Deep Learning-Based Knee Osteoarthritis Severity
Assessment Using X-ray Imaging
Kothoju Navyeesh -21BCE5655, Kanakala Hinduja -21BCE1116, Nagavarapu Divya Charitha -21BCE5299
Dr. Jani Anbarasi L
School of Computer Science and Engineering
Introduction Methodology Results
Algorithm Accuracy
• Knee OA is a widespread, degenerative joint disease-causing • Proposed System: Develop an end-to-end deep learning
chronic pain, stiffness, and limited mobility, primarily affecting pipeline to automatically classify the severity of knee Inception-V3 94.57 %
people over 55. Due to lack of awareness, many ignore early osteoarthritis from X-ray images using deep convolutional
symptoms until the condition worsens. With the global rise in techniques DenseNet, MobileNet, and InceptionV3 for optimal MobileNet 94.75 %
aging and obesity, OA cases are expected to reach 130 million performance.
by 2050. • Dataset Description: This project uses the Kaggle Knee
DenseNet-121 56 %
Osteoarthritis X-ray dataset, which contains a total of 8260 X-
• OA involves the gradual breakdown of knee cartilage, leading ray images labeled according to the Kellgren and Lawrence
to painful bone-on-bone contact. Risk factors include aging, (KL) grading scale (Grade 0 to 4). The dataset is split into 5778
obesity, joint injuries, repetitive use, and genetic predisposition. training images, 826 validation images, and 1656 testing
Conclusion
Women are more commonly affected, especially after age 45. images..
This project successfully demonstrates the application of deep
• Data Preprocessing: Apply Contrast Limited Adaptive
• This study explores the use of deep learning, specifically CNN learning techniques for automatic severity classification of knee
Histogram Equalization (CLAHE) to enhance image contrast,
models like DenseNet, MobileNet, and InceptionV3, to classify osteoarthritis using X-ray images. By leveraging powerful CNN
followed by resizing and normalization. Implement data
OA severity from X-ray images. Each model offers unique architectures such as DenseNet, MobileNet, and InceptionV3, and
augmentation techniques such as rotation, flipping, and
advantage MobileNet’s lightweight design, and InceptionV3’s enhancing model performance through preprocessing techniques
brightness adjustments to improve model generalization and
multi-scale processing collectively aiming to improve like CLAHE and data augmentation, the system achieves robust and
address class imbalance.
diagnostic accuracy and efficiency in medical imaging. accurate predictions. This work highlights the potential of AI-driven
• Model Architectures: Fine-tune state-of-the-art CNN models:
solutions in advancing medical imaging and improving patient
Objective DenseNet, MobileNet, and InceptionV3 using transfer learning,
outcomes
and ensemble them to boost classification accuracy and ensure
• To compare and evaluate the performance of the proposed robust severity prediction of knee osteoarthritis.. Contact Details
model against individual deep learning models, ensuring Mail ID Mobile No
robustness and reliability in predictions
Architecture Diagram
[email protected] 7013672574
• To extract critical features from X-ray images using advanced
architectures such as MobileNet and InceptionV3, thereby
[email protected] 8341415068
enhancing prediction accuracy and model interpretability.
[email protected] 7671809424
• To provide valuable insights into the clinical implications of n
the study findings, supporting better management and treatment References
strategies for knee osteoarthritis.
Chen, Pingjun (2018), “Knee Osteoarthritis Severity Grading Dataset”,
• To develop an automated and accurate system for estimating Mendeley Data, V1, doi: 10.17632/56rmx5bjcr.1
the severity of knee osteoarthritis using X-ray images by
leveraging state-of-the-art CNN models such as DenseNet, Knee Osteoarthritis Severity Detection. (2024, November). Knee
MobileNet, and InceptionV3 with the aim of enhancing early Osteoarthritis Severity Detection 01 Dataset. Roboflow Universe.
diagnosis and improving clinical decision-making in OA https://universe.roboflow.com/knee-osteoarthritis-severity-detection/
management.. knee-osteoarthritis-severity-detection-01.
Scope of the Project
Abdullah, S. S., & Rajasekaran, M. P. (2022). Automatic detection and
This project aims to classify the severity of knee osteoarthritis classification of knee osteoarthritis using deep learning approach. La
from X-ray images using deep learning models: DenseNet, radiologic medica, 127(4), 398-406.
MobileNet, and InceptionV3. It focuses on providing an accurate,
automated diagnostic tool to support early detection and improve Saleem, M., Farid, M. S., Saleem, S., & Khan, M. H. (2020). X-ray
clinical decision-making. image analysis for automated knee osteoarthritis detection. Signal,
Image and Video Processing, 14, 1079-1087.
