Research Paper on Segmentation Methods in the Medical Imaging Process

Introduction

Medical imaging is a vital process that helps health professionals to diagnose and treat certain diseases (18). The study of medical images relies on the optical interpretation of radiologists. The process of Medical imaging is time-consuming and accurate result depends on the experience of the radiologist. To overcome these challenges, it is necessary to use computer-aided systems such as digital image processing to visually interpret digital information (2). Digital image processing does not work independently but it requires the use of other techniques such as machine learning and pattern recognition. In digital image processing, there are three layers which must be completed to have to produce an accurate image (5). These layers' image processing, analysis and image understanding. Image segmentation is a very critical step that must be completed in image analysis and it is done to extract useful information from the image (19). The result generated from image segmentation is very important in influencing the accuracy and correctness of image measurements and therefore the use of the computer-aided system is of importance when carrying out medical imaging.

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Save 25%

Image segmentation is used in several areas of medicine, it is used for diagnosis, the study of anatomical structures, localization of pathology, treatment and computerized surgery. Because anatomical structures of the animal body are variable and complex, medical image segmentation is a useful step that must be included when studying anatomical structure of animals (15). Image segmentation is, therefore, a process through which digital images are divided into different segments. In medicine, it allows healthcare officers to identify certain objects within the image. It is difficult to do it by using naked eyes but with the help of a computer algorithm, it works more effectively (10). Image segmentation takes place in three stages namely image processing which requires the radiologists to eliminate useless data from the image while initial object discrimination is the division of the image into two different attributes (20). Finally, the last stage is object boundary clean-up which involves reduction of the object into a pixel width. As a result of the emergence of various technologies, there are different approaches to image segmentation which depends on certain image qualities.

Detecting discontinuities is a property which requires image separation through alteration of intensities and it involves the use of an algorithm such as edge recognition. Another quality required when segmenting image is detecting similarities where an image is divided into various regions that have the same qualities (16). It includes a segmentation algorithm such as thresholding and merging. Thresholding is a widely used approach in which an image is embodied as a group of pixels that has a higher value equal to or less than threshold values (22). Another important approach commonly used in region segmentation is clustering approach which requires partitioning of an image into two separate groups of pixels that have the same space qualities.

Edge Detection Segmentation

In this approach, there is an attempt to solve image segmentation by discovering or identifying edges of an image between various parts. These edges must have been exposed to a higher intensity during the higher transition and they are joined to create adjacent object margins forming binary images (2). Edge-based segmentation is two number namely grey histogram and gradient methods (4). In image processing, edge detection is an established field and it is used as a base for another segmentation method. The disconnected edge only can be discovered through edge detection and in order to execute image segmentation, it is important to have close region boundaries (14). Required edges involve borders that exist in such objects. In addition, it is possible to apply a segmentation technique to other edges received from edge detectors. Integrated edge segmentation was created by Lindeberg (8) and it can be used in segmenting edges into a straight and curved edge.

Threshold Method

It is possible to choose a threshold algorithm manually on the basis of the earlier information by image information (21). The algorithms used in this method of image segmentation are also separated in accordance with edge based, hybrid and region based (13). The edge-based algorithm is commonly used because it is associated with edge information and it can allow for the depiction of the object structure by the edge point. There are other algorithms categorized as common edge detection comprises of the Canny edge detector and Laplacian edge detectors (17). They are usually applied in identifying edge pixels and at the same time removes any action that causes noise. In most cases, canny edge detectors take advantage of the threshold of the gradient magnitude to determine the required edge pixels and prevent them by way of the non-maximal clampdown and hysterics thresholding. Because the actions taken in this algorithm are associated with pixels, the identified edge corresponds with discrete pixels thus they are incomplete (23). It is, therefore, necessary to use post-processing such as morphological activities that link the breaks or removes all the holes.

The use of this technique can ensure that 3D is segmented accurately but the only problem it has is that it makes it very hard to process the image of textured blob objects (12). It is also a very simple approach that can be used in segmenting images that has light objects that are located in dark areas (1). Thresholding method always relies on image space regions such as features of the image (4) and it is capable of changing the multilevel image to dual image. It helps in the selection of threshold T that separates image pixels into different parts and divide the objects at the back. Pixel (x,y) is used as object segment only when its intensity is more than the threshold value Such as f(x,y)T (3, 11). Through the choice of thresholding values, there are two thresholding techniques (12) usually called global and local. In a situation where T is constant, the method is referred to global while it is not constant it is called local thresholding. In most cases, global thresholding becomes unsuccessful when there is uneven background illumination while local thresholding is applied in rewarding uneven illumination (8). Thresholding method has some weaknesses which make it ineffective in several ways. Because it cannot be used in multichannel images and also does not consider spatial features of the image as a result of its sensitivity to noise (4).

