Pre-trained Deep Models for Automated Detection of 10 Types of Bone Fractures: A Comparative Study with ResNet50, EfficientNetB3 and MobileNet
Article Sidebar
Main Article Content
Abstract
Rapid and accurate diagnosis of bone fractures is key to avoiding complications. Convolutional neural networks (CNN) have demonstrated high efficiency in the classification of radiographs, achieving results comparable to those of specialists. This study evaluated MobileNet, ResNet50, and EfficientNetB3 to classify 10 fracture types using 1,129 balanced images: 90% for training (1,017) and 10% for validation (112). Preprocessing included resizing to 256x256 pixels, conversion to RGB, normalization and one-hot encoding. The training was carried out in Google Colab with RMSprop optimizer, categorical_crossentropy loss and early stopping. ResNet50 and EfficientNetB3 achieved 95.54% accuracy and F1-score >0.94, outperforming MobileNet (91.96% accuracy). Confusions occurred mainly in visually similar fractures. The CNNs evaluated are viable for the automatic classification of bone fractures, constituting a useful diagnostic support tool, especially in areas with a shortage of specialists. It is proposed as future work to include an independent test set and data augmentation to improve generalization.
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
When an author creates an article and publishes it in a journal, the copyright passes to the journal as part of the publishing agreement. Therefore, the journal becomes the owner of the rights to reproduce, distribute and sell the article. The author retains some rights, such as the right to be recognized as the creator of the article and the right to use the article for his or her own scholarly or research purposes, unless otherwise agreed in the publication agreement.
How to Cite
References
T. S. Omofoye et al., «Backlogs in formal interpreta-tion of radiology examinations: a pilot global sur-vey», Clin Imaging, vol. 106, feb. 2024, doi: 10.1016/j.clinimag.2023.110049.
G. Wahid, Ammara Haroon, Mehreen Samad, y Naila Tamkeen, «Causes of Delay in Radiological Reporting and ways to Reduce them», Journal of Sai-du Medical College, Swat, vol. 12, n.o 3, pp. 133-137, sep. 2022, doi: 10.52206/jsmc.2022.12.3.697.
T. Urakawa, Y. Tanaka, S. Goto, H. Matsuzawa, K. Watanabe, y N. Endo, «Detecting intertrochanteric hip fractures with orthopedist-level accuracy using a deep convolutional neural network», Skeletal Radiol, vol. 48, pp. 239-244, 2018, doi: 10.1007/s00256-018-3016-3.
Y. Xie et al., «Artificial intelligence diagnostic model for multi-site fracture X-ray images of extremities based on deep convolutional neural networks», Quant Imaging Med Surg, vol. 14, n.o 2, pp. 1930-1943, feb. 2024, doi: 10.21037/qims-23-878.
N. T. Htun y K. M. M. Tun, «Fuzzy-based Image Enhancement and Ensemble CNN Model for Bone Fracture Detection and Classification System», en 2024 5th International Conference on Advanced Infor-mation Technologies (ICAIT), 2024, pp. 1-6. doi: 10.1109/ICAIT65209.2024.10754918.
K. Mittal, K. S. Gill, R. Chauhan, y A. Kapruwan, «Innovative Fracture Diagnosis: MobileNet CNN Approach for Precise Bone Fracture Detection and Classification», en 2024 International Conference on In-telligent Systems for Cybersecurity (ISCS), 2024, pp. 1-5. doi: 10.1109/ISCS61804.2024.10581396.
L. O. Carmo et al., «An increasing number of convo-lutional neural networks for fracture recognition and classification in orthopaedics», Bone Jt Open, vol. 2, pp. 879-885, 2021, doi: 10.1302/2633-1462.210.BJO-2021-0133.
F. Mohammad, S. Al-Ahmadi, y J. Al-Muhtadi, «Block-Deep: A Hybrid Secure Data Storage and Diagnosis Model for Bone Fracture Identification of Athlete From X-Ray and MRI Images», IEEE Access, vol. 11, pp. 142360-142370, 2023, doi: 10.1109/ACCESS.2023.3330914.
M. Yaseen, M. Ali, S. Ali, A. Hussain, M. Il Joo, y H. C. Kim, «Cervical Spine Fracture Detection and Classification Using Two-Stage Deep Learning Methodology», IEEE Access, vol. 12, pp. 72131-72142, 2024, doi: 10.1109/ACCESS.2024.3398061.
S. C. Medaramatla, C. V. Samhitha, S. D. Pande, y S. R. Vinta, «Detection of Hand Bone Fractures in X-Ray Images Using Hybrid YOLO NAS», IEEE Ac-cess, vol. 12, pp. 57661-57673, 2024, doi: 10.1109/ACCESS.2024.3379760.
L. Mu et al., «Fine-Tuned Deep Convolutional Net-works for the Detection of Femoral Neck Fractures on Pelvic Radiographs: A Multicenter Dataset Vali-dation», IEEE Access, vol. 9, pp. 78495-78503, 2021, doi: 10.1109/ACCESS.2021.3082952.
S. Torne et al., «VGG-16, VGG-16 With Random Forest, Resnet50 With SVM, and EfficientNetB0 With XGBoost-Enhancing Bone Fracture Classifica-tion in X-Ray Using Deep Learning Models», IEEE Access, vol. 13, pp. 25568-25577, 2025, doi: 10.1109/ACCESS.2025.3534818.
