Normal cells of the human body grow and divide to generate new cells in order to meet the body needs. If these normal cells grow old or destroyed, they die. So the new cells take their place. Cancer occurs when these new cells are generated and the body does not need them; in addition, the old and destroyed cells do not die as they should, so these additional cells are constructed as a mass tissue known as a lump, growth or tumor [1].

One of the main lethal cancers is breast cancer [1,2]. It occurs in both males and females, but the breast cancer in men is scarce [1,3]. Around 25% of all cancers in woman are breast cancer, and approximately 20% of cancers causing death are breast cancers. It primarily sites in ducts (tubas that transfers milk to the nipple) and in the lobules (glands which make milk) in the breast by taking the form as micro-calcifications or masses [1]. Commonly, breast cancer cannot be avoidable, but at least early detection gives a big opportunity of the treatment of this disease [2]. One of the best tools for scanning the structure of the breast is mammogram [1,2]. Mammogram images are classified depended on the categories as malignant, benign and normal. Where benign and malignant are abnormal; benign tumors can be treated (not cancerous). In contrast, to malignant tumors (cancerous), this can damage neighboring tissues and spread to the rest parts of body [1,2]. Mammogram images are represented as very accurate and complex images to be interpreted [1,3]. Moreover, many cases (about 25%) of cancers are not detected in the screen. Therefore, there is a necessary need to improve computer-aided diagnosis (CAD) systems to help a radiologist in his diagnosis for the cancer [3,4]. For these reasons, our aim in this paper is to put the scope on applying different thresholding techniques on mammogram images as a try to determine which one presents best results.

This paper is organized as follows: Section 2 is a brief introduction of segmentation, and mention types of segmentation used in mammograms. It also defined the thresholding, and related works for thresholding used in mammogram images. Section 3 illustrated the formulation of the structure of an image. Section 4 describes Otsu method, and the techniques related to it. Section 5 is about the thresholding evaluation methods. Section 6 is a discussion for experimental results. Conclusions appear in Section 7.

Segmentation in image processing plays a central role in detection the region of interest from background. Its inputs are images and the outputs are the properties were obtained from those images [1,2]. Segmentation of intensity images (such as mammograms), contains four types of methods: thresholding techniques, boundary-based methods, region-based methods, and hybrid techniques that used both boundary and region criteria [1].

The thresholding in mammograms images is based on separated the histogram into background and breast tissues. Depending on the value of threshold all pixels less than the threshold are classified as background, and the reminder pixels are breast or vice versa [1-5]. As a related work there are many researchers used (global or local) thresholds to split the breast region from background; like in [6] presents simple histogram thresholding. Others like in [7-9] used global thresholding to extract the regions of interest (ROI) from the background. In contrary, in [10] local thresholds are used; in addition to region growing and morphological filter. In [5] the researcher applied entropy threshold to segment microcalcifications in mammographic X-ray images. Finally, in [11] a comparison has been done of automatic thresholding methods including Otsu’s method, Kapur ’s method, Pun’s method, Johannes’s method Iterative Selection, Fuzzy Sets, and semi-supervised thresholding method-Black PCT. After the comparison among these methods is done, the experimental outcomes ensured that Otsu method is the best in determination tumor cells.

In general, the previous thresholding techniques have disadvantages including some need manual adjustment of thresholds, and others require local statistics to get the thresholds (local thresholds) [5]. Therefore, we will use one of the best automatic thresholding selection techniques (Otsu method). Its concept based on selected the optimal threshold that maximizes the separability classes in gray levels [12]. Furthermore, many thresholding techniques have been appeared to revise Otsu technique; like [13-16]; each one developed Otsu method in a specific way. This research focuses on implemented thresholding techniques (Otsu method and the techniques related to it) on mammograms.

To analyze and process any image we should know that an image is generated from a set of pixels denoted as n; for each image level there are a set of pixels denoted as n_i. Therefore, the total number of pixels is defined as:

Gray level histogram is normalized and regarded as a probability distribution:

The gray level of an image is [0… L−1]. where the gray level 0 is the darkest and the gray level L−1 is the lightest.

