122

Bioinformatics of the Brain

skull is the outcome of the pixel subtraction processes. Dandıl [9] proposed a

skull stripping algorithm that includes morphological operations like erosion,

dilation and filling. In addition, the study conducted a comparative analysis

with Otsu’s method to showcase the efficacy of the suggested skull-stripping

method. Roy and Maji [29] presented a novel intensity-based algorithm for

skull stripping. The steps of the algorithm include adaptively calculating the

threshold value and using morphological operations such as opening and clos-

ing. The authors compared the results of the algorithm to those of other widely

used skull-stripping tools, such as BET, BSE, and ROBEX. The results of the

experiment demonstrated that the proposed method can be utilized for syn-

thetic as well as real images. Reddy et al. [18] distinguished and removed the

skull regions of the brain based on opening and closing morphological opera-

tions. Duarte et al. [30] suggested a paradigm for brain extraction, consisting of

data collection, preprocessing, and extraction of the largest connected compo-

nent, in which they utilized digital image processing approaches at each stage.

The purpose of the preprocessing stage was to improve contrast and remove

any potential noise from the T1-weighted MRI. The process of extracting the

largest connected component—the brain—from an image realized by identify-

ing its largest element and then using mathematical morphological operators

to take it out.

In a review study on skull extraction methods, Kalavathi and Prasath [31]

stated that although numerous approaches to skull stripping have been put

forth, the issue is still not thought to be fully resolved. Numerous systems

in the literature perform well on some datasets (T1-weighted images), but

when the study populations or acquisition settings are altered, they are unable

to yield results that are acceptable. Hence, studies that provide applicable

solutions to all the problems presented by skull stripping methods are one of

the demanding research areas in the field of brain tumor detection [31].

4.3.3

Segmentation

Image segmentation infers partitioning an image into separate regions accord-

ing to the characteristics of the pixels in order to reduce the complexity of

the image and make its analysis simpler. In this way, pixels with the same

characteristics on the image will be grouped together and the image will be

divided into subgroups or regions of the same type. Segmentation is utilized

in the simplest cases to distinguish objects from the background, therefore the

segmented image becomes a binary image containing only the foreground and

background [32].

Image segmentation criteria can be stated with the following mathematical

expressions [33]. The image R can be partitioned into n regions R1, R2, ..., Rn

such that:

1.

n

i=1 ^{Ri = R}

2.

Ri is a connected set, i = 1, 2, ..., n