ROI impact on the characterization of knee osteoarthritis using fractal analysis

This paper presents a preliminary study of the influence of the positioning of Regions Of Interest (ROI) for the characterization of bone texture on radiographs for the diagnosis of knee OsteoArthritis (OA) progression. Characterization of the bone texture is of great interest to doctors because it would improve the prognostic in the clinical routine. In general, studies mainly focus on the descriptors while neglecting the choice of ROI positioning. Using fractal descriptors, the objective of this work is to highlight the impact of the ROI for the diagnosis of knee OA by considering the couple (descriptor, ROI). This study was performed over 1054 knees from 616 subjects composed of stable and progressor patients. Achieved statistical tests demonstrated the importance of the choice of the ROI to improve the clinical diagnosis.