Lightweight YOLOv8 for Multi-Anatomical Fracture Detection in Emergency Radiology
Main Article Content
Abstract
Although deep learning-based fracture detection models have shown potential results for their diagnostic parameters under experimental settings, limited research has been conducted on applying light weight one-stage object detection frameworks to clinical emergency radiological procedures for varied anatomical regions. An off-the-shelf YOLOv8n deep learning architecture was tested using a dataset consisting of 4,148 plain X-ray images from various anatomical sites like elbow, finger, forearm, humerus, shoulder, and wrist with a standard data augmentation process and a fixed spatial resolution of 416×416 pixels. The performance metrics for algorithmic localization was determined using conventional mean average precision measures, precision values, and recall rates at 0.50 and 0.50:0.95 intersection-over-union levels with a fine-grained class-by-class breakdown to identify specific structural weaknesses. Terminal localization module showed an aggregate mAP@50 of 0.262, paired with a macro-averaged precision of 0.389 and a macro-averaged recall of 0.277 along with a drastic multi-level mAP@50-95 drop to 0.095. Diagnostic performance exhibited high variance where the localization of humerus demonstrated a maximum mAP@50 of 0.584, while the separate localization of elbow, shoulder, and wrist were all found to be below 0.130 threshold level. Lightweight one-step regression models lack adequate localization sensitivity for automated multilocation skeletal fracture detection. The discovered accuracy-implementation trade-off implies that the use of anatomy-specific model ensembles, larger input sizes, and dual-stage system design is an absolute must prior to any ethical and legal justification of automation in the realm of orthopedic emergencies.
Article Details
This work is licensed under a Creative Commons Attribution 4.0 International License.
References
[1] R. Y. L. Kuo et al., “Artificial Intelligence in Fracture Detection: A Systematic Review and Meta-Analysis,” Radiology, vol. 304, no. 1, pp. 50–62, Jul. 2022, doi: 10.1148/radiol.211785.
[2] L. Duron et al., “Assessment of an AI Aid in Detection of Adult Appendicular Skeletal Fractures by Emergency Physicians and Radiologists: A Multicenter Cross-sectional Diagnostic Study,” Radiology, vol. 300, no. 1, pp. 120–129, Jul. 2021, doi: 10.1148/radiol.2021203886.
[3] V. Bousson et al., “Artificial Intelligence for Detecting Acute Fractures in Patients Admitted to an Emergency Department: Real-Life Performance of Three Commercial Algorithms,” Academic Radiology, vol. 30, no. 10, pp. 2118–2139, Oct. 2023, doi: 10.1016/j.acra.2023.06.016.
[4] T. Sukhbaatar, A. Davies, A. Koye, M. Hashem, and S. Sivaloganathan, “Artificial intelligence in virtual fracture clinics: a systematic review of imaging and clinical-text tools,” Journal of Orthopaedic Surgery and Research, vol. 21, no. 1, p. 176, Feb. 2026, doi: 10.1186/s13018-025-06656-5.
[5] A. Guermazi et al., “Improving Radiographic Fracture Recognition Performance and Efficiency Using Artificial Intelligence,” Radiology, vol. 302, no. 3, pp. 627–636, Mar. 2022, doi: 10.1148/radiol.210937.
[6] R. Bachmann et al., “Improving traumatic fracture detection on radiographs with artificial intelligence support: a multi-reader study,” BJR|Open, vol. 6, no. 1, Dec. 2023, doi: 10.1093/bjro/tzae011.
[7] Y. Zhang, X. Li, F. Wang, B. Wei, and L. Li, “A Comprehensive Review of One-stage Networks for Object Detection,” in 2021 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC), IEEE, Aug. 2021, pp. 1–6. doi: 10.1109/ICSPCC52875.2021.9564613.
