Red Blood Cell Counter Using Embedded Image Processing Techniques Mariya Yeldhos* and K.P. Peeyush 1 Department of Biomedical instrumentation engineering, Faculty of Engineering, Avinashilingam Institute for Home science and Higher Education for Women, Coimbatore-641108, India. 1 Department of Electronics and Communication Engineering, Amrita school of engineering, Amrita Vishwa Vidyapeetham, Coimbatore, India-641112. Corresponding author: mariya.yeldhos@gmail.com ABSTRACT Red blood cells or erythrocytes have a special protein called as hemoglobin, which helps in the transport of oxygen from lungs to all living tissues and removes carbon dioxide by exhalation. Erythrocyte count gives information regarding the patient health status on hematological diseases and can be differentiated into mainly anemia and polycythemia. Conventional method of erythrocyte counting by hemocytometer may lead to human errors. Machines developed for erythrocyte counting causes high cost and time consumption. This work emphasize on FPGA implementation of image processing algorithm for segmented erythrocyte counting. Ycbcr Color conversion algorithm is used for erythrocyte segmentation from the blood smear image and connected circular hough transform is used for counting. The proposed algorithm is implemented on the FPGA board and the accuracy of the technique is evaluated. KEYWORDS: Clustering, FPGA, System generator, Simulink. Citation: Yeldhos M, Peeyush KP (2018) Red Blood Cell Counter Using Embedded Image Processing Techniques. Research Reports 2:e1-e5. doi:10.9777/rr.2018.10325. www.companyofscientists.com/index.php/chd e1 Research Reports
I. INTRODUCTION Blood is a body fluid that helps in the transportation of oxygen and nutrients, forming blood clots to prevent excess blood loss. It also plays a vital role in regulating body temperature and protecting the immune system. The main components in the blood are plasma, red blood cells, white blood cells and platelets. Based on the count of leukocytes and erythrocytes, proper functioning of immune system, hidden infections and other undiagnosed medical conditions can be detected. White blood cells or leukocytes, plays important role in the human immune system that protects the body against foreign bodies and other infections. They are 3,500 to 10,500 leukocytes per microliter of blood, which accounts only for one percent in healthy people [1]. Red blood cells or erythrocytes have a special protein called as hemoglobin, which helps in the transport of oxygen from lungs to all living tissues and removes carbon dioxide by exhalation. Normal range of erythrocytes varies from 4.7-6.1 million erythroctes per microliter of blood in male and 4.2-5.4 million erythrocytes per microliter of blood in females [2]. Leukocytes can be divided into granulocytes and agranulocytes. Granulocyte which has granules consists of neutrophils, eosinophils and basophil. Agranulocytes does not contain granules and consists of lymphocytes and monocytes. Based on the abnormal count of leukocytes they can be divided into leukocytopenia and leukocytes. Leukocytopenia is the reduction in the leukocytes in blood. It can be caused due to chemotherapy, radiation therapy, HIV, bone marrow diseases, autoimmune disorders, infections etc. This is called as immuno-suppression as it leads to the weakening of the immunity. Leukocytosis is caused by abnormal increase in the leukocyte count due to tissue damage, smoking, viral infections, inflammatory conditions, leukemia etc. Individual with low count erythrocytes are anemic and with high count can develop polycythemia. Some causes of anemia are nutritional deficiency, trauma, bone marrow disorders, kidney failure etc and for polycythemia it includes dehydration, lung diseases, smoking and congenital heart diseases. Manual counting using hemocytometer is the common method to count leukocytes and erythrocytes in laboratories. The number of blood components can be counted from the quadrants of counting chamber by using a microscope. Manual counting of blood cells for large samples are prone to human errors, tiresome and time consuming. Automatic modern analyzers for counting are present, but the complexity and the cost of device are high. Image processing methods can be implemented for the segmenting and counting of white blood cells and red blood cells from the blood smear images. This compares the image processing techniques used to segment and count erythrocytes. II. MATERIALS AND METHODS A. Database For leukocyte image processing, from Acute lymphoblastic leukemia image database (ALL-IDB) 108 blood smear images were obtained which consists of 49 blast cell images and 59 normal images. These images have been captured with an optical laboratory microscope attached with a high resolution camera. The blood smear images in dataset are in JPG format. The leukocytes are blue color as the blood smear is stained by the Wright s stain, and thus it appears darker than the background. B. Segmentation techniques for leukocytes The segmentation of leukocytes from the blood smear images is the foremost step in image processing. The Leukocytes are segmented using two techniques that include, Color based K-means clustering and Ycbcr color conversion. www.companyofscientists.com/index.php/chd e2 Research Reports
