Osteoarthritis Severity Estimator
Chede Sumith Kumar
Dept. of Computer Science Engineering Vasavi College of Engineering Hyderabad, India sumithkumar070104@gmail.com
Seelam Dinesh Reddy
Dept. of Computer Science Engineering Vasavi College of Engineering Hyderabad, India dineshreddyseelam726@gmail.com
D. Lakshmi Prasanna
Dept. of Computer Science Engineering Vasavi College of Engineering Hyderabad, India duvva.prasanna5@gmail.com
Abstract—Knee osteoarthritis (OA) is a degenerative joint disorder affecting millions worldwide, leading to pain, reduced mobility, and diminished quality of life. Early detection and accurate assessment of OA severity are crucial for timely medical intervention. This project presents a deep learning- based approach to predict the severity of knee osteoarthritis using X-ray images. Two distinct datasets were utilized—one comprising scanned clinical X-rays and the other consisting of images captured via mobile phones or cameras to simulate real- world user inputs.
To handle these varying image qualities and formats, we developed two specialized ensemble models. The first ensemble, designed for scanned X-rays, integrates ResNet50 and Xception architectures. The second ensemble, targeting user-captured images, combines EfficientNet and DenseNet models. These ensembles enhance classification performance through complementary feature extraction capabilities.
The proposed solution demonstrates the potential of AI in medical diagnostics, offering a scalable and accessible tool for both clinical and remote settings to support healthcare professionals in evaluating knee OA severity.
Keywords— Pose Estimation, Real-Time Feedback, Exercise Monitoring, Computer Vision, Human Activity Recognition.