The probability of occurrence of the two classes can be denoted as the following:

(3)

The mean and variance of the foreground and background are denoted respectively as the following:

, (4)

, (5)

It worth to mention that in each image there is a specific thresholding algorithm used to get an optimal threshold, which separated the object from the background.

Otsu method is based on established the optimal threshold that (minimizing intra-class variance) or maximizing inter-class variance. The following equations will represent the within-class variance, and the between class variance respectively.

The final form of between-class variance can also be denoted as the following:

Otsu method is known as: It’s simple, effective, easier to apply and choosing the optimal threshold is automatically and stably [12].

The position of the optimal threshold is displayed in Figure 1 where first class represents the background, while the second class represents the object.

The quality of thresholding technique is a critical issue; it various depending on the type of the thresholding technique and the kind of image. In order to analyze the performance of the thresholding techniques, there are different evaluation methods used to measure their robustness and efficiency. In our study we used two evaluation methods Region Non-Uniformity (NU) and Inter-Region Contrast (GC) [17]. Then, we compare the results of the five thresholding techniques to determine which technique is the best in determination the region of interest (object) from the background.

In the following examples we applied the thresholding techniques on negative mammogram images in order to bring the part of interest from bright region (second mode) to dark region (first mode) and then determine the threshold values. It is worth to mention that the radiologist diagnosis the first Figure 3(a1) normal breast masses, Figure 4(a2) is an image for micro calcifications, while Figure 5(a3) represents breast tissues with small malignnant tumor. In the following examples we will show the performance of each thresholding techniques.

In the first example Figure 3(a1) is the original image of normal breast tissues. In general, and as we see all the five thresholding techniques detect the normal breast tissues. In Figure 3(b), the result of Otsu method that separated the object from the background successfully with T = 157. Figure 3(c), Valley emphasis method also extracted the object from the background with T = 150. Figure 3(d) The third method neighborhood valley emphasis method gave the same result of segmentation of the valley emphasis method with T = 149. Figure 3(e), Variance and intensity contrast method gave T = 141 to separate the object from the background. Finally, Figure 3(f) is the variance discrepancy method with T = 141 that also extracted the normal tissues from the background. The best region non uniformity NU = 0.112841 is introduced from Otsu method, on the other side the best inter region contrast GC = 0.556721 is get from both variance and intensity contrast technique and variance discrepancy technique.

Most early breast cancer can be diagnosed by detecting micro calcifications in the mammogram image. Figure 4(a) is an example for the micro calcifications in the mammogram image. As we see in Figure 4(b) Otsu method did not detect the micro calcifications, and its optimal threshold T = 53. In Figure 4(c) valley emphasis method failed in isolated the micro calcifications from the background; its optimal threshold T = 49. The third method is neighborhood valley emphasis method with threshold T = 78 gives bad results as shown in Figure 4(d). Variance and intensity method succeed in isolation the micro calcifications from the complex intensity background. Its optimal threshold is get from λ = 0.05. Moreover, this method presents the smallest value of region non uniformity NU = 0.00799638 and the smallest

average = 0.169415. On the other hand, the smallest inter region contrast GC = 0.319315 is presented from variance discrepancy method.

Figure 5(a3) is the original image with a small tumor. Figure 5(b) Otsu method with T = 190 did not detect the tumor; moreover, the object details are not clear. The reminder methods like Figure 5(c) valley emphasis method failed to extract the tumor from the rest normal tissues. Figure 5(e) variance and intensity contrast method also failed in detection the tumor, and it reported incorrect threshold with white image. Figure 5(f) is variance discrepancy method also failed in detection the tumor from the background. Only, Figure 5(d) the threshold that is obtained from neighborhood valley emphasis method succeeded in detection the tumor from the background with optimal threshold T = 114 NU = 0.614376, GC = 0.606957, and the average = 0.610667. Moreover, the radiologist diagnoses the tumor with all its protrusions as malignant tumor. Table 1 displays the results of the five thresholding methods including the optimal threshold, region non uniformity, inter region contrast and the average.