Predicting Severity Of Osteoarthritis Using Recurrent Convolutional Neural Networks (Rcnn) And Medical Imaging Data

Abstract

Cartilage degeneration causing functional incapacity and discomfort defines common joint condition Osteoarthritis (OA). Correct classification of OA severity from knee X-ray images determines diagnosis and treatment course. Conventional methods of OA severity classification rely on hand inspection or basic machine learning techniques, which could not effectively capture the complex trends in imaging data. Using a dataset of 9,786 knee X-ray images, this work uses Recurrent Convolutional Neural Networks (RCNN) to project OA severity. Combining recurrent layers in the RCNN architecture helps to capture temporal correlations in spatial information, hence enhancing classification performance. Defining the dataset are five KL grades: 0 (healthy) to 4 (severe). This beats by around 5%, 7%, and 3%, respectively standard CNN, Deep Neural Network (DNN), and Deep Convolutional Neural Network (DCNN), with a precision of 87.2%, recall of 88.9%, accuracy of 89.5%, and F1-score of 88.0%.