[8] K. He, G. Gkioxari, P. Dollár, and R. B. Girshick, “Mask R-CNN,” 2017 IEEE International Conference on Computer Vision (ICCV), pp. 2980–2988, 2017, [Online]. Available: https://api.semanticscholar.org/CorpusID:206771194
[9] J. Zou and M. R. Arshad, “Detection of whole body bone fractures based on improved YOLOv7,” Biomedical Signal Processing and Control, vol. 91, p. 105995, May 2024, doi: 10.1016/j.bspc.2024.105995.
[10] D. Bhattacharya, S. Das, T. Biswas, and G. Mandal, “ResNet based backbone integrated YOLO framework for bone fracture detection,” Scientific Reports, vol. 16, no. 1, p. 12954, Mar. 2026, doi: 10.1038/s41598-026-41782-y.
[11] T. Alam et al., “An Integrated Approach using YOLOv8 and ResNet, SeResNet & Vision Transformer (ViT) Algorithms based on ROI Fracture Prediction in X-ray Images of the Elbow,” Current Medical Imaging Formerly Current Medical Imaging Reviews, vol. 20, Nov. 2024, doi: 10.2174/0115734056309890240912054616.
[12] A. Ahmed, A. S. Imran, A. Manaf, Z. Kastrati, and S. M. Daudpota, “Enhancing wrist abnormality detection with YOLO: Analysis of state-of-the-art single-stage detection models,” Biomedical Signal Processing and Control, vol. 93, p. 106144, Jul. 2024, doi: 10.1016/j.bspc.2024.106144.
[13] C. Chien, R. Ju, K. Chou, and J. Chiang, “YOLOv9 for fracture detection in pediatric wrist trauma X‐ray images,” Electronics Letters, vol. 60, no. 11, Jun. 2024, doi: 10.1049/ell2.13248.
[14] F. Hržić, S. Tschauner, E. Sorantin, and I. Štajduhar, “Fracture Recognition in Paediatric Wrist Radiographs: An Object Detection Approach,” Mathematics, vol. 10, no. 16, p. 2939, Aug. 2022, doi: 10.3390/math10162939.
[15] S. C. Medaramatla, C. V. Samhitha, S. D. Pande, and S. R. Vinta, “Detection of Hand Bone Fractures in X-Ray Images Using Hybrid YOLO NAS,” IEEE Access, vol. 12, pp. 57661–57673, 2024, doi: 10.1109/ACCESS.2024.3379760.
[16] H. Wang, Z. Li, and D. Zhang, “A Two-Stage Deep Learning Method for Auxiliary Diagnosis of Upper Limb Fractures Based on ResNet-50 and Enhanced YOLO,” Mathematics, vol. 13, no. 11, p. 1858, Jun. 2025, doi: 10.3390/math13111858.
[17] M. Nikouei et al., “Small object detection: A comprehensive survey on challenges, techniques and real-world applications,” Intelligent Systems with Applications, vol. 27, p. 200561, Sep. 2025, doi: 10.1016/j.iswa.2025.200561.
[18] P. K. Darabi, “Bone Fracture Detection: Computer Vision Project,” Jan. 18, 2024. doi: 10.13140/RG.2.2.14400.34569.
[19] Z. Cai, K. Zhou, and Z. Liao, “A Systematic Review of YOLO-Based Object Detection in Medical Imaging: Advances, Challenges, and Future Directions,” Computers, Materials & Continua, vol. 85, no. 2, pp. 2255–2303, 2025, doi: 10.32604/cmc.2025.067994.
[20] E. J. Topol, “High-performance medicine: the convergence of human and artificial intelligence,” Nature Medicine, vol. 25, no. 1, pp. 44–56, Jan. 2019, doi: 10.1038/s41591-018-0300-7.
[21] K. Goddard, A. Roudsari, and J. C. Wyatt, “Automation bias: a systematic review of frequency, effect mediators, and mitigators,” Journal of the American Medical Informatics Association, vol. 19, no. 1, pp. 121–127, Jan. 2012, doi: 10.1136/amiajnl-2011-000089.
