Knee Osteoarthritis detection and classification from balanced and imbalanced x-ray images using deep learning

Knee osteoarthritis (OA) problem is commonly identified in the elderly. Knee osteoarthritis (OA) can be identified in patients by following the traditional processes such as X-rays, CT scan and MRI. The dataset contains nearly 9786 knee x-rays images. Our methodology leverages a comprehensive dataset of knee X-ray images, encompassing both healthy and osteoarthritic cases. The U-Net model is trained on this dataset to learn the intricate patterns and variations associated with OA. This training enables the model to perform precise segmentation of knee joint structures, distinguishing between healthy and osteoarthritic regions. The study will show how well the U-Net-based method works by detecting knee osteoarthritis from X-ray pictures with a high level of sensitivity and specificity. Additionally, the automated segmentation of the affected areas offers useful data for medical practitioners, assisting in early identification and individualized treatment planning. By providing a potent instrument for the early diagnosis of knee osteoarthritis, this research contributes to the field of medical image analysis. Our model first segments the portion by using the UNet model and then with the help of ResNet50 shows the accuracy of 97.45%.