06-11-2012, 05:57 PM
Denoising, Segmentation and Characterization of Brain Tumor from Digital MR Images
1Denoising, Segmentation.pdf (Size: 2.44 MB / Downloads: 62)
Abstract
The objective of this paper is to present an automated segmentation method which allows rapid identification of
Tumor tissues/pathological structure with an accuracy and reproducibility comparable to those of manual
segmentation. The authors uses the wiener filter for the removal of noise and then applies a new marker based
watershed segmentation method using image processing and digital processing algorithms to detect Tumor
tissues of Brain. This method is simple and intuitive in approach and provides higher computational efficiency
along with the exact segmentation of an image. The proposed technique has been implemented on MATLAB 7.3
and the results are compared with the existing techniques.
Introduction
A brain tumor is an abnormal growth of cells within the brain or inside the skull, which can be cancerous or
non-cancerous (Marcel, Elizabeth, NathanMoon, Koen & Guido, 2003). There are more than 120 types of brain
tumors. Today, most medical institutions use the World Health Organization (WHO), classification system to
identify brain tumors. The best way to determine brain tumor is to perform a type of brain scan called a Magnetic
Resonance Imaging (MRI) or a scan called a Computed Tomography (CT) scan. Brain tumors are the leading
cause of solid tumor death in children under age 20 now surpassing acute lymphoblast leukemia, and are the
third leading cause of cancer death in young adults ages 20-39. In the states, the overall incidence of all primary
brain tumors is more than 14 per 100,000 people (Matt, 2007).
Proposed Marker Based Watershed Segmentation Method
According to the definition of geography, a watershed is the ridge that divides areas drained by different river
systems and catchments basins is the geographical area draining into theriver. The basic principle of watershed
technique is to transform the gradient of a grey level image in a topographic surface, where the values of f(x, y)
are interpreted as heights and each local minima embedded in an image is referred as a catchments basins. If we
imagine rain falling on the defined topographical surface, then water would be collected equally in all the
catchments basins. The watershed transformation can be built up by flooding process on a gray tone image and
may be illustrated by Figure 1. The basic watershed algorithm is well recognized as an efficient morphological
segmentation tool which has been used in a variety of gray scale image processes & video processing
applications. However, a major problem with the watershed transformation is that it produces a large number of
segmented regions in the image around each local minima embedded in theimage. Over segmentation problem in
the morphological watershed segmentation for irregular-shaped objects is usually caused by spurious minima in
the inverse distance transform. Figure 2 shows the over segmentation result produced by the basic watershed
segmentation method. In this, different catchment basin is represented by different colors and watershed
ridgelines are shown by white lines. A solution to sort out this problem is to introduce markers and flood the
gradient image starting from these markers instead of regional minima .The Marker based Watershed
Segmentation method possesses several important properties that makes it highly usable for various kinds of
image segmentation problems. Implementation of this method involves various processing steps which can be
arranged in a meaningful manner that is shown in block diagram 1.
Implementation
The proposed algorithm is tested on Personal computer (i3 processor, 2.1GHZ,320GB HDD, 2GB RAM) using
the software Matlab 7.3 with four image slices of thickness 0.5cm in JPEG Format. These test images are
executed in less than 13 seconds including pre-processing time on a personal computer. It also removes the
problem of over segmentation in the process of watershed of negative distant transform.
Results
The final segmentation result obtained by proposed Marker based Watershed algorithm is shown in Figure 3(a)
and Figure 3(b) shows the ROI of a complete tumor. Similarly, the second, third and fourth image Segmentation
is shown in Figure 4, Figure 5, Figure 6.The results of the proposed algorithm for four test images are
summarized in Table 1. Deviation between experimental data and data produced by the proposed method is also
presented in Table 1.The values of K (height threshold) are optimized for proper segmentation of different
images representing different tumor size and locations. And optimized values of K for four test images are shown
in Table 1. The execution time of the algorithm for all test images are also tabulated in Table 1.
Conclusions
The developed algorithm is used to know about the location and size of the tumor. This method uses the user
defined input parameters such as threshold for the analytical calculations. The optimal value of the threshold is
highly dependent on shape and location of the tumor as well as on the different views of the images. In this
method four images, each from different data sets are processed using the marker controlled algorithm. The
result shown in Table 1 which illustrates that the developed algorithm is 8-10 times faster than current state of art
in medical image segmentation. Each test image is executed in less than 15 seconds including the pre-processing
time on a personal computer. The execution time for all of test images is summarized in Table 1. The marker size
can be controlled interactively by the user itself.