Multiregional Fuzzy Thresholding Segmentation Completed by Spatial Median Aggregation: Modeling and Segmentation of Early Pathological Findings of Articular Cartilage

In the field of the clinical orthopedics, articular cartilage is one of the essential object which is evaluated. From the clinical point of view, the early osteoarthritis changes are increasingly challenging especially due to their insufficient contras against image background. Therefore, those findings are often only subjectively estimated from MR records. We have proposed a mathematical model based on the multiregional thresholding methodology, which is able to differentiate of physiological articular cartilage from osteoarthritic findings. Segmentation method is composed from two parts: brightness modeling via fuzzy triangular functions, and spatial median aggregation procedure taking into account the spatial pixel information which makes the model robust against noise and artifact which are often incorrectly classified. A method was tested on the sample of MR data from Proton Dense sequence and Fat Suppression sequence with satisfactory results.