Application of Image Processing Algorithms for Brain Tumor Analysis in 2D and 3D Leading to Tumor’s Positioning in Skull: Overview

Segmentation of brain tumors has been found challenging throughout in the field of image processing. Different algorithms have been applied to the segmentation of solid or cystic tumors individually but little work has been done for solid cum cystic tumor. The papers reviewed in this article only deal with the case study of patients suffering from solid cum cystic brain tumor as this type of tumor is rarely found for the purpose of research. The research work conducted so far on this topic has been reviewed. The study begins with 2D (Two Dimensional) segmentation of tumor using MATLAB. It is then extended to study of slices of tumor and its volume calculation using open source software named 3D Slicer which represents the tumor in 3D. This software can intake the 2D slices and process them to give a combined 3D view. Various techniques are available in the software. According to the particular requirement an appropriate algorithm can be chosen. This paper gives a promising hierarchy for volume calculation of tumor and the three dimensional view. Further we can also find the position of tumor in the skull using the same software. This piece of work is a valuable guideline for the researchers interested in segmentation and three dimensional representations of different areas of human body. The models extracted out using the given algorithms can also be treated for matching and comparison of any future research. This will also aid surgeons and physicians in efficient analysis and reporting techniques.


INTRODUCTION
D iseases like tumors are always found to be dangerous and are to be handled with special care. Tumor may be benign or malignant [1].
In any case if the tumor is found in the brain then expert physicians and surgeons have to make the proper treatment and its analysis. This is made by conducting the MRI (Magnetic Resonant Imaging) scan or CT (Computed Tomography) scans of the brain [2][3]. MRI is considered to be a very sensitive test and a single movement may cause the patient to go through the test again as the magnetic resonant imaging machine works on the principle of quantum mechanics dealing with spin of protons and neutrons [4]. There are different pulse sequences which are applied to samples resulting in the distinctive reconstruction of images. These are useful in generating different contrasts between the tissues, thus providing a wide variety of choices to the radiologists. Some commonly used MRI scans include T1weighted MRI, T2 weighted MRI and FLAIR (Fluid Attenuated Inversion Recovery) [5]. In T1 weighted image the image contrast and brightness are determined.
On T1 weighted images fat appears bright and fluid dark.
Whereas on T2 weighted images fluid appears bright and fat dark. Therefore, T1 weighted images are used to study solid part of tumor and T2 weighted images are used to study the cystic part. Other than these, FLAIR uses Inversion-Recovery Pulse Sequence. It is basically used to null signal from fluids. Signal from any particular tissue can be suppressed by selecting appropriate inversion time. It can be used in brain imaging to null the signals from CSF (Cerebral-Spinal Fluid). Work has been done to show the brain images into their respective Grey and White matters [6]. During MRI even a slight motion of the patient could produce artifacts on image [7]. It may be noise. The noise reduction filter is very important for the image enhancement [8]. Noise reduction then leads to segment out the ROI (Region of Interest). Segmentation turns out to be a decisive step in image processing [9]. It includes preprocessing This review is carried out considering various research papers which give different techniques for 2D and 3D views and analysis. This paper will begin with review of 2D papers and followed by 3D analysis. Each paper with its results is studied individually.

Materials
The material includes a dataset of cases of patients

2D Segmentation using T2 and T1 Weighted Images
The first technique used for segmenting out tumor begins with morphological reconstruction of the images. The visible maximum in this view whereas coronal and sagittal views can be taken as well [10]. Morphological reconstruction shows accurate results. Solid part is segmented using contrast enhanced T1 where as cystic part is taken from T2. The procedure begins by using a filter for noise reduction which is important for the image enhancement. Instead of taking the mean, the median of the neighborhood pixels are used to find the output using the default 3x3 neighborhoods as the mean filter resulted in merging the edges of tumor [10]. The texture of tumor is same as the rest of brain so the Mean filter results in blurring the image. To trace the boundaries automatically Sobel mask will be used which will give the information by using the method of convolution of two kernels [10].
Sobel results are more accurate than other filters. Other than the brain tumor, many edges were also detected which will be difficult to segment if any other filter was applied.
The above kernels in Equation (1) are used to detect the horizontal (h y ) and vertical (h x ) edges.
The kernels will be specified for horizontal and vertical edges respectively and then gradient method is applied to detect the edges [10]. Morphological reconstruction will be done initiating with thickening the boundaries by dilating using disk shape followed by eroding [12]. The morphological operations for Image Segmentation and Tracking Edges include procedures in which object undergoes the opening i.e. removing small objects and closing [13].  1(a). T2 WEIGHTED FIG.1(b). T1 WEIGHTED CONTRAST ENHANCED block diagram as shown in Fig. 2. The transparency which is superimposed on original image is shown in Fig. 3.The arrow is pointing at the tumor detected in Fig. 3.

2D Segmentation Using T1 Images
The second technique is based on the study of intensities of the region of interest [15]. This method is applied on the T1 weighted images only. It begins with using the median filter which also has the advantage of edge keeping along with the noise removal. After filtering the edges are detected using the canny option of MATLAB [15]. Among In order to segment out the solid part convert the image to label from binary. An array of structure will be obtained.
The main part of the technique lies in extracting the solidity of areas applying the command of region props (MATLAB). The commands applied on the basis of intensity and will result in labeled tumor.
The cystic part will be segmented by applying the region growing method [15]. Cystic part appears with other parts of brain. Apply the Bwareaopen (MATLAB) command on the image with cystic part to remove the unwanted region. Dilation and erosion technique will be used here.
Finally the OR logic operation, conversion RGB and to transparency in applied. The segmented tumor is then imposition in original one [15].In the block diagram red portion indicates the tumor detected.
In this technique it is to be kept noticed that as the tumor size changes the structure element will also change its size. Therefore, the threshold level will vary for different cases and has to be kept changing accordingly both for the solid and cystic part. The above methodology is illustrated as in block chart Fig. 4. Further a comparison of few similar algorithms with the above discussed algorithm is given in Table. 1.

Technique for 3D Volumetric Representation of Human Skull with Tumor
This study is further extended to the modeling of skull from original data set. T1 or T2 any of the images can be taken as in both of them the boundary or the edges of the skull in all the slices is visible. It will give the presence of tumor in the skull and analyze what trajectory is to be followed from external of the brain to the tumor. This is started by loading all the images into slicer libraries [19]. The images are then directly loaded in to the editor module for mapping of complete areas visible along with the edges. The surface model maker will then combine all the labeled slices to give the complete view of the skull [20]. Here triangle smoothing can be applied to have a neat view of the skull. This model is also stored with its tumor in the model hierarchy tree present in the models module. Call all the parts one by one and will get a complete view of the original skull bearing tumor which is shown in Fig. 6. The algorithm turns out to be simpler as compared to neurofuzzy techniques applied for the segmentation of tumor and then its modeling in order to show the lesion in the brain [21].The objective of neurosurgical resection of brain lesions is, maximal cutting out of tumor or lesion with minimal permanent damage to the neighboring normal brain tissue [22]. It can be made possible once the tumor distance and location from the skull is visible.

CONCLUSIONS
All the above given algorithms have been applied to various cases and gives good results with 98% efficiency on an average.
The segmentation being an important part of image analysis is done effectively in 2D .The study of 3D analysis also gives promising results of volume calculation and the model of tumor. The tumor representation in skull will be helpful for the surgical purposes. The skull modeling will facilitate the surgeons in biopsy and radiotherapy.