A Survey on KOA-CCTNet: Optimized Compact Convolutional Transformer for Osteoarthritis Severity Detection

Abstract

Knee osteoarthritis(OA) is a prevalent degenerative a condition affecting the joints that significantly impacts the quality of life. Early and accurate diagnosis is critical for effective management and treatment. This project, titled “Automated Knee Osteoarthritis Prediction and Classification utilizes Xray imaging along with deep learning methods to propose an innovative method for diagnosing and grading knee(OA) by applying deep learning models. The system is developed using Python for backend computation, Flask as the web framework, and HTML, CSS, and JavaScript for the front-end. Two stateof-the-art advanced AI models, VGG16 and MobileNetV2, have been employed to classify X-ray images of knee joints into five categories on the Kellgren and Lawrence grading system: Normal, Doubtful, Mild, Moderate, and Severe. The dataset consists of 8bit grayscale X-ray images obtained through recognized hospitals and diagnostic centers using the PROTEC PRS 500E X-ray machine. Each image has been reviewed and labeled manually by two medical experts for accuracy and reliability. The VGG16 and MobileNetV2 models have been thoroughly evaluated. We measured performance using well-established metrics, namely accuracy, precision, and recall and F1-score,Confusion Matrix, have been computed to assess each model comprehensively. The findings of this study highlight the capability through the use of deep learning to facilitate automated KOA detection and grading.

Introduction

Knee osteoarthritis (KOA) is a widespread degenerative joint disorder primarily affecting older adults and individuals with risk factors such as obesity or previous injuries. It leads to progressive cartilage loss, pain, stiffness, and reduced mobility, making it a major global health concern. Diagnosis typically relies on radiographic evaluation using the Kellgren–Lawrence (KL) grading system, but this manual process is subjective, time-consuming, and inconsistent across clinicians.

Advancements in artificial intelligence, especially deep learning, offer promising solutions for automated KOA detection. Convolutional neural networks (CNNs) capture local image features, while transformers model global context, but each has limitations when used alone. To address these gaps, the study proposes an enhanced hybrid approach using a modified Compact Convolutional Transformer (KOA-CCTNet) designed for accurate, automated KOA grading.

A comprehensive literature review highlights the limitations of existing KOA studies and various deep-learning applications in medical imaging, emphasizing issues such as data imbalance, overfitting, lack of external validation, and insufficient model interpretability.

The proposed methodology introduces a robust pipeline:

• Dataset creation from four sources and region-specific cropping produced 10,232 high-quality KOA radiographs.

• DCGAN-based augmentation addressed class imbalance, expanding the dataset to 110,232 images with validated synthetic quality.

• Preprocessing techniques (Adaptive Histogram Equalization and Fast Non-Local Means filtering) enhanced contrast and reduced noise.

• Transfer learning baselines using common CNN models confirmed the need for more advanced architectures.

• Transformer model comparison identified CCT as the most suitable backbone.

• KOA-CCTNet improvements, including reduced encoder depth, optimized tokenization, tuned hyperparameters, and ablation studies, yielded a final accuracy of 94.58%, outperforming all CNN and transformer baselines.

Key contributions include enhanced feature clarity, balanced dataset construction, and improved robustness even with reduced training data. Overall, KOA-CCTNet demonstrates a clinically meaningful, efficient, and accurate framework for automated KOA grading, leveraging both local and global image features for superior diagnostic performance.

Conclusion

The developed KOA-CCTNet framework demonstrated showed enhanced effectiveness in classifying KOA severity grades, achieving 94.58% accuracy and outperforming both CNN and transformer-based baselines. By integrating advanced augmentation, preprocessing, and a modified CCT architecture, the model ensured robustness even with reduced datasets.

