Using Fuzzy Clustering to Detect the Tumor Area in Stomach Medical Images

Although the number of stomach tumor patients reduced obviously during last decades in western countries, but this illness is still one of the main causes of death in developing countries. The aim of this research is to detect the area of a tumor in a stomach images based on fuzzy clustering. The proposed methodology consists of three stages. The stomach images are divided into four quarters and then features elicited from each quarter in the first stage by utilizing seven moments invariant. Fuzzy C-Mean clustering (FCM) was employed in the second stage for each quarter to collect the features of each quarter into clusters. Manhattan distance was calculated in the third stage among all clusters' centers in all quarters to disclosure of the quarter that contains a tumor based on the centroid value of the cluster in this quarter, which is far from the centers of the remaining quarters. From the calculations conducted on several images' quarters, the experimental outcomes show that the centroid value of the cluster in each quarter was greater than 0.9 if this quarter did not contain a tumor while the value of the centroid value for the cluster containing a tumor was less than 0.4.For examples, in a quarter no.1 for STOMACH_1 medical image, the centroid value of the cluster was 0.973 while the value of the cluster centroid in quarter no.3 was 0.280. For this reason the tumor area was found in quarter no.(3) of the medical image STOMACH_1. Also, the centroid value of the cluster in a quarter no.2 was 0.948 for STOMACH_2 while, the value of the cluster centroid in quarter no.4 was 0.397. For this reason the tumor area was found in a quarter no.4 of the medical image STOMACH_2.


Introduction:
The stomach is a segment of the gastrointestinal (GI) tract which receives food from the esophagus, mixes it, breaks it down, and then passes it on to the small intestine in smaller portions. In the stomach, the main secretory functions and digestion process begin. However, the most critical and important function of the stomach in digestive physiology is perhaps gastric motility. In this way, the functions of the stomach are mainly three: (i) the storage of large quantities of food to be further processed in the duodenum, and lower intestinal tract, (ii) the mixing of this food with gastric secretions to form a semi-fluid mixture, and (iii) to slow down the emptying of that semi-fluid mixture into the small intestine at a rate suitable for proper digestion and absorption. Regarding the motor activity, the stomach must consume glucose to generate the power necessary to carry out the digestion process. Although glucose consumption in the stomach is relatively low, it can affect the glucose concentration in the bloodstream (1).
A stomach traditionally is regarded as hollow muscular sac which starts the second digestion phase. So far this view simply disregards the fact that it is the most sophisticated endocrine organ with unique biochemistry, physiology, microbiology and immunology. Whole ingested materials which include the nutrition have to negotiate first this organ and as such the stomach is arguably the most significant piece within the tract of GI. The single biological duty of gastric acid secretion not only starts the digestive operation but acts also as a first defense line against microbes transferred by food-borne. Ordinary gastric physiology and morphology may be disrupted via infection of Helicobacter pylori, the ultimate popular chronic bacterial infection in the world and the etiological factor for most gastric cancer and peptic ulcers (2). Cancer is one of the ultimate 45<8 significant concerns in the life of people. It has the highest death average after the diseases of heart. Also, it has high death average psychologicaleconomic and impairment results which rise due to the delay in diagnosis and accurate treatment. In accordance to the statistics introduced via The Organization of World Health , the cancer of stomach is the second widespread cancer in the world. As 50% of this cancer is inoperable and with available treatments there is no opportunity for more than 5 years life for the sufferers (3). A stomach is both an ingested food reservoir and a mixing and digestion chamber. It continues the procedure of chemical and mechanical digestion within the assistance of a group of gastric enzymes and its different smooth muscles layers, before funneling food turned into chyme into the duodenum. The stomach exists in the upper left quadrant of the abdomen and it is a J-shaped organ consists predominantly of involuntary smooth muscle. A food bolus enters the stomach via the lower oesophageal sphincter that quickly closes to prevent regurgitation of secretions' gastric (4).
There are four main regions in the stomach: 1) The cardia (or cardiac region) is the point where the esophagus connects to the stomach and through which food passes into the stomach. 2) Located inferior to the diaphragm, above and to the left of the cardia, is the domeshaped fundus. 3) Below the fundus is the body, the main part of the stomach. 4) The funnel-shaped pylorus connects the stomach to the duodenum. The wider end of the funnel, the pyloric antrum, connects to the body of the stomach. The narrower end is called the pyloric canal, which connects to the duodenum. The smooth muscle pyloric sphincter is located at this latter point of connection and controls stomach emptying. In the absence of food, the stomach deflates inward, and its mucosa and sub-mucosa fall into a large fold called a ruga. The addition of an inner oblique smooth muscle layer gives the muscularis the ability to vigorously churn and mix food. The convex lateral surface of the stomach is called the greater curvature; the concave medial border is the lesser curvature. The stomach is held in place by the lesser omentum, which extends from the liver to the lesser curvature, and the greater omentum, which runs from the greater curvature to the posterior abdominal wall (5). Figure 1 illustrates the stomach regions, the cardia, fundus, body, and pylorus.

