118

Bioinformatics of the Brain

imaging properties as tumors and are displayed darker on T1-weighted scans

and brighter on T2-weighted scans [6, 8].

4.3

Brain Tumor Detection

The overall flow of brain tumor detection consists of three phases. The pre-

processing phase eliminates noise and undesirable artifacts in the MRI scan,

the skull stripping phase eliminates the skull and any surrounding areas from

the brain, the segmentation phase partitions the MRI scan into separate re-

gions and detects the tumor. The sections that follow will provide a full in-

troduction to these phases.

4.3.1

Pre-processing and Enhancement

The primary objective of the image pre-processing step is to prevent erroneous

results that may occur during the segmentation process. The presence of noise

and other undesirable artifacts in the MRI scan can reduce the success of the

segmentation process. To tackle this issue, techniques such as image enhance-

ments and noise reduction methods are implemented to improve the image’s

quality.

4.3.1.1

Histogram Equalization

A histogram is a distribution that shows the frequency of occurrence of each

intensity value in the image. It serves as a visual tool to depict the contrast

(the highest and lowest pixel intensities difference) in an image. A histogram

solely provides statistical information and does not reveal information about

the position of pixels. It is worth noting that multiple images can have iden-

tical histograms. The clustering of intensity values in particular areas in an

image affects the quality of the image. In these cases, histogram equalization

is utilized to improve the image’s quality.

In histogram equalization, the goal is to achieve a homogeneous distri-

bution of the image’s intensity. The definition of histogram equalization in

mathematics can be stated as follows [12]:

sk = T(rk) = (L −1)

k



j=0

pr(rj) = (L −1)

k



j=0

nj

n

where k = 0, 1, · · ·, L −1

(4.1)

In Equation 4.1, rk and sk indicate the input and processed kth pixel

intensity values respectively. Term L is the maximum intensity value (L=2ⁿ

for n bit image), n is the total number of pixels, nj is the frequency of jth

intensity value and pr(rj) is the probability of frequency of intensity rj.