An Efficient Automatic Segmentation Method For Leukocytes

Transcription

1 An Efficient Autoatic Segentation Method For Leukocytes Biji G Assistant Professor Dept of Electrical Engineering Govt Engineering College Bartonhill Thiruvananthapura,695035,India biji2engg@gail.co Abstract Blood tests are of the ost iportant and counting of leukocytes in peripheral blood is coonly used in basic clinical diagnosis. A ajor requireent for this paper is an efficient ethod to segent cell iages. This work presents an accurate segentation ethod for autoatic count of white blood cells. First a siple thresholding approach is applied to give initial labels to pixels in the blood cell iages. The algorith is based on inforation about blood sear iages, and then the labels are adjusted with a shape detection ethod based on large regional context inforation to produce eaningful results. This approach akes use of knowledge of blood cell structure, the experiental result shows that this ethod is ore powerful than traditional ethods that use only local context inforation. It can perfor accurate segentation of white blood cells even if they have unsharp boundaries. Keywords: Leukocytes, Thresholding, Pixels, Peripheral Blood, Segentation. 1. INTRODUCTION There are any different classes of white blood cells present in blood sear iages. Differential count of these various types of cells gives valuable inforation and plays an iportant role in the diagnosis of differential diseases. It s a tedious task to count class of cells anually. An autoatic counter using coputer vision helps this edical test rapidly and accurately. The first step of this autoatic analysis is a segentation of blood cells iages, which differentiates eaningful objects fro the background. In these case attepting to identify white blood cells with an accurate segentation ethods. This step is crucial because the result is the basis of subsequent analysis. The success of classification depends ainly on the correct segentation of iages. It s also difficult and challenging proble due to the coplex nature of the cells and uncertainty in the icroscopic iages. Cells often overlap each other and have variation of different size and shape. The contrast between the cells boundary and the background varies according to illuination in consistency. Many segentation ethods of blood cell iages have been proposed. Histogra thresholding, edge detection and regional growing ethods are often used. Thresholding techniques always can t produce eaningful result since no spatial inforation is used during the selection of the segentation threshold. They are often cobined with atheatical orphology operation. Edge detection ethod perfor poorly on cell iage because not all cells boundaries are sharp, so it s difficult to get all the edge inforation and locate the cells Most of these entioned ethods are on short decision processes and often ake wrong crisp decisions. Relaxation ethods are proposed to avoid it. A fuzzy patch label relaxation algorith used patches that provide ore useful and eaningful context inforation that copact local context to obtain better segentation results. But soeties the inforation of a few neighbour patches still not enough to ake correct decision. For exaple, too bright illuination often leads International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :

2 to bright gaps inside cytoplas, and then a part of cytoplas ay be labeled as red blood cells since it is separate fro the other parts of the cytoplas and is siilar to a red blood cell in position relationship. The neighbour patches play poorly on recovering this istake. Hence a shape detection ethod is proposed to provide fast and accurate segentation by using large regional context inforation. Since ore inforation is used to help ake decision, the approach is ore robust and efficient. In this paper presents a white blood cell segentation approach that consists of three steps. First is siple thresholding approach cobined with atheatical orphology operation is applied to give initial labels to pixel in the blood cell iages. The algorith is based on priori inforation about blood sear iages derived fro a learning process. Next the labels are adjusted with a shape detection ethod based on large regional context inforation, which akes use of knowledge of blood cell structure. At last the regions of white blood cells are arked and the shapes of the regions are arranged to for rounded boundaries. The following sections give a detailed explanation of the proposed ethod along with soe experients results. 2. PROPOSED METHOD 2.1 Initial Segentation The goal of initial segentation is to separate four different regions roughly, background, red cells, cytoplas and nucleus. Inforation of colours, brightness and gradients are used in thresholding to create initial labels of the pixels. The bright white or yellow regions correspond to background. The dark regions correspond to nucleus. The regions that have interediate brightness and sall gradients correspond to red cells. Other regions are labeled as cytoplas teporarily. Most thresholds are derived fro priori inforation of blood sear iages. But soe sensitive thresholds should be selected autoatically to adapt to the illuination variation, which greatly influence the segentation result. To select an optial threshold, we change it fro a saller value to larger one by a sall step and segent the iage with it. If the segentation is correct with this threshold, the edges of the segented iage should have relatively large gradients, which correspond the real edges of different regions. So we calculate the nuber of the edge points that have relatively large gradients while the threshold changes and select the threshold that leads to the largest nuber of edge points to serve as the optial threshold. The regions labeled cytoplas after thresholding are not real cytoplas regions. They are repartitioned to three different parts according to their connection status with other coponents. Regions connected with only red cell regions are labeled as red cell. Regions connected only with nucleus and regions connected with both are relabeled separately. Every tie the labels of pixels change, the different types of regions are soothened with atheatical orphology operation to get connected regions and eliinate single points and lines caused by noises. After initial labeling with thresholding ethod and orphology operation, the blood sear iages are segented roughly into four regions. Pixels that are sure enough to belong to background or nucleus are deterined and other labels of pixels will be adjusted in the following steps. 2.2 Label Connection with Shape Detection Now the initial nucleus regions along with cytoplas regions around the ake up the initial leukocyte regions. They are not real leukocyte regions since there are any other regions in the such as red cell and stain regions which ay be connected with real leukocyte regions. Soe of these false regions can be eliinated by a scan ethod. Real leukocyte regions are connected regions with convex shapes. False regions ay be connected with the, but there still will have soe relatively large gaps between the. So when we scan the iages in a certain direction, back ground regions will be detected between false International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :

3 regions and real regions. Separate regions in this direction can be eliinated. That is although the false regions are connected with real leukocyte regions soewhere; they are separate in certain directions and can be detected by scanning in these directions.fuzzy C-Means is a clustering ethod which allows single data belong to ore than one clusters. This ethod (Dunn, 1973) is a pattern recognition based on iniization of the following objective function given as Equation (2) below: The FCM algorith attepts to partition a finite collection of eleents X={X1,X2,.. Xn} into a collection of c fuzzy clusters with respect to soe given criterion. Given a finite set of data, the algorith returns a list of cluster centres C={C1,C2,.. Cn} and a partition atrix Like the K-eans clustering, the FCM ais to iniize an objective function: where: J u ij N c i 1 j 1 u ij x c i j (1) c k 1 x x i i 1 c c j k (2) c j N uij.. xi i1 N uij i (3) where each eleent uij tells the degree to which eleent xi belongs to cluster Cj. This differs fro the k-eans objective function by the addition of the ebership values uij and the fuzzifier, with 1. The fuzzifier deterines the level of cluster fuzziness. A large results in saller eberships * is any nor expressing the siilarity between any easured data and the center. The algorith is coposed of the following steps: 1. Initialize U=[uij] atrix, U(0) 2. At k-step: calculate the centers vectors C(k)=[cj] with U(k) 3. Update U(k), U(k+1) 4. If U(k+1) - U(k) < then STOP; otherwise return to step 2. After the scan process a large region containing several white blood cells has been segented out. Then we ll give accurate segentation in saller regions each containing a single white blood cell.nucleus regions are always in the centre of the leukocyte. There are any cytoplas regions around the, which ay be labeled incorrectly. The real cytoplas regions can be picked out with a regularity detection ethod. The distance fro a boundary point of the cell to center International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :

4 point of the nucleus is calculated to serve as the radius. If a real leukocyte region is issed, a sharp decrease can be detected while we check the radiuses of boundary points in turn. If a false region is added into the leukocyte region, a sharp increase will be detected. These two cases both lead to great variation in radius. So the cytoplas regions that together for a region with the least variation of the radius should be labeled real leukocyte regions. 2.3 Shape Arrangeent Most pixels have been correctly labeled after the previous processes. But the unsharp boundaries between the cells and the background often lead to sall regions around the white blood cells, which ay be edge regions of the cells and influence the shapes of the cells. The labels of these sall regions can be deterined according to their positions. Hough transforation is applied to find a circle in the iage of edge points, which corresponds to the rough region of the cell. Sall regions over this circle are added to the cell region and the sall gaps inside cell regions are filled to for an orbicular shape of the cell. Although we adjust labels of points at every step to avoid aking false segentation, there is no guarantee that the segentation results are correct all the tie. So we identify the regularity of the final boundary of cells by calculating the variation of the radius. Great variation corresponding to an irregular shape eans possible incorrect segentation. The irregular edges are arked to reind users that these edges should be checked anually to avoid istakes 3. EXPERIMENTAL RESULTS The ethod has been applied to 25 slides and 128 white blood cells have been segented fro background and overlapping red cells successfully. The boundaries of the have been traced out and irregular shapes are arked to avoid istakes FIGURE (1.a.1): Original Iage. FIGURE (1.a.2): Segented Iage. FIGURE (1.b.1): Original Iage. FIGURE (1.b.2): Segented Iage. International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :

5 FIGURE (1.c.1): Original Iage. FIGURE (1.c.2): Segented Iage. International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :

6 4. CONCLUSION The proposed shape detection ethod uses inforation fro a large region that contains a whole white blood cell rather than inforation fro a sall local region or a few neighbour patches. Therefore it is powerful to obtain reliable and eaningful segentation results and trace out accurate boundaries of white blood cells. Although the labels of the pixels are adjusted step by step to avoid wrong crisp decisions that are difficult to be reversed, it is not an iterative approach which akes it faster than ost relaxation ethods. 5. FUTURE SCOPE OF WORK The shape detection and segentation of white blood cell can be approxiated by a quasi circular for or a circular for. The autoatic segentation of WBC s in coplicated and cluttered iages can be approached by circle detection optiization ethods. 6. REFERENCES [1] D.Anorganingru, Cell segentation with edian filter and atheatical orphology operation, Proceedings on International Conference on Iage Analysis and Processing, pp , [2] C.Di Rubeto, A.Depster, S.Khan, B.JArra, SEgentation of blood iages using orphological operators,, Proceedings of 15th International conference on Pattern recognition, Vol.3,pp , 2000 [3] M.B.Jeacocke, B.C.Lovell, A ulti-resolution algorith for cytological iage segentation, Proceedings of International conference on second Australian and New Zealand if Intelligent Inforation Systes, pp , [4] J.Theerapattanakul, J.Plodpai and C.Pintavirooj, An efficient ethod for segentation step of autoated white blood cell classifications, in TENCON, Bangkok 2004 [5] T. Z. Muda, R. A. Sala, Blood cell iage segentation using hybrid K-Means and edian-cut algoriths IEEE International Conference on Control Syste, Coputing and Engineering (ICCSCE), [6] G.Cao, C.Zhong, L.Li and J.Dong, Detection of Red blood cell in Urine icrograph, in ICBBE, [7] F. H. A. Jabar, W. Isail, K. A Rahi, and R. Hassan,"K-Means Clustering For Acute Leukeia Blood Cells Iage", Proceedings of the International Conference on Soft Coputing and Coputational Matheatics (ICSCCM), [8] H.Raoser, V.Laurian, H.Bischort and R.Ecker, Leukocyte segentation and classification in blood sear iages, Proceedings in IEEE Engineering in Medicine and Biology, Shanghai International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :

7 [9] H. Zhang, J. E. Fritts, S.A. Goldan, Iage Segentation Evaluation: A survey of unsupervised ethods, IEEE proceedings on Coputer Vision and Iage Understanding, Volue 110, Issue 2, Pages , May [10] H.T.Madhloo, S.A.Karee, H.Ariffin, A.A.Zaidan, H.O Alanazi and B.B.Zaidan, An autoated white Blood cell Nucleus Localization and Segentation using Iage Arithetic and Autoatic threshold, Journal of Applied Sciences, Vol.10, No.11, pp , 2010 [11] P. Filipczuk, M. Kowal, A. Obuchowicz, Autoatic Breast Cancer Diagnosis Based on K- Means Clustering and Adaptive Thresholding Hybrid Segentation, Iage Processing and Counications Challenges. 3, , [12] A. S. Saa and R. A. Sala, Adaptation of K-Means Algorith for Iage Segentation, International Journal of Inforation and Counication Engineering, [13] N. H. Harun, M. Y. Mashor, and R. Hassan, Autoated Blasts Segentation Techniques Based on Clustering Algorith for Acute Leukaeia Blood Saples, Journal of Advanced Coputer Science and Technology Research , [14] J. Freixenet, X. Munoz, D. Raba, J. Marti, X. Cufi, Yet another survey on iage segentation: Region and boundary inforation integration.,notes in Coputer Science Volue 2352, pp , [15] D. Coaniciu and P. Meer, " Mean shift: A robust approach toward feature space analysis, IEEE Transactions on Pattern Analysis and Machine Intelligence,24: , [16] M. Kunt, M. Benard, and R. Leonardi, Recent results in high copression iage coding IEEE Transactions Circuits Syste, vol.34, 1987, pp International Journal of Iage Processing (IJIP), Volue (12) : Issue (3) :