An Ensemble Deep Learning Approach for Automated Knee Osteoarthritis Detection using X-Ray Images

Hema padmini Ganji
Dept of ECE-embedded systems,Sidhardha educational academy group of institutions, C.Gollapalli, Tirupati.
Affliated to JNTUA
Ananthapur,India.
ganjipadmini@gmail.com

Abstract— Knee osteoarthritis is a common type of joint problem where the cartilage in the knee slowly wears away over time. It often causes long-term pain, stiffness, and reduced movement, especially in older people. Finding KOA early and correctly assessing its severity is important for better treatment and to stop more joint damage from happening. Conventional diagnosis is mainly conducted via visual assessment of knee X-ray images utilizing the Kellgren-Lawrence (KL) grading system. Reading X-ray pictures by hand takes a lot of time and can vary depending on the doctor's skill and personal opinions. To address these limitations, this study proposes an automated deep learning-based framework for the detection and classification of knee osteoarthritis from radiographic images. The suggested method uses transfer learning with two strong convolutional neural network models, InceptionV3 and NASNetLarge, to get detailed features from knee X-ray images. Before starting the training of the model, certain image processing steps like resizing, normalizing, and augmenting the images are done. These steps help in making the data better quality and allow the model to perform well on different types of data. The system sorts knee images into various levels of seriousness according to the KL grading scale. Additionally, a group approach is used to bring togetherthe predictions from both models, which helps make the classification more dependable and boosts the overall results.