Diagnosing and grading knee osteoarthritis from X-ray images using deep neural angular extreme learning machine

Knee osteoarthritis is becoming more and more common, as a result there is a need for automated diagnostic system, that triggers an essential necessity in medical image analysis and greatly impacts health rate due to spreading of Knee Osteoarthritis. However, the current knee imaging modalities processes cause more disruption in the medical detection framework because of the high time-consumption and intricacies in the workload. It is a highly demanding endeavour for Knee Osteoarthritis segmentation due to its irregular appearance, distinct densities, osteophytes formation, and joint space narrowing. Presently, Deep learning-based techniques have been applied to identify the knee joint regions and assist radiologists in the initial diagnosis and scrutinizing the severity level. The proposed work presents a modified deep angular extreme learning machine model based deep network architecture with Center of Gravity and Angular Momentum-based segmentation model for obtaining optimal regions of interest optimization algorithm, which increases the efficiency and performance of the Knee Osteoarthritis diagnostic model.Two publicly accessible benchmark datasets have been engaged with the pre-processing task accomplished by image normalization and data augmentation techniques. The Deep Neural angular extreme learning machine comprises of one input layer, two hidden layers and one output layer. In the input layer, raw X-ray images are obtained from two datasets, namely, Knee Osteoarthritis Dataset with Severity Grading and OAI (Osteo arthritis Initiative). Moreover, Center of Gravity and Angular Momentum-based segmentation model for obtaining optimal regions of interest optimization algorithm, has analysed the optimal set of gathered features. In the experimental evaluation, the proposed model outperformed than CNN Random Forest Kneighbors (CRK), Dense Net 169, by 5.49%, 11.12% respectively in terms of accuracy for Knee Osteoarthritis Dataset with Severity Grading. The developed Deep Neural extreme learning machine model attained better accuracy than CNN Random Forest Kneighbors (CRK), Dense Net 169 for the recall rate of Grade 0, Grade 1, Grade 2, Grade 3, Grade 4 knee osteo arthritis by 4.48%, 6.68% respectively for OAI dataset. The obtained precision of the developed model is 2.79%, 6.28%, 5.00%, 10.00%, maximized than CNN Random Forest Kneighbors (CRK), Dense Net 169 respectively, for the segmentation of the knee joint in the Knee Osteoarthritis Dataset with Severity Grading and OAI dataset. The investigational analysis validates the proposed Deep Neural extreme learning machine model's effectiveness and performance in obtaining high accuracy in comparison with other state-of-the-art techniques.