[22] E. Nagy, M. Janisch, F. Hržić, E. Sorantin, and S. Tschauner, “A pediatric wrist trauma X-ray dataset (GRAZPEDWRI-DX) for machine learning,” Scientific Data, vol. 9, no. 1, p. 222, May 2022, doi: 10.1038/s41597-022-01328-z.
[23] M. W. Kim et al., “Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room,” Clinical and Experimental Emergency Medicine, vol. 8, no. 2, pp. 120–127, Jun. 2021, doi: 10.15441/ceem.20.091.
[24] M. F. Russe et al., “AI-based X-ray fracture analysis of the distal radius: accuracy between representative classification, detection and segmentation deep learning models for clinical practice,” BMJ Open, vol. 14, no. 1, p. e076954, Jan. 2024, doi: 10.1136/bmjopen-2023-076954.
[25] T.-Y. Lin et al., “Microsoft COCO: Common Objects in Context,” D. Fleet, T. Pajdla, B. Schiele, and T. Tuytelaars, Eds., Cham: Springer International Publishing, 2014, pp. 740–755. doi: 10.1007/978-3-319-10602-1_48.
[26] M. Cohen et al., “Artificial intelligence vs. radiologist: accuracy of wrist fracture detection on radiographs,” European Radiology, vol. 33, no. 6, pp. 3974–3983, Dec. 2022, doi: 10.1007/s00330-022-09349-3.
[27] Z. Huang, X. Wang, L. Huang, C. Huang, Y. Wei, and W. Liu, “CCNet: Criss-Cross Attention for Semantic Segmentation,” in 2019 IEEE/CVF International Conference on Computer Vision (ICCV), IEEE, Oct. 2019, pp. 603–612. doi: 10.1109/ICCV.2019.00069.
[28] N. Ramadanov et al., “Artificial intelligence-guided distal radius fracture detection on plain radiographs in comparison with human raters,” Journal of Orthopaedic Surgery and Research, vol. 20, no. 1, p. 468, May 2025, doi: 10.1186/s13018-025-05888-9.
[29] A. Kekatpure, A. Kekatpure, S. Deshpande, and S. Srivastava, “Development of a diagnostic support system for distal humerus fracture using artificial intelligence,” International Orthopaedics, vol. 48, no. 5, pp. 1303–1311, May 2024, doi: 10.1007/s00264-024-06125-4.
[30] E. Ozkaya, F. E. Topal, T. Bulut, M. Gursoy, M. Ozuysal, and Z. Karakaya, “Evaluation of an artificial intelligence system for diagnosing scaphoid fracture on direct radiography,” European Journal of Trauma and Emergency Surgery, vol. 48, no. 1, pp. 585–592, Feb. 2022, doi: 10.1007/s00068-020-01468-0.
[31] J. W. Choi et al., “Using a Dual-Input Convolutional Neural Network for Automated Detection of Pediatric Supracondylar Fracture on Conventional Radiography,” Investigative Radiology, vol. 55, no. 2, pp. 101–110, Feb. 2020, doi: 10.1097/RLI.0000000000000615.
[32] B. Zeng, H. Wang, J. Xu, P. Tu, L. Joskowicz, and X. Chen, “Two-Stage Structure-Focused Contrastive Learning for Automatic Identification and Localization of Complex Pelvic Fractures,” IEEE Transactions on Medical Imaging, vol. 42, no. 9, pp. 2751–2762, Sep. 2023, doi: 10.1109/TMI.2023.3264298.
[33] R. I. Sumon et al., “Automatic Fracture Detection Convolutional Neural Network with Multiple Attention Blocks Using Multi-Region X-Ray Data,” Life, vol. 15, no. 7, p. 1135, Jul. 2025, doi: 10.3390/life15071135.
[34] M. Tariq and K. Choi, “YOLO11-Driven Deep Learning Approach for Enhanced Detection and Visualization of Wrist Fractures in X-Ray Images,” Mathematics, vol. 13, no. 9, p. 1419, Apr. 2025, doi: 10.3390/math13091419.