Automatic Segmentation of Impaired Joint Space Area for Osteoarthritis Knee on X-ray Image using Gabor Filter Based Morphology Process

Abstract

Segmentation is the first step in osteoarthritis classification. Manual selection is time-consuming, tedious, and expensive. The system is designed to help medical doctors to determine the region of interest of visual characteristics found in knee Osteoarthritis (OA). We propose a fully automatic method without human interaction to segment Junction Space Area (JSA) for OA classification on impaired x-ray image. In this proposed system, right and left knee detection is performed using using Contrast-Limited Adaptive Histogram Equalization (CLAHE) and template macthing. The row sum graph and moment methods are used to segment the junction space area of knee. Overall we evaluated 98 kneess of patients. Experimental results demonstrate an accuracy of the system of up to 100% for detection of both left and right knee and for junction space detection an accuracy 84.38% for the right knee and 85.42% for the left. The second experiment using gabor filter with parameter α=8, θ=0, Ψ=[0 Π/2], γ=0,8 and N=8 and row sum graph give an accuracy 92.63% for the right knee and 87.37% for the left. And the average time needs to process is 65.79 second. For obvious reasons we chose the results of the fourth to segment junction area in both right and the left knee.