DEVELOPMENT OF A CONVOLUTIONAL NEURAL NETWORK TO ASSESS THE SEVERITY OF KNEE OSTEOARTHRITIS
Cite as: A.S. Mannaa G.V. Muratova. Development of a convolutional neural network to assess the severity of knee osteoarthritis // Izvestiya SFedU. Engineering Sciences – 2024. – N. 6. - P. 155-163. doi: 10.18522/2311-3103-2024-6-155-163
Abstract
A method In this paper, we propose a novel method for the automated assessment of knee osteoarthritis
severity, utilizing advanced machine learning techniques, specifically a deep neural network. Osteoarthritis
is one of the most prevalent degenerative joint diseases, and its timely diagnosis is crucial for
ensuring effective treatment. Traditional methods for visually assessing X-ray images of the knee joint
present several limitations, including subjectivity and reliance on the experience of the clinician. Therefore,
the development of automated medical image analysis techniques has become increasingly relevant.
Osteoarthritis of the knee joint is one of the most common and severe degenerative diseases leading to a
significant decrease in the quality of life of patients. Traditional methods of diagnosing osteoarthritis,
such as visual assessment of X-ray images, depend on the subjective opinion of a specialist and his experience,
which can lead to variations in the accuracy of diagnosis and timely detection of pathology. Therefore,
the development and implementation of methods for automated analysis of medical images is highly
relevant and has potential clinical value. In this study, we designed and trained a specialized neural network
based on the ResNet-34 architecture, which has demonstrated significant effectiveness in solving
computer vision problems. The network was modified to incorporate two parallel branches, each containing a spiral linear structure and four hidden layers. This design enables more precise identification of the
knee joint area. Additionally, the architecture facilitates optimization of the loss function to account for
varying pathological characteristics, such as different degrees of joint degradation, and to address the
issue of class imbalance—a common challenge in medical imaging datasets. To further enhance model
performance, the neural network was trained on two distinct datasets stratified by gender (male and female).
This approach improved overall image quality and reduced the impact of noise introduced by artifacts
during radiographic imaging. Moreover, we employed the ImagePixelSpacing technique during data
preparation to standardize image resolution at 256 × 256 pixels, allowing for more accurate processing
of fine details and structures within the knee joint. The network training employed state-of-the-art optimization
techniques, resulting in a high level of classification accuracy. To evaluate the effectiveness of the
proposed model, the Kappa test was utilized, confirming the reliability of baseline determinations.
The model achieved an average accuracy of 93.76%, as demonstrated by the multiclass T-test, indicating
its strong potential for clinical application. Additionally, the model’s area under the curve (AUC) score
was 0.97, surpassing the results reported in previous studies in this domain. In conclusion, this research
contributes significantly to the field of medical informatics and computer-based medical image analysis by
offering an innovative solution for the automated assessment of osteoarthritis. This method has the potential
to profoundly improve diagnostic accuracy and treatment outcomes in clinical settings. In addition,
these results demonstrate the potential of the model as a reliable tool for automated assessment of the
degree of osteoarthritis, which can not only improve the accuracy of diagnosis, but also facilitate the work
of medical specialists. Further research may include adapting the model to analyze other joints and integrating
additional functionality, such as predicting disease progression based on sequential scans.
References
on deep learning: data from the osteoarthritis initiative, Journal of Healthcare Engineering, 2021,
pp. 1-10. Available at: https://doi.org/10.1155/2021/5586529.
2. Tolpadi A.A., Lee J.J., Pedoia V. and Majumdar S. Deep learning predicts total knee replacement from
magnetic resonance images—scientific reports, 2020, 10 (1), pp. 6371. Available at:
https://doi.org/10.1038/s41598-020-63395-9.
3. Kokkotis C., Moustakidis S., Papageorgiou E., Giakas G. and Tsaopoulos D.E. Machine learning in
knee osteoarthritis: a review, Osteoarthritis and Cartilage Open, 2020, 2 (3), pp. 100069. Available at:
https://doi.org/10.1016/j.ocarto.2020.100069.
4. Guan B., Liu F., Mizaian A.H., Demehri S., Samsonov A., Guermazi A. and Kijowski R. Deep learning
approach to predict pain progression in knee osteoarthritis, Skeletal radiology, 2022, pp. 1-11.
5. Yeoh P.S.Q., Lai K.W., Goh S.L., Hasikin K., Hum Y.C., Tee Y.K. and Dhanalakshmi S. The emergence
of deep learning in knee osteoarthritis diagnosis, Computational intelligence and neuroscience,
2021, pp. 1-20. Available at: https://doi.org/10.1155/2021/4931437.
6. Kim D.H., Lee K.J., Choi D., Lee J.I., Choi H.G. and Lee Y.S. Can additional patient information improve
the diagnostic performance of deep learning for the interpretation of knee osteoarthritis severity, Journal of
Clinical Medicine, 2020, 9 (10), pp. 3341. Available at: https://doi.org/10.3390/jcm9103341.