References

[1] A. Cui, H. Li, D. Wang, J. Zhong, Y. Chen, and H. Lu, “Global, regional prevalence, incidence and risk factors of knee osteoarthritis in population-based studies,” EClinicalMedicine, vols. 29–30, Art. no. 100587, Dec. 2020, doi: 10.1016/j.eclinm.2020.100587. [2] P. P. F. M. Kuijer, H. F. van der Molen, and S. Visser, “A health-impact assessment of an ergonomic measure to reduce the risk of work-related lower back pain, lumbosacral radicular syndrome and knee osteoarthritis among floor layers in The Netherlands,” Int. J. Environ. Res. Public [3] Health, vol. 20, no. 5, p. 4672, Mar. 2023, doi: 10.3390/ijerph20054672. [4] V. K. V., V. Kalpana, and G. H. Kumar, “Evaluating the efficacy of deep learning models for knee osteoarthritis prediction based on KellgrenLawrence grading system,” e-Prime - Adv. Electr. Eng., Electron. Energy, vol. 5, Art. no. 100266, Sep. 2023, doi: 10.1016/j.prime.2023.100266. [5] S. Castagno, M. Birch, M. van der Schaar, and A. McCaskie, “Prediction of the Rapid Progression of Knee Osteoarthritis Using Automated Machine Learning: A Novel Precision Health Approach for Chronic Degenerative Diseases,” SSRN, 2024. [Online]. Available: https://ssrn.com/abstract=4561796. Accessed: Aug. 2, 2024. [6] A. A. S. Afroze, R. Tamilselvi, and M. G. P. Beham, “Machine learning based osteoarthritis detection methods in different imaging modalities: A review,” Current Med. Imag., vol. 19, no. 14, pp. 1628–1642, Dec. 2023, doi: 10.2174/1573405619666230130143020. [7] Y. C. Park, K. J. Park, B. H. Goo, J. H. Kim, B. K. Seo, and Y. H. [8] Baek, “Oriental medicine as collaborating treatments with conventional treatments for knee osteoarthritis: A PRISMA-compliant systematic review and meta-analysis,” Medicine, vol. 102, no. 29, Art. no. E34212, 2023, doi: 10.1097/MD.0000000000034212. [9] S. V. Chaugule, S. V. Chaugule, and V. S. Malemath, “Towards identifying key features in the classification of knee osteoarthritis—An enhanced feature fusion based deep network model,” Tech. Rep., 2023. [10] Y. Soda, T. Kano, and M. Nakamura, “Kinematically aligned total knee arthroplasty for valgus osteoarthritis of more than 10?: Is it still a challenging surgery?” Open J. Orthopedics, vol. 13, no. 9, pp. 355– 369, 2023, doi: 10.4236/ojo.2023.139035. [11] H. Khalid, M. Hussain, M. A. Al Ghamdi, T. Khalid, K. Khalid, M. A. Khan, K. Fatima, K. Masood, S. H. Almotiri, M. S. Farooq, and A. Ahmed, “A comparative systematic literature review on knee bone reports from MRI, X-rays and CT scans using deep learning and machine learning methodologies,” Diagnostics, vol. 10, no. 8, p. 518, Jul. 2020, doi: 10.3390/diagnostics10080518. [12] J. J. Bjerre-Bastos, M. A. Karsdal, M. Boesen, H. Bliddal, A. BayJensen, J. R. Andersen, and A. R. Bihlet, “The acute and long-term impact of physical activity on biochemical markers and MRI measures in osteoarthritis—Perspectives for clinical osteoarthritis research,”Transl. Sports Med., vol. 3, no. 5, pp. 384–394, Sep. 2020, doi:10.1002/tsm2.155. [13] Y. X. Teoh, A. Othmani, S. L. Goh, J. Usman, and K. W. Lai, “Deciphering knee osteoarthritis diagnostic features with explainable artificial intelligence: A systematic review,” arXiv, Aug. 2023, arXiv:2308.09380. [14] Y. Nasser, M. El Hassouni, D. Hans, and R. Jennane, “A discriminative shape-texture convolutional neural network for early diagnosis of knee osteoarthritis from X-ray images,” Phys. Eng. Sci. Med., vol. 46, no. 2, pp. 827–837, Jun. 2023, doi: 10.1007/s13246-023-01256-1. [15] S. Paul and P.