Classification of WBC for Leukemia detection A Project Report Submitted in Partial Fulfillment of Requirements for the Degree of PDF

Preview

Summary

Classification of WBC for Leukemia detection A Project Report Submitted in Partial Fulfillment of Requirements for the Degree of Bachelor of Technology by Chimed Palden (2012CSB1009) Gaurav Kumar (2012CSB1011) Kunal Yadav (2012CSB1018) Department of Computer Science & Engineering Indian Institut...

Description

Classification of WBC for Leukemia detection

A Project Report Submitted in Partial Fulfillment of Requirements for the Degree of

Bachelor of Technology

by Chimed Palden (2012CSB1009) Gaurav Kumar (2012CSB1011) Kunal Yadav (2012CSB1018)

Department of Computer Science & Engineering Indian Institute of Technology Ropar Rupnagar 140001, India May 2016

Abstract Leukaemia is cancer of blood-forming tissues, including bone marrow. Many types exist such as acute lymphoblastic leukaemia, acute myeloid leukaaemia, and chronic lymphocytic leukaemia. At the moment, identification of blood disorders is through visual inspection of microscopic images of blood cells. From the identification of blood disorders, it can lead to classification of certain diseases related to blood. We intend to develop a method for detection of leukemia types using microscopic blood sample images. Analyzing through images is very important as from images diseases can be detected and diagnosed at earlier stage. From there, further actions like controlling, monitoring and prevention of diseases can be done. Images are used as they are cheap and do not require expensive testing and lab equipment.

Acknowledgements We would like to express our gratitude to our mentor Dr. Deepti R. Bathula who gave us the golden opportunity to do this wonderful project on the topic Classification of WBC for Leukemia detection and for providing their invaluable guidance, comments and suggestions throughout the course of the project which helped us in doing a lot of Research and we came to know about so many new things.

Honor Code We certify that we have properly cited any material taken from other sources and have obtained permission for any copyrighted material included in this report. We take full responsibility for any code submitted as part of this project and the contents of this report.

Chimed Palden (2012CSB1009)

Gaurav Kumar (2012CSB1011)

Kunal Yadav (2012CSB1018)

Certificate It is certified that the B. Tech. project ”Classification of WBC for Leukemia detection” has been done by Chimed Palden (2012CSB1009), Gaurav Kumar (2012CSB1011), Kunal Yadav (2012CSB1018) under my supervision. This report has been submitted towards partial fulfillment of B. Tech. project requirements.

Dr. Deepti Bathula Project Supervisor Department of Computer Science & Engineering Indian Institute of Technology Ropar Rupnagar-140001

Contents Contents

v

List of Figures

vi

1 Introduction 1.1 General Classification . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Dataset Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 2 2

2 Pipeline Followed 2.1 WBC identication . . . . . . . . . . . . . . . . . . 2.2 Identication and separation of grouped leukocytes 2.3 Image Cleaning . . . . . . . . . . . . . . . . . . . 2.4 Feature Extraction . . . . . . . . . . . . . . . . . 2.5 Arrow Removal . . . . . . . . . . . . . . . . . . .

3 4 5 6 7 7

3 Classification

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

9

4 Result

10

5 Conclusion

11

References

12

v

List of Figures 1.1 1.2

Lymphocyte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comparison between healthy and affected Leukocytes . . . . . . .

1 1

2.1 2.2 2.3 2.4 2.5 2.6 2.7

Pipeline Followed . . . . . . . . . . . . . . Steps in WBC identification . . . . . . . . Example Zack Algorithm . . . . . . . . . . Image before and after applying Watershed Final separation results and image cleaning Original Image . . . . . . . . . . . . . . . Image after Arrow Removal . . . . . . . .

. . . . . . .

3 4 5 6 6 7 8

3.1

SVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

vi

. . . . . . . . . . . . . . . . . . . . . Algorithm . results. . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

Chapter 1 Introduction The analysis of white blood cells (WBCs) allows for the detection of acute lymphoblastic leukemia (ALL), a blood cancer that can be fatal if left untreated.

Figure 1.1: Lymphocyte The round, uniform nucleus and small amount of cytoplasm surrounding it are the best identifying characteristics for this cell. (Bottom) A comparison between lymphocytes suering from ALL: a healthy lymphocyte, followed by lymphoblasts classied as L1, L2 and L3 respectively.

Figure 1.2: Comparison between healthy and affected Leukocytes

1

Lymphocytes are regularly shaped and have a compact nucleus with regular and continuous edges, whereas lymphoblasts are irregularly shaped and contain small cavities in the cytoplasm, termed vacuoles, and spherical particles within the nucleus, termed nucleoli.

1.1

General Classification

Acute leukemia is characterized by a rapid increase in the number of immature blood cells. The crowding that results from such cells makes the bone marrow unable to produce healthy blood cells. Immediate treatment is required in acute leukemia because of the rapid progression and accumulation of the malignant cells. Chronic leukemia is characterized by the excessive buildup of relatively mature, but still abnormal, white blood cells. Typically taking months or years to progress, the cells are produced at a much higher rate than normal, resulting in many abnormal white blood cells. Specific Types: Acute myeloid leukemia (AML) Chronic myeloid leukemia (CML) Acute lymphocytic leukemia (ALL) Chronic lymphocytic leukemia (CLL)

1.2

Dataset Used

We have taken the image dataset from Department of Information Technology Universit degli Studi di Milano. ALL-IDB (Acute Lymphoblastic Leukemia Image Database for Image Processing) is a public and free dataset of microscopic images of blood samples, specifically designed for the evaluation and the comparison of algorithms for segmentation and image classification. Link to dataset:http://homes.di.unimi.it/scotti/all/

2

Chapter 2 Pipeline Followed

Figure 2.1: Pipeline Followed

3

2.1

WBC identication

WBC identication consists of several phases: Conversion from RGB to CMYK color model. Histogram equalization or contrast stretching operations. Segmentation by threshold using Zack algorithm. Background removal operation.

