Automated Osteoporosis Detection and Severity Prediction using CNN with VGG19 on Knee Radiographs

Abstract

Automated Osteoporosis Detection And Severity Prediction Using CNN With VGG19 On Knee Radiographs develops a deep learning model to automatically classify osteoporosis severity using knee X-ray images. A Convolutional Neural Network (CNN), combined with a pre-trained VGG19 model, extracts critical features for accurate classification. The model is fine-tuned with multiple a convolutional and max-pooling layers, followed by dense layers for improved learning. Techniques such as data augmentation, learning rate scheduling, and early stopping are applied to enhance model performance. The classification process is optimized using the Adam optimizer and evaluated with standard metrics. The results demonstrate promising accuracy in distinguishing between various osteoporosis classes, providing reliable and consistent assessments to support radiologists in clinical settings.

Introduction

The study “Automated Detection of Osteoporosis and Severity Prediction with CNN and VGG19 on Knee Radiographs” presents a deep learning approach to classify osteoporosis severity from knee X-ray images. It uses a convolutional neural network (CNN) integrated with a pre-trained VGG19 model to extract important features and accurately differentiate between Healthy, Osteopenia, and Osteoporosis classes. The model is enhanced with techniques like data augmentation, learning rate scheduling, and early stopping to improve generalization and prevent overfitting. Training is performed with the Adam optimizer, and evaluation metrics include accuracy, precision, recall, and F1-score.

Results demonstrate that VGG19 outperforms ResNet50 in classification accuracy (91.1% vs. 89.1%) and F1-scores, particularly excelling in detecting osteoporosis. The model’s high Area Under Curve (AUC) of 0.97 further confirms its reliability. Automated classification reduces radiologists' workload by providing fast, consistent, and accurate osteoporosis diagnoses.

The literature review highlights previous research applying CNNs and transfer learning for osteoporosis detection, addressing challenges like limited data and emphasizing deep learning’s potential in medical imaging.

Conclusion

Osteoporosis is a serious condition that affects millions worldwide, and early detection can make a significant difference in patient outcomes. In this project, we successfully developed an automated system for osteoporosis detection and severity classification using deep learning techniques, specifically a CNN model integrated with VGG19. By leveraging knee X-ray images, our approach provides a reliable, efficient, and consistent method for diagnosing osteoporosis stages healthy, osteopenia, and osteoporosis.```` Our results demonstrate that deep learning can assist radiologists in making faster and more accurate diagnoses, reducing subjectivity and improving healthcare efficiency.

References

[1] I. M. Wani and S. Arora, “Osteoporosis diagnosis in knee X-rays by transfer learning based on convolution neural network,” Multimed. Tools Appl., vol. 82, no. 9, pp. 14193–14217, Apr. 2023, doi: 10.1007/s11042-022-13911-y. [2] Y. He, J. Lin, S. Zhu, J. Zhu, and Z. Xu, “Deep learning in the radiologic diagnosis of osteoporosis: a literature review,” J. Int. Med. Res., vol. 52, no. 4, p. 03000605241244754, Apr. 2024, doi: 10.1177/03000605241244754. [3] A. Jaman Shawon, I. I. Mostafa Gazi, H. Rashid Hiya, and A. Roy, “Osteoporosis Prediction Using VGG16 and ResNet50,” Int. J. Innov. Sci. Res. Technol. IJISRT, pp. 2489–2492, May 2024, doi: 10.38124/ijisrt/IJISRT24APR2565. [4] H. Xie et al., “A few-shot learning framework for the diagnosis of osteopenia and osteoporosis using knee X-ray images,” J. Int. Med. Res., vol. 52, no. 9, p. 03000605241274576, Sep. 2024, doi: 10.1177/03000605241274576. [5] M. Shen, “Utilizing Deep Learning for Osteoporosis Diagnosis through Knee X-Ray Analysis,” in Proceedings of the 2024 International Conference on Artificial Intelligence and Communication (ICAIC 2024), vol. 185, Y. Wang, Ed., in Advances in Intelligent Systems Research, vol. 185. , Dordrecht: Atlantis Press International BV, 2024, pp. 553–560. doi: 10.2991/978-94-6463-512-6_58. [6] S. M. Naguib, M. K. Saleh, H. M. Hamza, K. M. Hosny, and M. A. Kassem, “A new superfluity deep learning model for detecting knee osteoporosis and osteopenia in X-ray images,” Sci. Rep., vol. 14, no. 1, p. 25434, Oct. 2024, doi: 10.1038/s41598-024-75549-0. [7] A. M. Sarhan et al., “Knee Osteoporosis Diagnosis Based on Deep Learning,” Int. J. Comput. Intell. Syst., vol. 17, no. 1, p. 241, Sep. 2024, doi: 10.1007/s44196-024-00615-4. [8] T.-Y. Yen, C.-S. Ho, Y.-P. Chen, and Y.-C. Pei, “Diagnostic Accuracy of Deep Learning for the Prediction of Osteoporosis Using Plain X-rays: A Systematic Review and Meta-Analysis,” Diagnostics, vol. 14, no. 2, p. 207, Jan. 2024, doi: 10.3390/diagnostics14020207. [9] A. Siddiqua, R. Hasan, A. Rahman, and A. S. M. Miah, “Computer-Aided Osteoporosis Diagnosis Using Transfer Learning with Enhanced Features from Stacked Deep Learning Modules,” Dec. 12, 2024, arXiv: arXiv:2412.09330. doi: 10.48550/arXiv.2412.09330.

Copyright

Copyright © 2025 Srinivasa Ramanaidu M, Narava Tilak, Bonu Likhitha, Pysaki Pavani, Nallana Vasanth Sai. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.