Brain Tumor Detection Using Image Processing Techniques

135

basins of the image I. Watershed transformation is expressed as the comple-

ment of the Xhmax basins set within the I image [56].

4.4

Related Work

Several studies have investigated the segmentation of MRI brain images to

detect and extract tumor areas. A review of selected literature on brain tumor

segmentation techniques and their applications is presented in this section.

In their study, Madhukumar and Santhiyakumari [57] evaluated the ca-

pabilities of Fuzzy C-means and k-means segmentation methods to classify

tissues (gray matter, white matter, cerebro-spinal fluid, necrotic focus, vaso-

genic edema and background) in brain MRI images. In the course of the ex-

periments, Fuzzy C-means classified three tissue classes and generated empty

clusters, whereas k-means classified six classes. k-means demonstrated better

ability to identify vasogenic edema, white matter, gray matter and necrotic

focus than Fuzzy C-means.

Dhage et al. [17] accomplished brain tumor segmentation by using the

Watershed algorithm and determined the position and shape of the tumor in

the MRI image through the use of connected component labeling.

Kaur and Sharma [58] investigated the existing methods for brain tumor

detection and segmentation in brain MRI images and reached the following

findings. Although the intensity-based thresholding techniques yield good re-

sults, they are not effective for images with significant intensity differences.

While region-based segmentation techniques work well for images with high

contrast, they are ineffective for images with low contrast. Edge-based and

clustering-based segmentation techniques obtain better results but fail for

noisy images.

A technique for identifying and localizing brain tumors from MRI scans

was presented by Hazra et al [59]. During the pre-processing stage, filtering

and image enhancement techniques were applied to the image converted to

grayscale. The edge detection stage was performed using Sobel, Prewitt, and

Canny algorithms. In the last stage, thresholding-based segmentation and k-

means clustering techniques were used to detect tumor-affected areas in the

MRI.

Mittal et al. [60] proposed an effective algorithm consisting of pre-

processing, segmentation and output stages to segment tumor from MRI im-

ages. In the first stage, they converted the input image into gray scale, and

applied a high pass filter to remove noises and a median pass filter to enhance

the quality of image. In the second stage, they utilized the Otsu thresholding

method together with the Watershed technique to realize the image segmenta-

tion. In the last stage, they carried out morphological operations to segmented

the image and detected the tumor on the image.