Determination of Osteoarthritis Using Histogram of Oriented Gradients and Multiclass SVM

Abstract

Abstract-Knee Osteoarthritis is most ordinary kind of joint inflammation, which often occurs in one or both the knee joints. Osteoarthritis is additionally called as 'wear and tear' process of joint that results in dynamic disintegration of articular cartilage. Cartilage is smooth substantial layer that ensures movement to occur effortlessly. In Osteoarthritis, the cartilage is inclined towards the destruction as it loses elasticity and becomes brittle.Osteoarthritis is regularly investigated from radiographic evaluation after clinical examination. In any case, a visual evaluation made by the restorative physician depends on experience that varies subjectively and is profoundly reliant on their experience. Subsequently, in order to make diagnostic process more systematic and reliable, evolution of imaging based analysis for early recognition of Osteoarthritis is required. The objective of this study is to develop a machine vision approach for investigation of Knee Osteoarthritis using region based and active shape model. The computation involves histogram of oriented gradient (HOG) method. The processed HOG elements are computed using multiclass SVM for evaluating Osteoarthritis based on Kellgren and Lawrence (KL) grading system. The classification rate of 97.96% for Grade-0, 92.85% for Grade-1, 86.20% for Grade-2, 100% for Grade-3 & Grade-4 is obtained. The results are promising and competitive which are validated by the medical experts.

Detection of Osteoarthritis in Knee Radiographic Images using Artificial Neural Network

“…[9][10]. Further, the computational part was modified using histogram of gradients which were classified using multiclass SVM by achieving the enhanced accuracy of 95% [11]. Later, the implementation was carried out to automatically detect the region of interest using different segmentation methods.…”

Analysis of Multi-modal Biometrics System for Gender Classification Using Face, Iris and Fingerprint Images

“…K-Nearest Neighbor classifier: will classify the class label based on measuring the distance between testing and training data. KNN [46,48,49] will classify by suitable K value which in turn finds the nearest neighbor and provides a class label to un-labeled images. Depending on the types of problem, a variety of different distance measures can be implemented.…”

Early Detection of Osteoarthritis based on Cartilage Thickness in Knee X-ray Images

“…S. S. Gornale et al, have used contour based segmentation method for the analysis of OA using Knee x-ray images [11][12][13][14]. The semi-automated approach was utilized for extracting the region of interest.…”

“…The accuracies of 87.92% and 88.88% were obtained respectively [11][12]. Further utilizing the same method for extracting region of interest histogram gradients were computed and classified using multiclass SVM achieving the accuracy of 95% [13]. The implementation was further conceded using different segmentation methods like Prewitt, Sobel, Texture based and Otsu's based methods [14].…”