-Y. Chen, “Vision transformers are robust learners,” in Proc. AAAI Conf. Artif. Intell., vol. 36, no. 2, pp. 2071–2081, 2022, doi: 10.1609/aaai.v36i2.20103. [16] K. Mohiuddin, M. A. Alam, M. M. Alam, P. Welke, M. Martin, J. Lehmann, and S. Vahdati, “Retention is all you need,” in Proc. 32nd ACM Int. Conf. Inf. Knowl. Manage., pp. 4752–4758, Oct. 2023, doi: 10.1145/3583780.3615497. [17] A. Hassani, S. Walton, N. Shah, A. Abuduweili, J. Li, and H. Shi, “Escaping the big data paradigm with compact transformers,” arXiv, Apr. 2021, arXiv:2104.05704. [18] G. K. M. and A. D. Goswami, “Automatic classification of the severity of knee osteoarthritis using enhanced image sharpening and CNN,” Appl. Sci., vol. 13, no. 3, p. 1658, Jan. 2023, doi: 10.3390/app13031658. [19] A. S. Mohammed, A. A. Hasanaath, G. Latif, and A. Bashar, “Knee osteoarthritis detection and severity classification using residual neural networks on preprocessed X-ray images,” Diagnostics, vol. 13, no. 8, p. 1380, Apr. 2023, doi: 10.3390/diagnostics13081380. [20] S. S. Abdullah and M. P. Rajasekaran, “Automatic detection and classification of knee osteoarthritis using deep learning approach,” La Radiologia Medica, vol. 127, no. 4, pp. 398–406, Apr. 2022, doi: 10.1007/s11547-022-01476-7. [21] S. A. El-Ghany, M. Elmogy, and A. A. A. El-Aziz, “A fully automatic fine-tuned deep learning model for knee osteoarthritis detection and progression analysis,” Egyptian Informat. J., vol. 24, no. 2, pp. 229– 240, Jul. 2023, doi: 10.1016/j.eij.2023.03.005. [22] T. Tariq, Z. Suhail, and Z. Nawaz, “Knee osteoarthritis detection and classification using X-rays,” IEEE Access, vol. 11, pp. 48292–48303, 2023, doi: 10.1109/ACCESS.2023.3276810. [23] S. Olsson, E. Akbarian, A. Lind, A. S. Razavian, and M. Gordon, “Automating classification of osteoarthritis according to Kellgren-Lawrence in the knee using deep learning in an unfiltered adult population,” BMC Musculoskeletal Disorders, vol. 22, no. 1, pp. 1–8, Dec. 2021, doi: 10.1186/s12891-021-04722-7. [24] A. Qadir, R. Mahum, and S. Aladhadh, “A robust approach for detection and classification of KOA based on BILSTM network,” Comput. Syst. Sci. Eng., vol. 47, no. 2, pp. 1365–1384, 2023, doi: 10.32604/csse.2023.037033. [25] U. Yunus, J. Amin, M. Sharif, M. Yasmin, S. Kadry, and S. Krishnamoorthy, “Recognition of knee osteoarthritis (KOA) using YOLOv2 and classification based on convolutional neural network,” Life, vol. 12, no. 8, p. 1126, 2022. [26] P. Chen, “Knee osteoarthritis severity grading dataset,” Mendeley, Sep. 2018, doi: 10.17632/56rmx5bjcr.1. [27] S. Gornale and P. Patravali, “Digital knee X-ray images,” Mendeley, Jun. 2020, doi: 10.17632/t9ndx37v5h.1. [28] Kaggle, “CGMH Osteoarthritis Images,” Dataset, Accessed: Jul. 31, 2020. [Online]. Available: https://www.kaggle.com/datasets/tommyngx/cgmh-oa [29] M. Gordon and E. Akbarian, “Knee osteoarthritis classification according to Kellgren-Lawrence,” Tech. Rep., 2021, doi:10.23698/aida/koa2021. [30] S. S. Gornale, P. U. Patravali, and R. R. Manza, “Detection of osteoarthritis using knee X-ray image analyses: A machine vision based approach,” Int. J. Comput. Appl., vol. 145, no. 1, pp. 20–26, Jul. 2016, doi: 10.5120/ijca2016910544. [31] W. Fang, F. Zhang, V. S. Sheng, and Y. Ding, “A method for improving CNN-based image