Segmentation of Lumbar [L1-L4] AP Spine X-ray images using various Level Set methods to detect Osteoporosis

The automatic segmentation of lumbar spine x-ray images is a crucial step involved with the minimally invasive surgery of the spine. In most of the advanced methods, human intrusion is still essential because of Lumbar spine x-ray imaging has limitations, such as irregular boundaries, image noise and topological variation. Therefore, the paper employs various segmentation methods such as a multiphase statistical and variational level set (MSVLS) approach is presented for simultaneous bias correction and tissue segmentation in x-ray images. Secondly, the new evolution of a level set following distance regularized level set evolution (DRLSE). Due to the distance regularization effect, reinitialization is not necessary, as well as any numerical errors it may cause. Finally, Lumbar spine segmentation is carried out with the help of canny edge filters and morphological operations on X-ray images. Unfortunately, despite the filter's ability to detect lumbar spine edges, the final edges they produce are still unsatisfactory. In order to solve this problem, we use Euler numbers to extract the lumbar spine region, before implementing edge detection with a filter. Most of the segmented image matches the ground truth very closely, which is a convincing conclusion from the implementation. In the experiment we used AP lumbar spine datasets and compared with other state-of-the-art methods. Accurate segmentation helps to detect Osteoporosis. (Abstract)