Image segmentation: Determination of joint space area inhand radiographs

This paper is an attempt to investigate the possibility of automating assessment of joint damage in hand radiographs. The goal is to design a segmentation algorithm to obtain the area which is an object, the joint space. Image segmentation is the process of partitioning a digital image into several segments. For this study has collected 46 hands radiographic images. Pre-processing done using adaptive threshold and morphological gradient concept, then do segmentation with watershed transformation. Analysis results of segmentation based on the level of accuracy, sensitivity and specificity using the ROC. Results of segmentation with pre-processing, produces images that are not over-segmentation and produces a clearer image with accuracy value of 95.75%, sensitivity of 70.51% and specificity of 96.99% in a high percentage.