Automatic Spine Extraction in Posteroanterior Radiographs Based on Spine's Features

摘要

Nowadays, radiographs are used more widely in clinical diagnostics. Therefore, automatic detection and extraction of spine in radiographic image are problems which are paid more attention to quantify the severity of spinal diseases. This will help reduce the number of clinical evaluation by a doctor and will allow studies based on large database. The purpose of this paper is to present a new approach to automatic detection of the spine in posteroanterio radiographic image. First, we define a region of interest (ROI) based on a set of radiographs. This region contains sufficient information and data of the vertebra. Filters with threshold based on the brightness and the features of images and spine, are used to remove redundant information. Then a mask is brought in to save the position of the spine. Based on the determined position of the spine, a method is applied to smooth the edge and to add parts of the spinal disability. In addition, the mask processing will detect the direction of curvature of the spine to facilitate other processes. The result which is obtained from the steps above for good quality X-ray image is almost entirely derived spine and less noise, although the poor quality images still have high noise surrounding the bone.