© 2017, IJARCSM S All Rights Reserved 125 | P a ge
IS SN: 2321- 7782 (Online)
e-ISJN: A4372- 3114
Impact Factor: 6.047
Volume 5, Issue 6, June 2017
International Journal of Advance Research in
Computer Science and Management Studies
Research Article / Survey Paper / Case Study
Available online at: www.ijarcsms.com
Boundary Refinement Technique Using Laplace Transform for
CT Spine Image Segmentation
Aditi B. Pardeshi
1
Department of Computer Engineering
Late G. N Sapkal College of Engineering,
Anjaneri, Nasik – India
Prof. J. V. Shinde
2
Department of Computer Engineering
Late G. N Sapkal College of Engineering,
Anjaneri, Nasik – India
Abstract: Image segmentation is designated as partitioning an image into a predetermined number of non-overlapping
regions. In medical applications, it is a primary processing most systems that support medical diagnosis, surgical planning
and treatments. In general, this procedure is done manually by clinicians, which may be monotonous and time-consuming.
To overcome the problem, a number of interactive segmentation methods have been proposed in the literature. But each
method has some drawbacks, so we proposed a method which uses prior knowledge of anatomy. In this paper, we propose
method to perform segmentation for CT images thoracic and lumbar vertebrae by applying Multiatlas-based segmentation
with Laplace Transform. A total 12 CT spine images which consist of thoracic and lumbar vertebrae are used for evaluation.
The proposed method is fully automatic as well as efficient. Evaluation of method gives average dice coefficient of 0.90.
Keywords: CT, Image segmentation, Label Fusion, Laplace Transform, Multiatlas, Pre-processing, spine, vertebrae.
I. INTRODUCTION
This Truthful analysis of the spine and spinal structures from medical images is an essential tool in many medical
applications of spinal imaging. Study of the detailed shape of individual vertebrae can significantly assist early diagnosis,
surgical planning and follow-up opinion of several spinal pathologies, such as spinal deformities, degenerative disorders,
tumours and trauma. Also segmentation of individual vertebrae by computer-assisted methods may provide additional support
to identification and handling of vertebral fractures. Robust and efficient segmentation algorithms on medical images are a
challenging research topic of increasing interest especially from last few years. In case of medical applications, manually
segmenting a vertebra is time consuming and tiresome. Therefore semi-automated or fully-automated methods are required for
most medical applications which increase the correctness, stability, and reproducibility of the analysis, so that it will allow
clinicians to focus more on their other work. The output taken from image segmentation process is the principal parameter for
the quality of advance image processing process. Image segmentation algorithms play a fundamental role in medical
applications, i.e., diagnosis of diseases related to heart, spine, pelvis, prostate, brain, knee and blood vessel. Therefore, Image
segmentation is still a very interesting topic of research for the field of image processing [1].
Atlas is well-defined as the grouping of an intensity image (template) and its segmented image (the atlas labels). After
registering the atlas template & the target image, the atlas labels are spread to the target image. This process is termed as atlas-
based segmentation. In current years, researchers have investigated different registration techniques to align atlases to query
subjects and also strategies for atlas construction. The segmentation of the vertebrae is not easy, mainly due to the complex
shape, variable architecture of vertebrae across the population and neighbouring anatomy of similar intensity (e.g. other
vertebrae, other bones and/or other tissues) [2].A number of segmentation algorithms for computed tomography (CT) images of
spine have been proposed. Segmentation of vertebrae was accomplished by unsubstantiated image processing approaches such