Region-Based Segmentation

When it is compared to edge detection, region-based segmentation is easy to use and it is also more resistant to noise (4, 6). The methods such as region growing are used in segmentation algorithm and it is a procedure that integrates pixels into different groups (2,3). There are four processes that can be used in the region growing and they include identifying a group of pixels in the main image (7), identification of the same criterion like grey level intensity, growing region and finally ensure that the growing regions are stopped. Splitting of the region and amalgamation instead of selecting seed points allows users to subdivide the image into arbitrary split regions and at the end merge it (2,4) to meet the condition required when segmenting the image.

Segmentation Based on Clustering

In this segmentation method, there is a need to identify a given cluster to help in the classification of pixels (17). It does not require training stage but there are available data that can be used to train members. The clustering method is applied in cases where classes have been identified and the same criterion is explained by different pixels (2). The same pixels are then grouped into given sets to form clusters. In most cases, the pixels grouped together into clusters are usually done on the basis of the theory of maximizing the intraclass similarity and maximizing the interclass similarities (7). To ensure there is the high quality of the clustering result, it is necessary to use a good similarity measure and an effective implementation process. In most cases, the clustering algorithm usually categorized into hard clustering, K-means clustering and others.

Hybrid Image Segmentation

Image segmentation has some general problems on how it can be used to divide the image into homogenous sections in a manner that allow for the combination of two adjacent objects. Various methods are available that can be used to partition error-free images like histogram based (11). This method uses both edge-based and region based methods of segmentation and it requires partitioning of the image into various groups before merging similar groups together through the application of split and merge methods.

Conclusion

In this paper, there is an overview of different segmentation methods that are used in the medical imaging process. The various image segmentation methods discussed in this paper include edge detection, region-based and hybrid segmentation methods. All of these image segmentations are used to support medical practitioners when carrying out their duties.

References

[1] Rastgarpour M., and Shanbehzadeh J., Application of AI Techniques in Medical Image Segmentation and Novel Categorization of Available Methods and Tools, Proceedings ofthe International MultiConference of Engineers and Computer Scientists 2011 Vol I, IMECS 2011, March 16-18, 2011, Hong Kong.

[2] Zhang, Y. J, An Overview of Image and Video Segmentation in the last 40 years, Proceedings of the 6th International Symposium on Signal Processing and Its Applications, pp. 144-151, 2001.

[3] Wahba Marian, An Automated Modified Region Growing Technique for Prostate Segmentation in TransRectal Ultrasound Images, Master's Thesis, Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Ontario, Canada, 2008.

[4] W. X. Kang, Q. Q. Yang, R. R. Liang,"The Comparative Research on Image Segmentation Algorithms", IEEE Conference on ETCS, pp. 703-707, 2009.

[5] N. R. Pal, S. K. Pal,"A Review on Image Segmentation Techniques", Pattern Recognition, Vol. 26, No. 9, pp. 1277- 1294, 1993.

[6] H. Zhang, J. E. Fritts, S. A. Goldman,"Image Segmentation Evaluation: A Survey of unsupervised methods", computer vision and image understanding, pp. 260-280, 2008

[7] K. K. Singh, A. Singh,"A Study of Image Segmentation Algorithms for Different Types of Images", International Journal of Computer Science Issues, Vol. 7, Issue 5, 2010.

[8] T. Lindeberg and M.-X. Li "Segmentation and classification of edges using minimum description length approximation and complementary junction cues", Computer Visionand Image Understanding, vol. 67, no. 1, pp. 88--98, 1997.

[9] P. Karch, I. Zolotova,"An Experimental Comparison of Modern Methods of Segmentation", IEEE 8th International Symposium on SAMI, pp. 247-252, 2010.

[10] H. G. Kagami, Z. Beij,"Region-Based Detection versus Edge Detection", IEEE Transactions on Intelligent information hiding and multimedia signal processing, pp. 1217-1221, 2009.

[11] L.Aurdal,"Image Segmentation beyond thresholding", Norsk Regnescentral, 2006.

[12] Y. Zhang, H. Qu, Y. Wang,"Adaptive Image Segmentation Based on Fast Thresholding and Image Merging", Artificial Reality and Telexistence-Workshops, pp. 308-311, 1994.

[13] Y. Chang, X. Li,"Adaptive Image Region Growing", IEEE Trans. On Image Processing, Vol. 3, No. 6, 1994. Prof. Dinesh D. Patil et al, International Journal of Computer Science and Mobile Computing Vol.2 Issue. 1, January- 2013, pg. 22-27 2013, IJCSMC All Rights Reserved 27

[14] T. Gevers, V. K. Kajcovski,"Image Segmentation by direct region subdivision", Proceedings of...