C. Segmentation technique for erythrocytes The red blood cells in the Fig. 1 are segmented using the Ycbcr color conversion, morphological operators and watershed algorithm. Subtracting leukocyte segmented image and original binary image, erythrocytes from the image can be segmented out. The morphological operator such as opening is applied on the segmented image to remove disjoints as shown in Fig. 3. After segmenting the erythrocytes from the blood smear image, the overlapping and clumped erythrocytes in the segmented image are removed using watershed algorithm as shown in Fig.4. Watershed algorithm is a region based segmentation technique which improves the segmentation accuracy [3]. Euclidean distance transform is used to find the distance from a pixel to the adjacent non-zero valued pixel [4]. The erythrocytes are counted using circular hough transform after the watershed algorithm as the overlapped cells will be separated. Fig 2. Blood smear image Fig 3. RBC count before watershed transform Fig 4. Watershed transformed image Fig 5. RBC count after watershed transform Fig 1. Erythrocyte segmentation steps using Ycbcr conversion, morphological operator and watershed technique. D. Counting of leukocytes and erythrocytes After the segmentation of leukocytes from the blood smear image, leukocytes are counted using connected labeling algorithm in Simulink. It labels and counts the connected components in a binary image. Based on the 8 point connectivity, the nearest pixels in the leukocyte segmented image are grouped as scanning is done horizontally and www.companyofscientists.com/index.php/chd e3 Research Reports
vertically. Erythrocytes are devoid of nucleus and are in circular in shape. Thus circular though transform can be used to detect the presence of the circular objects in the images and to count the segmented erythrocytes. Using this technique objects that are not circular in shape are eliminated, thus increasing the accuracy of erythrocyte count. E. Hardware implementation in Zynq 7000 SoC A simulink model for erythrocyte counting is developed in Matlab using library browser. System generator is a block in Xilinx blockset library, which is configured with Zynq 7000 SoC. The designed simulink model consists of preprocessing and postprocessing blocksets that are present in computer vision and DSP system toolbox. Counting of erythrocyte are done using label block in morphological operations of simulink library. The model designed to count leukocytes is simulated in Matlab simulink and code is downloaded to Zynq board using system generator [5]. III. RESULTS AND CONCLUSION Results The counting results of the erythrocytes segmented algorithms are tabulated below Blood component Erythrocyte Table 1. Segmentation technique Ycbcr color conversion and morphological operators Combination of Ycbcr color conversion, morphological operators and watershed algorithm Accuracy 86.82 90.98 The results from Table 1 shows that for Erythrocyte segmentation, with watershed algorithm the accuracy will be high for Ycbcr technique, clumped and overlapping cells will be removed. Conclusion The comparison between different techniques for erythrocyte segmentation is done. Segmentation method for erythrocytes counting with high accuracy is implemented in Zynq 7000 SoC. Based on the count of erythrocytes in blood smear image they can be differentiated into anemia and polycythemia. Hardware implementation of erythrocytes counting improves the counting efficiency of blood cell components, as it requires less power and complexity. ACKNOWLEDGEMENT The authors sincerely acknowledge their gratitude towards Mr. Fabio Scotti, for providing them with high quality blood smear image database. CONFLICT OF INTEREST Authors declares there is no conflict of interest REFERENCES [1] V. Piuri, F. Scotti, "Morphological classification of blood leucocytes by microscope images", in Proceedings of the 2004 IEEE International Conference. on Computational Intelligence for Measurement Systems and Applications (CIMSA 2004), Boston, MA, USA, pp. 103-108, July 12-14, 2004. [2] J. M. Sharif, M. F. Miswan, M. A. Ngadi, S. Hj, and M. Mahadi, Red Blood Cell Segmentation Using Masking and Watershed [3] Algorithm : A Preliminary Study, International Conference on Biomedical Engineering (ICoBE), Penang, Malaysia, February 27-28, 2012. [4] H. Cheng, X. Jiang, Y. Sun and J. Wang, "Color image segmentation: advances and prospects", Pattern Recognition, vol. 34, no. 12, pp. 2259-2281, 2001. [5] N. John, Viswanath, A., Sowmya, V., and Soman, K. P., Analysis of various color space models on effective single image super resolution, Advances in Intelligent Systems and Computing, vol. 384, pp. 529-540, 2016 [6] L. Vincent, and P. Soille, Watersheds in Digital Spaces: An Efficient Algorithm Based on Immersion Simulations, IEEE Pattern Analysis and Machine Intelligence, vol. 13, no. 6. pp. 583 598, 1991. [7] J. Roerdink and a Meijster, The Watershed Transform: Definitions, Algorithms and Parallelization Strategies, Fundamenta Informaticae, vol. 41, no. 1 2, pp. 187 228, 2000. www.companyofscientists.com/index.php/chd e4 Research Reports
[8] Athmasri. B. Krishnan and K. P. Peeyush, Blood group determination using vivado system generator in Zynq SoC, IFMBE Proceedings, vol. 52, pp. 166 169, 2015. www.companyofscientists.com/index.php/chd e5 Research Reports