C. Maier, J. B. Thatcher, V. Grover, y Y. K. Dwivedi, «Cross-sectional research: A critical perspective, use cases, and recommendations for IS research», Int J Inf Manage, vol. 70, p. 102625, jun. 2023, doi: 10.1016/J.IJINFOMGT.2023.102625.
«Bone Break Classification Image Dataset». Accedi-do: 14 de julio de 2025. [En línea]. Disponible en: https://www.kaggle.com/datasets/pkdarabi/bone-break-classification-image-dataset
N. Vasker, M. Hasan, M. Nuha, S. Jahan, M. Tahsin, y Md. Y. Emon, Real-time Classification of Bone Frac-tures Utilizing Different Convolutional Neural Network Approaches. 2023. doi: 10.1109/ICCIT60459.2023.10441387.
S. Turk, O. Bingol, A. Coşkunçay, y T. Aydin, «The impact of implementing backbone architectures on fracture segmentation in X-ray images», Engineering Science and Technology, an International Journal, p., 2024, doi: 10.1016/j.jestch.2024.101883.
P. M, S. M, R. N, y S. S, «Edge AI-based Bone Frac-ture Detection using TFlite», International Journal of Innovative Research in Advanced Engineering, p., 2025, doi: 10.26562/ijirae.2025.v1204.04.
A. Khanapure, H. Kashyap, A. Bidargaddi, S. Habib, y A. Anand, «Bone Fracture Detection with X-Ray images using MobileNet V3 Architecture», 2024 IEEE 9th International Conference for Convergence in Technology (I2CT), pp. 1-8, 2024, doi: 10.1109/I2CT61223.2024.10544356.
R. Bhuria y S. Gupta, «X-Ray Insights: Comprehen-sive Dataset for Bone Fracture Detection Across Di-verse Anatomical Regions», 2024 5th International Conference on Smart Electronics and Communication (ICOSEC), pp. 1242-1247, 2024, doi: 10.1109/ICOSEC61587.2024.10722406.
M. Goel y G. Singh, «Fracture Detection using Mo-bileNet Model», 2024 4th International Conference on Ubiquitous Computing and Intelligent Information Sys-tems (ICUIS), pp. 1574-1579, 2024, doi: 10.1109/ICUIS64676.2024.10866070.
S. Thota, P. Kandukuru, M. Sundaram, A. Ali, S. Muzamil, y B. Bindu, «Deep Learning based Bone Fracture Detection», 2024 International Conference on Smart Systems for applications in Electrical Sciences (ICSSES), pp. 1-7, 2024, doi: 10.1109/ICSSES62373.2024.10561360.
S. R. Sannasi Chakravarthy, N. Bharanidharan, C. Vinothini, V. Vinoth Kumar, T. R. Mahesh, y S. Guluwadi, «Adaptive Mish activation and ranger optimizer-based SEA-ResNet50 model with ex-plainable AI for multiclass classification of COVID-19 chest X-ray images», BMC Med Imaging, vol. 24, n.o 1, dic. 2024, doi: 10.1186/s12880-024-01394-2.
A. S. B. Karno et al., «Classification of cervical spine fractures using 8 variants EfficientNet with transfer learning», International Journal of Electrical and Com-puter Engineering (IJECE), p., 2023, doi: 10.11591/ijece.v13i6.pp7065-7077.
S.-T. Huang, L.-R. Liu, M.-F. Tsai, M.-Y. Huang, y H.-W. Chiu, «Developing a Deep Learning Model Using Transfer Learning from EfficientNet-b3 to De-tect Knee Fracture on X-ray Images», Proceedings of the 2023 7th International Conference on Medical and Health Informatics, p., 2023, doi: 10.1145/3608298.3608352.
K. Teeyapan, «Abnormality Detection in Musculo-skeletal Radiographs using EfficientNets», 2020 24th International Computer Science and Engineering Confer-ence (ICSEC), pp. 1-6, 2020, doi: 10.1109/ICSEC51790.2020.9375275.
H. Amin, A. Darwish, A. E. Hassanien, y M. Soli-man, «End-to-End Deep Learning Model for Corn Leaf Disease Classification», IEEE Access, vol. 10, pp. 31103-31115, 2022, doi: 10.1109/ACCESS.2022.3159678.
L. Yeh et al., «A deep learning-based method for the diagnosis of vertebral fractures on spine MRI: retro-spective training and validation of ResNet», Europe-an Spine Journal, vol. 31, pp. 2022-2030, 2022, doi: 10.1007/s00586-022-07121-1.
Y. Li et al., «Differential diagnosis of benign and malignant vertebral fracture on CT using deep learning», Eur Radiol, vol. 31, pp. 9612-9619, 2021, doi: 10.1007/s00330-021-08014-5.
K. Saini y R. Devi, «A systematic scoping review of the analysis of COVID-19 disease using chest X-ray images with deep learning models», Journal of Au-tonomous Intelligence, vol. 7, dic. 2023, doi: 10.32629/jai.v7i2.928.
A. P. Bonifaz, C. C. Rodriguez, y R. P. Esparza, «Diagnostic Reference Levels for Common X-ray Procedures in Peru», Cureus, vol. 13, p., 2021, doi: 10.7759/cureus.18566.