recognition using DCGAN,” Comput., Mater. Continua, vol. 57, no. 1, pp. 167–178, 2018, doi: 10.32604/cmc.2018.02356. [32] M. Puttagunta and R. Subban, “A novel COVID-19 detection model based on DCGAN and deep transfer learning,” Proc. Comput. Sci., vol.204, pp. 65–72, Jan. 2022, doi: 10.1016/j.procs.2022.08.008. [33] D. J. Bora, A. K. Gupta, and F. A. Khan, “Comparing the performance of LAB? and HSV color spaces with respect to color image segmentation,” arXiv, Jun. 2015, arXiv:1506.01472. [34] A. V. Reddy, R. Thiruvengatanadhan, M. Srinivas, and P. Dhanalakshmi, “Region based segmentation with enhanced adaptive histogram equalization model with definite feature set for sugarcane leaf disease classification,” Int. J. Intell. Syst. Appl. Eng., vol. 12, no. 1s, pp. 428– 442, 2024. [35] P. Yugander, K. Abhishek, P. S. Reddy, G. Manideep, T. Sahithi, and M. Jagannath, “Extraction of blood vessels from retinal fundus images using maximum principal curvatures and adaptive histogram equalization,” in Proc. 1st Int. Conf. Electr. Electron. Inf. Commun. Technol. (ICEEICT), Feb. 2022, doi: 10.1109/ICEEICT53079.2022.9768517. [36] Y. Mousania, S. Karimi, and A. Farmani, “Optical remote sensing, brightness preserving and contrast enhancement of medical images using histogram equalization with minimum cross-entropy-Otsu algorithm,” Opt. Quantum Electron., vol. 55, no. 2, pp. 1–22, Feb. 2023, doi: 10.1007/s11082-022-04341-z. [37] A. Buades, B. Coll, and J. M. Morel, “A non-local algorithm for image denoising,” in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., Jun. 2005, pp. 60–65. [38] A. Qi, D. Zhao, F. Yu, A. A. Heidari, Z. Wu, Z. Cai, F. Alenezi, R. F. Mansour, H. Chen, and M. Chen, “Directional mutation and crossover boosted ant colony optimization with application to COVID-19 X-ray image segmentation,” Comput. Biol. Med., vol. 148, Art. no. 105810, Sep. 2022, doi: 10.1016/j.compbiomed.2022.105810. [39] S. Montaha, S. Azam, A. K. M. R. H. Rafid, P. Ghosh, M. Z. Hasan, M. Jonkman, and F. De Boer, “BreastNet18: A high accuracy fine-tuned VGG16 model evaluated using ablation study for diagnosing breast cancer from enhanced mammography images,” Biology, vol. 10, no. 12, p. 1347, Dec. 2021. [40] K. Fatema, S. Montaha, M. A. H. Rony, S. Azam, M. Z. Hasan, and M. Jonkman, “A robust framework combining image processing and deep learning hybrid model to classify cardiovascular diseases using a limited number of paper-based complex ECG images,” Biomedicines, vol. 10, no. 11, p. 2835, Nov. 2022, doi: 10.3390/biomedicines10112835. [41] H. A. Sanghvi, R. H. Patel, A. Agarwal, S. Gupta, V. Sawhney, and A. S. Pandya, “A deep learning approach for classification of COVID and pneumonia using DenseNet-201,” Int. J. Imag. Syst. Technol., vol. 33, no. 1, pp. 18–38, Jan. 2023, doi: 10.1002/ima.22812. [42] S. Montaha, S. Azam, A. K. M. R. H. Rafid, M. Z. Hasan, A. Karim, K. M. Hasib, S. K. Patel, M. Jonkman, and Z. I. Mannan, “MNet-10: A robust shallow convolutional neural network model performing ablation study on medical images assessing the effectiveness of applying optimal data augmentation technique,” Frontiers Med., vol. 9, Art. no. 924979, Aug. 2022. [43] F. J. M. Shamrat, S. Azam, A. Karim, K. Ahmed, F. M. Bui, and F. De Boer, “High-precision multiclass classification of lung disease through customized MobileNetV2 from chest X-ray images,” [44] Comput. Biol. Med., vol. 155, Art. no. 