7. Ahmed S.M. and Mstafa R.J. A comprehensive survey on bone segmentation techniques in knee osteoarthritis
research: from conventional methods to deep learning, Diagnostics, 2022, 12 (3), pp. 611.
Available at: https://doi.org/10.3390/diagnostics12030611
8. Kotti M., Duffell L.D., Faisal A.A. and McGregor A.H. Towards automatically assessing osteoarthritis
severity by regression trees & SVMs, 2013.
9. Kotti M., Duffell L.D., Faisal A.A. and McGregor A.H. Detecting knee osteoarthritis and its discriminating
parameters using random forests, Medical engineering &physics, 2017, 43, pp. 19-29.
Available at: https://doi.org/10.1016/j.medengphy.2017.02.004.
10. Köktaş N.Ş., Yalabik N., Yavuzer G. and Duin R.P. A multi-classifier for grading knee osteoarthritis
using gait analysis // Pattern Recognition Letters, 2010, 31 (9), pp. 898-904. Available at:
https://doi.org/10.1016/j.patrec.2010.01.003.
11. de Dieu Uwisengeyimana J. and Ibrikci T. Diagnosing knee osteoarthritis using artificial neural networks
and deep learning, Biomedical Statistics and Informatics, 2017, 2 (3), pp. 95.
12. Brahim A., Jennane R., Riad R., Janvier T., Khedher L., Toumi H. and Lespessailles E. A decision
support tool for early detection of knee OsteoArthritis using X-ray imaging and machine learning: data
from the osteoarthritis initiative, Computerized Medical Imaging and Graphics, 2019, 73, pp. 11-18.
Available at: https://doi.org/10.1016/j.compmedimag.2019.01.007.
13. Antony J., McGuinness K., Moran K. and O’Connor N.E. 2017. Automatic detection of knee joints and
quantification of knee osteoarthritis severity using convolutional neural networks, In Machine Learning
and Data Mining in Pattern Recognition: 13th International Conference, MLDM 2017, New York,
NY, USA, July 15-20, 2017, Proceedings 13. Springer International Publishing, 2017, pp. 376-390.
Available at: https://doi.org/10.1007/978-3-319-62416-7_27.
14. Pedoia V., Lee J., Norman B., Link T.M. and Majumdar S. Diagnosing osteoarthritis from T2 maps
using deep learning: an analysis of the entire Osteoarthritis Initiative baseline, 2019.
15. Pedoia V., Lee J., Norman B., Link T.M. and Majumdar S. Diagnosing osteoarthritis from T2 maps
using deep learning: an analysis of the entire Osteoarthritis Initiative baseline cohort, Osteoarthritis
and cartilage, 2019, 27 (7), pp. 1002-1010. Available at: https://doi.org/10.1016/j. joca.2019.02.800.
16. Chan S, Dittakan K. Osteoarthritis stages classification to human joint imagery using texture analysis:
A comparative study on ten texture descriptors, Recent Trends in Image Processing and Pattern
Recognition, 2019, pp. 209-225.
17. Antony J, McGuinness K, Moran K, O’Connor N. Automatic detection of knee joints and quantification
of knee osteoarthritis severity using convolutional neural networks, Machine Learning and Data
Mining in Pattern Recognition (MLDM), 2017, pp. 376-390.
18. Tiulpin A., Thevenot J., Rahtu E., Lehenkari P., Saarakkala S. Automatic knee osteoarthritis diagnosis
from plain radiographs: A deep learning-based approach, Sci Rep, 2018, 8, 1727.
19. Pingjun C., Linlin G., Xiaoshuang S., Kyle A., Lin Y. Fully automatic knee osteoarthritis severity grading
using deep neural networks with a novel ordinal loss, Comput Med Imaging Graph, 2019, 75, pp. 84-92.
20. Hu J., Shen L., Sun G. Squeeze-and-excitation networks, 2018 IEEE/CVF Conf on Computer Vision
and Pattern Recognition, 2018, pp. 7132-7141.
21. Tiulpin A., Saarakkala S. Automatic grading of individual knee osteoarthritis features in plain radiographs
using deep convolutional neural networks, Osteoarthr Cartil, 2020, 28 (1): S308. DOI:
10.1016/j.joca.2020.02.480.
22. Norman B., Pedoia V., Noworolski A. Applying densely connected convolutional neural networks for
staging osteoarthritis severity from plain radiographs, J Digit Imaging, 2019, 32, pp. 471-477.
23. Huang G., Liu Z., Weinberger KQ. Densely connected convolutional networks. IEEE Conf on Computer
Vision and Pattern Recognition (CVPR), 2017, 8, pp. 2261-2269.
24. He K., Zhang X., Ren S., Sun J. Deep residual learning for image recognition, IEEE Conf on Computer
Vision and Pattern Recognition (CVPR), 2016, pp. 770-778.
25. Kwon S.B., Han H., Lee M.C., Kim H.C. Machine learningbased automatic classification of knee osteoarthritis
severity using gait data and radiographic images. IEEE Access, 2020, 8, pp. 120597-120603
