Analysis of Cervical X-Ray Images for Vertebra Localization and Segmentation

Abstract

This study is related to Image segmentation and region selection which plays an important role in the fields of medical imaging, object recognition, and computer vision. We are specifically interested in Vertebra segmentation and localization from X-ray images. A reliable identification and Segmentation of vertebras are necessary because of neurological and oncological applications. In the context of robotic surgery, correct knowledge extraction is necessary about the shape and Individual positions of vertebras. Although different vertebras show different characteristics, neighboring vertebras are typically very similar so the task of automatic identification and segmentation is difficult. In these days, medical image processing has become a necessary step in diagnosing and identifying problems that are diagnosed with the help of X-ray, CT, and MR. Medical image processing facilitates medical professionals in accurately diagnosing the problem and proposing its treatment. As medical imaging offers additional information about the patient, thus it becomes more important in medical treatment. For automated vertebra assessment system, it is essential to segment and extract vertebras from the x-ray. The focus of this study is on semi-automated vertebra identification and segmentation. X-ray images are high noise and poor contrast images so this is a challenging task. In this method, we have calculated a mean model by using vertebras of different shapes and angles this mean model is required for identification and segmentation. The selection of an area of interest can be automatic by using a mask or manual during the process. The localization process use Generalized Hough transform is to identify the template image in the X-ray, specifically, it identifies the most likely area of the template in form of points. Further processing is required for identification and segmentation of multiple vertebras. These points are then clustered into the desired number of clusters, we are using fuzzy c mean clustering for this step. Fuzzy c mean will give us centroid and then we will calculate intervertebral points. To make separate regions affine transform is used on these points. The dataset NHANES II is used during experimentations in this study is real world standardizes dataset.