106646, Mar. 2023, doi: 10.1016/j.compbiomed.2023.106646. [45] R. Indraswari, R. Rokhana, and W. Herulambang, “Melanoma image classification based on MobileNetV2 network,” Proc. Comput. Sci., vol. 197, pp. 198–207, Jan. 2022, doi: 10.1016/j.procs.2021.12.132. [46] Y. E. Almalki, M. Zaffar, M. Irfan, M. A. Abbas, M. Khalid, K. S. Quraishi, T. Ali, F. Alshehri, S. K. Alduraibi, A. A. Asiri, M. A. A. Basha, A. Alduraibi, M. K. Saeed, and S. Rahman, “A novel-based Swin transfer based diagnosis of COVID-19 patients,” Intell. Autom. Soft Comput., vol. 35, no. 1, pp. 163–180, 2023, doi: 10.32604/iasc.2023.025580. [47] J. Feng, H. Zhang, M. Geng, H. Chen, K. Jia, Z. Sun, Z. Li, X. Cao, and B. W. Pogue, “X-ray cherenkov-luminescence tomography reconstruction with a three-component deep learning algorithm: Swin transformer, convolutional neural network, and locality module,” J. Biomed. Opt., vol. 28, no. 2, pp. 1–18, Feb. 2023, doi: 10.1117/1.jbo.28.2.026004. [48] A. Marefat, M. Marefat, J. H. Joloudari, M. A. Nematollahi, and R. Lashgari, “CCTCOVID: COVID-19 detection from chest X-ray images using compact convolutional transformers,” Frontiers Public Health, vol. 11, Art. no. 1025746, Feb. 2023, doi: 10.3389/fpubh.2023.1025746. [49] I. U. Khan, S. Azam, S. Montaha, A. A. Mahmud, A. K. M. R. H. Rafid, M. Z. Hasan, and M. Jonkman, “An effective approach to address processing time and computational complexity employing modified CCT for lung disease classification,” Intell. Syst. with Appl., vol. 16, Art. no. [50] 200147, Nov. 2022, doi: 10.1016/j.iswa.2022.200147. [51] C.-M. Lo and K.-L. Lai, “Deep learning-based assessment of knee septic arthritis using transformer features in sonographic modalities,” Comput. Methods Programs Biomed., vol. 237, Art. no. 107575, Jul. 2023. [52] O. Uparkar, J. Bharti, R. K. Pateriya, R. K. Gupta, and A. Sharma, “Vision transformer outperforms deep convolutional neural networkbased model in classifying X-ray images,” Proc. Comput. Sci., vol. 218, pp. 2338–2349, Jan. 2023, doi: 10.1016/j.procs.2023.01.209. [53] M. M. Al Rahhal, Y. Bazi, R. M. Jomaa, A. AlShibli, N. Alajlan, M. L. Mekhalfi, and F. Melgani, “COVID-19 detection in CT/X-ray imagery using vision transformers,” J. Personalized Med., vol. 12, no. 2, p. 310, Feb. 2022, doi: 10.3390/jpm12020310. [54] J. J. Mondal, M. F. Islam, S. Zabeen, and M. A. Manab, “InvoPotNet: Detecting pothole from images through leveraging lightweight involutional neural network,” in Proc. 25th Int. Conf. Comput. Inf. Technol. (ICCIT), Cox’s Bazar, Bangladesh, pp. 599–604, 2022, doi: 10.1109/ICCIT57492.2022.10055818. [55] W. Ou and S.-I. Kamata, “Skin lesion classification based on involution neural networks with Triplet++ attention generator,” in Proc. 8th Int. Conf. Biomed. Signal Image Process., Jul. 2023, pp. 1–6. [56] I. U. Khan, M. A. K. Raiaan, K. Fatema, S. Azam, R. U. Rashid, S. H. Mukta, M. Jonkman, and F. De Boer, “A computer-aided diagnostic system to identify diabetic retinopathy, utilizing a modified compact convolutional transformer and low-resolution images to reduce computation time,” Biomedicines, vol. 11, no. 6, p. 1566, May 2023, doi: 10.3390/biomedicines11061566. [57] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 30, pp. 1–11, 2017.

Copyright

Copyright © 2025 Dr. Madhavi Dasari, Adithya C M, Akshay R, Chandan R , Koushik Kumar R. 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.