Related Works
Various approaches for segmentation a tumor in a brain and diagnosis of cancer in a stomach were found. For example, by using image processing and artificial intelligence algorithms like primary preprocessors for increasing the quality and statistical characteristics of image, local binary pattern algorithm for elicit characteristics, image histogram algorithm to elicit impaired characteristics and support vector machine were used for accurate classification among impaired and suspected subjects and also for accurate diagnosis of impairment (6).Also, there was the application of two widely used algorithms for tumor detection (i) K-means clustering (ii) Fuzzy C Means clustering. The segmentation algorithms are compared to estimate the efficiency by evaluating the execution time and accuracy of the algorithm (7). The results of the proposed methodology show that the execution time was less in K-Means compared to Fuzzy C-Means clustering method in the process of detecting a quarter that contains a tumor, but the accuracy of the detection was low while the method of fuzzy C-Means method took a long time in the detection process but with high accuracy.

Proposed Methodology
The proposed methodology for detection tumor area in the stomach medical magnetic resonance imaging (MRI) images has three stages. The stomach images are divided into 4 quarters. When the image is divided into 8 or 16 quarters, the results of determining the area that containing the tumor do not change, but rather the number of clusters increases, and therefore requires more time to calculate the seven moments of discharge and more time to calculate the value of the fuzzy of each cluster and the tumor may be distributed over two or more quarters, so for ease of calculation and with the least time and determining only one quarter containing the tumor, the image is divided into four quarters only.

45<9
The stomach images are divided into four quarters and then features elicited from each quarter in the first stage by utilizing seven moments invariant. Fuzzy C-Mean clustering (FCM) was employed in the second stage for each quarter to collect the features of each quarter into clusters. Manhattan distance was calculated in the third stage among all clusters' centers in all quarters to detect the value of center that far away from other centers and determine quarter that contains tumor based on the center's value. The block diagram of the proposed methodology is illustrated in Fig.2.

Features Extraction Using Seven Moments Invariants
This approach has been widely used to image pattern recognition in a variety of applications because of their unchanged features on picture rotation, scaling, and translation. Moment invariants are helpful for calculating sets of region characteristics that can be utilized for shape recognition. Suppose (F (x, y)) defined a 2 dimension array of numbers in a spatial model. A Geometric moment of arrangement (p + q) is illustrated in Eq.1 (8). 45<: Unchanged scale could be obtained by utilizing normalized central moments. An important step for feature extraction is feature normalization. Complex image database use features that are generated by many different feature extraction algorithms with different kinds of sources. These feature vectors usually exist in a very high dimensional space. Not all of these features have the same range. The applied normalization process maps the extracted feature values to the range between [0, 1].

FCM
Fuzzy clustering is a strong unsupervised technique to data analysis and modes building. Objects on the boundaries between several classes are not forced to fully belong to one of the classes, but rather are assigned membership degrees between 0 and 1 indicating their partial membership. Clustering algorithm of FCM is one of the widely employed techniques between associated fuzzy models. Fuzzy clustering techniques can be employed to count the function of membership which defines to which degree the objects belong to clusters and employed to detect the overlapping clusters in the data-set. The FCM utilizes fuzzy partitioning such that a data point can belong to all groups with different membership grades between 1 and 0 (9). This algorithm can work by specifying membership for every data-point that corresponds with every center of cluster based on the basis of distance among the center of data point and a cluster. Further data is near to a center of the cluster further is its membership across a specific center of the cluster. The addition from membership for every data point must be equal to one. This algorithm specifies a membership value to the items of the data for the clusters within a scope of 1 to 0. Thus an amount of fuzzy sets of partial membership can be incorporated and forms overlapping clusters for supporting it (10). This algorithm requires an argument known a fuzzification (m) which has a domain [1, n]. This fuzzification argument employed to determine the fuzziness's degree in these clusters. As the amount of (m) equals 1 this very algorithm works as if it is a crisp portioning algorithm and the clusters' overlapping tend to be more for larger amounts of (m) (11). The algorithm computes the value of a membership via utilized Eq.6 (11).
where , (µ j (X i ) )is the membership of (X i ) in the cluster's ( j th ) , (d ji ) is the distance of (X i ) in the cluster (C i ) , (m) is the parameter of fuzzification , (p) is the number of specified cluster , and (d ki ) is the distance of (X i ) in the cluster (C k ). The latest cluster enters are computed with a value of memebership via employing Eq.7 (11). ∑ ( ) (7) Employing Eq. (7) to compute the latest values of the membership is done and is computed by employing Eq.8 (11).
Where, (C j ) is the center of ( j th ) cluster , (x i ) is i th data point , (m) is the parameter of the fuzzification and (µ j ) is the function that retrun the membership.The weighted average's form can be computed via utilizing this form specially. The fuzziness's degree in (x i 's ) current membership can be computed and this is multiplied by (x i ) (11).
Sum of the fuzzified membership divides the product acquired. Recent centroids are computed in this style.After clustering the seven moments in each cluster, the Manhattan for those features can be calculated via Eq. 9. The Manhattan distance between two components can be calculated by the summation of differences of their identical elements. The distance formula among the points A= (a1, a2, etc.) within the points B= (b1, b2, etc.) can be calculated via Eq.9 (12).