Figure 2.2: Steps in WBC identification

Conversion from RGB to CMYK color model The initial image is in the RGB scale. We first convert the image from RGB to CMYK scale because the leukocytes have no yellow component and appear completely black in the yellow component.

4

Histogram equalization or contrast stretching operations Contrast Stretching is done to increase the contrast so that the leukocytes can be easily thresholded. Segmentation by threshold using Zack algorithm

Figure 2.3: Example Zack Algorithm The triangle method is applied to the image histogram, resulting in a straight line that connects the highest histogram value (h[bmax]) and the lowest histogram value (h[bmin]), where bmax and bmin indicate the values of the grey levels where the histogram reaches its maximum and minimum, respectively. Then, the distance between the marked line and the histogram values between bmin and bmax is calculated. The intensity value where the distance d reaches its maximum denes the threshold value. Background removal operation It consists of removing the background of the image i.e. everything except the leukocytes. The threshold that we obtain from Zack Algorithm is used for background removal.

2.2

Identication and separation of grouped leukocytes

An important problem for the analysis of blood images is the presence of leukocyte agglomerates. Only in this phase we can detect and separate leukocyte

5

agglomerates, because in the previous phase we produced an image containing only the WBCs. Watershed segmentation is then applied to separate the grouped leukocytes.

Figure 2.4: Image before and after applying Watershed Algorithm

2.3

Image Cleaning

Image cleaning requires the removal of all of the leukocytes located on the edge of the image and all abnormal components (non leukocytes), which prevents errors in the later stages of the analysis process. Cleaning the image edge is a simple operation. First, the size of the area and the size of the convex area are computed for each leukocyte. The size of the area is used to calculate the mean area, which is necessary to determine and eliminate components with irregular dimensions.

Figure 2.5: Final separation results and image cleaning results.

6

2.4

Feature Extraction

In this phase we transform the images into data and then to extract information reflecting the visual patterns that pathologists refer to, while simultaneously extracting the descriptors that are most relevant to the subsequent classification process. We extracted three different types of descriptors from the sub-images: Roundness, Elongation and Eccentricity. Roundness: It is the measure of how close the cell is to a perfect circle. Formula to calculate Roundness: 4 ∗ pi ∗ Area/(convexperimeter)2 Elongation: It is the measure of how elongated a cell is. Formula to calculate Elongation: 1 − (minoraxis/majoraxis) Eccentricity: It is the ratio of the distance between the foci of the ellipse and its major axis length; this value is p between 0 and 1. Formula to calculate Eccentricity: (majoraxis)2 − (minoraxis)2 /(majoraxis)

2.5

Arrow Removal

The image database had arrow in them so a matlab file was created to remove the arrow from all the images. Sample images before and after operation are shown below:

Figure 2.6: Original Image

7

Figure 2.7: Image after Arrow Removal

In the arrow removal process we give the arrow same color as the background. Part of arrow in the background is given color as the background and part of arrow on the blood cell is given same color as the blood cell.

8

Chapter 3 Classification The classification process was carried out using 3 features (Roundness, Elongation and Eccentricity) using the Support Vector Machine (SVM) classier because this model is particularly suitable for binary classification problems for which the separation between classes depends on a large number of variables.

Figure 3.1: SVM

9

Chapter 4 Result Training was done on a subset of 40 images acquired from the same camera and under the same lighting conditions using a SVM. Testing was done on a subset of 10 images and we got an average accuracy of 86%. The performance of the proposed method for WBC identification is excellent in most cases. The worst results were achieved in images with significant overlapping between leukocytes, which are even difficult for human experts to identify.

10

Chapter 5 Conclusion We propose a method for the automatic identification and classification of leukocytes using microscopic images, providing an automated procedure to support the recognition of ALL. Our results indicate that the proposed method is able to efficiently identify the WBCs present in an image and to properly classify leukoblasts with great accuracy. The next step for this work will include further development of the identification, thresholding and segmentation phase. These improvements are necessary to increase the accuracy of counting WBCs and increase the overall accuracy of segmentation because accurate segmentation leads to a more robust extraction of shape features, which is essential for this type of problem. Moreover, it will be important to study and analyse the use of new features that may be decisive for this type of analysis. Then, the selection of the most discriminatory features will provide the highest level of accuracy. Further development of the proposed method could affect the separation of adjacent leukocytes, which is of considerable importance to account for all leukocytes in the image.

11

References Leucocyte classification for leukaemia detection using image processing techniques by Lorenzo Putzu, Universit degli studi di Cagliari Bennett, J.M., Catovsky, D., Daniel, M.T., Flandrin, G., Galton, D.A., Gralnick, H.R., Sultan, C.: Proposals for the classication of the acute leukemias. French-American-British (FAB) co-operative group. British Journal of Hematology, vol. 33, no. 4, pp. 451-458, August 1976 Inaba, H., Greaves, M., Mullighan, C. G.: Acute lymphoblastic leukaemia. The Lancet, vol. 381, no. 9881, pp. 1943-1955 June 2013

12...