GPU-Based Level Set Method for MRI Brain Tumor Segmentation using Modified Probabilistic Clustering

Research Objective

The goals of the research are to

• 1.

  The accurate segmentation of brain tumor with different size and shape in presence of intensity inhomogeneity and noise.

• 2.

  Automatic segmentation of brain tumor with high speed.

Introduction

Accurate image segmentation plays a vital role for anatomic structures inspection, disease identification as well as tracking of its progress in the field of biomedical image processing. It is also used for surgical planning and simulation. Now a day’s Image processing is used to identify the brain tumor. According to the WHO, more than 500,000 people undergo brain tumor treatment in a year. Tumors may appear at different places with distinct intensity. It may differ in size, shape, and location. There are two major types of brain tumor: First one is called benign type which does not spread and second one is known as Malignant which spread from any part of the body to the brain. Since brain are very complicated structure, therefore a precise segmentation is needed for detecting tumors, edema and necrosis tissue. Based on MRI technique, the study of the main cerebral tissues such as white matter, gray matter and CSF is carried out. In such studies, segmentation step is required whose aim is partitioning the intracranial volume into potentially overlapping parts such as WM, GM and CSF. MR images are one of the common ways to visualize brain structure. MRI image segmentation is very helpful for research and clinical study. The MRI scan is more reliable than CT scan for diagnosis because it does not affect the human body as it does not use any kind of radiation. MRI shares a common advantage with CT of high spatial resolution images but without ionizing radiation exposure. It possesses good contrast resolution for different tissues.

Various techniques related to brain tumor inspection are proposed by many researchers in recent past. The domain of automated inspection for brain tumor is found to be populated mostly by level set image Segmentation. The level set technique is a general framework for tracking dynamic interfaces and shapes. It was first developed as a way to model fluid boundaries, such as a flame front. In computer vision and pattern recognition the level set method (LSM) had been widely used for segmentation. The attractive advantage of the LSM is its ability to extract complex contours and to automatically handle topological changes, such as splitting and merging (S. Balla-Arabe, et al. 2013). The LSM belongs to the active contours model (ACMs) which is based on the Eulerian framework, i.e., the geometric representation of the active contour instead of the parametric representation which is based on the Lagrangian framework.