Suggested Algorithm
A suggested algorithm can be offered like: Algorithm 1. illustrates a suggested methodology main steps Input: -Images of medical stomach Output: -Detect the location of quarter that contains tumor Begin Step-1: For k=1 to count of images.
Step-2: acquire Imags[k] and divide Imags[k] into four quarter Step-3: For each quarter do Step-3-1: Feature are detected for all (Images[k]) by employing seven moments invariant and save the outcomes in (SM_Imags[k]) Step-3-2: Apply Fuzzy clustering algorithm to each quarter and save the outcomes in (SMF_Imags[k]).

End //for each quarter
Step-4: Apply Manhattan distance among the centers of clusters using Eq 9 to determine the faraway center.
Step-5: Disclosure of the quarter that contains a tumor based on the coordinates of the center of this quarter, which is far from the centers of the remaining quarters. End // for loop for k End.

Empirical Outcomes
The empirical outcomes of an offered methodology would be explained and offered in this section. C# language is utilized to implement the suggested method. Numbers of stomach medical images of type MRI in a database are 30 images. MRI used in the proposed methodology due to the fact that this type of image is distinguished as it deals with soft tissues , displays images with more details of the segment to be diagnosed within the body and is also capable of imaging in multiple axes and highlighting angles Images in the database are colored Joint photo Graphic Experts Group (JPEG) with size 200 × 200 pixels. The size of the MRI images has been standardized in the proposed system, which are 200 pixels in length and 200 pixels in width for easy handling The proposed methodology involves multiple stages:-1) A process of loading the images of medical stomach onto the project's form can be carried out in the headmost stage as explained in Fig. 3. The stomach images are divided into four quarters rather than divided into 8 or 16  2) In the second step, the fuzzy C-Means clustering was applied to each quarter using the C-Means clustering method as illustrated in Fig.6 for stomach_1 and Fig.7 for stomach_2. 3) Employing Manhattan distance to detect of the quarter that contains a tumor based on the value of the center of this quarter, which is far from the centers of the remaining quarters as displayed in Table 2. Figure 8 and 9 illustrate the detection process.   Table 3 shows the comparison for tumor detection between K-Means clustering method and fuzzy C-Means clustering method in term of time consuming and accuracy. From Table (3), it becomes clear to us that K-Menes method took a little time in the process of detecting a quarter that contains a tumor, but the accuracy of the detection was low while the fuzzy C-Means method took a long time in the detection process but with high accuracy.
The performance of the proposed methodology can be calculated based on accuracy. Accuracy is calculated by the measures of sensitivity and specificity (13). It is denoted as in Eq.10. (13) (10) Where TP displays true positives predication, TN displays true negative predication, FP displays false positives predication and FN displays false negative predication (13).

Conclusions:
The aim of this research is to detect the area of a tumor in a stomach images based on fuzzy clustering. A methodology utilized seven moments invariant to elicit the features of the stomach medical image which is simpler and easer to calculate. Utilizing the invariant ability to the affine transformations of features for handle various input conditions. FCM is really convenient to handle the affair concerning for understanding pattern capability, noise and incomplete data, mixed information of media, interaction of human and it supply the solutions quickly. The Manhattan distance is based on absolute value distance, as opposed to Euclidean distance. Absolute value distance should give more robust results, whereas Euclidean would be influenced by unusual value. From the calculations conducted on several quarters of images, the experimental outcomes show that the centroid value of the cluster in each quarter was greater than 0.9 if this quarter did not contain a tumor while the value of the centroid value for the cluster that containing a tumor was less than 0.4. Form Table 2 in the quarter no.1 for STOMACH_1 medical image, the centroid value of the cluster was 0.973 while the value of the cluster centroid in quarter no.3 was 0.280. For this reason the tumor area was found in quarter no.(3) of the medical image STOMACH_1. Also, the centroid value of the cluster in a quarter no.2 was 0.948 for STOMACH_2 while, the value of the cluster centroid in quarter no.4 was 0.397. For this reason the tumor area was found in a quarter no.4 of the medical image STOMACH_2.