  - Proceedings of the International Conference on Sustainable Computing and Artificial Intelligence (ICSCAI 2025)
PB  - Atlantis Press
SP  - 405
EP  - 416
SN  - 2352-5401
UR  - https://doi.org/10.2991/978-94-6239-674-6_34
DO  - 10.2991/978-94-6239-674-6_34
ID  - Ashish2026
ER  -