Knee Osteoarthritis Radiology Assistant

Abstract

Knee osteoarthritis is a degenerative joint disease affecting millions worldwide. Early detection and monitoring of knee osteoarthritis are critical for effective management of the disease. Magnetic resonance imaging (MRI) is a powerful tool for detecting knee osteoarthritis, but the interpretation of MRI images can be prolonged and subjective. In this paper, we propose a deep learning-based approach for the automatic diagnosis of knee osteoarthritis from MRI images. Our approach involves a deep learning architecture known as Dense Net, which has shown promising results in image classification tasks. We also incorporate a Squeeze-and- Excitation (SE) layer into the network, which can selectively emphasize informative features in the input images. We train and validate our approach using the OAI dataset of MRI images from patients with knee osteoarthritis and healthy controls. The images will then be fed into the Dense Net with SE layers to automatically classify them as either healthy or osteoarthritis. The python script generates a sample report for the MRI scans uploaded in the portal. The patient should be able to access their report but should also be able to generate a report if they have their MRI scans. The proposed model with Dense Net architecture gives an accuracy of 88.5%. The performance of our approach will be evaluated using various metrics such as accuracy, precision, recall, F1 score, confusion matrix, and area under the receiver operating characteristic curve (AUC-ROC).