Automatic Segmentation and Disentangling of Chromosomes in Q-

Band Prometaphase Images

Presented By:KAUSHIK BOSE - ECE/2009/003RAJENDRA PRASAD MITRA - ECE/2009/013SUMAN MANDAL - ECE/2009/022KOUSHIK BHATTACHARYYA - ECE/2009/018:

CONTENTS

1. Introduction2. Methodology3. Hurdles and complications4. Feature extraction5. Transformation and enhancement used6. Flow process7. Conclusion

INTRODUCTIONKaryotype analysis is a widespread procedure in cytogenetics to assess the possible presence of genetics defects. The procedure is lengthy and repetitive, so that an automatic analysis would greatly help the cytogeneticist routine work.

Using karyotype we can diagnose Aneuploidy i.e. an abnormal number of chromosomes in the cells of an individual.major example is Down's syndrome. Besides aneuploidies there are other genetic diseases, other chromosomal abnormalities and also the genetic mutations.

Hence, an automatic procedure to obtain the separated chromosomes, which are then ready for a subsequent classification step, would have significant implication.

METHODOLOGY• The first step for classification is the extraction of chromosome

features.• Chromosomes features include length,area,density profile and

its perimeter.• Polarization: Chromosomes in the acquired image are

randomly rotated. A polarization procedure is needed both to comply with the orientation standard adopted by cytogenetics.

• Feature Pre-processing: The dataset used in this study contains images acquired with different zoom and illumination conditions, hence data has to be normalized first.

• A prior trained Artificial neural network is used as classifier.• The Reassigning Algorithm: The human karyotype contains 23

chromosomes hence classification is constrained on this fact and classified.

HURDLES AND COMPLICATIONS

• The appearance of chromosomes depends on the stage of the cell division cycle and are visible as distinctive entity only in prometaphase stage.

• Dealing with touches and overlaps.• Nonsharp margins of the chromosomes.• Presence of staining debris due to restricted image resolution.• Uneven illumination of the field of view by the UV lamp.• Low resolution images.

FEATURE EXTRACTION• As of now we are able to extract chromosomes features using

MATLAB and Mathametica.• Using individual chromosomes dataset we apply certain image

enhancement technique and transforms to compute the required data.

• The classification features starts from an estimation of the chromosome medial axis, derived by a skeletonization procedure. Unfortunately, in many automatically segmented images this method provides a skeleton with a lot of spurious branches.

• Hence spurious branches have to be removed using spur reduction technique.

FEATURE EXTRACTION(cont.)

• After spur reduction and thinning procedure we are left up with medial axis which is of foremost importance,

• Using medial axis we can compute length of individual chromosomes.

• Density profile of each chromosomes is obtained by drawing perpendicular over medial axis.

• The area and perimeter of chromosomes is evaluated after suited transform over the binary image.

TRANSFORMS AND ENHANCEMENT USED

• fspecial(‘gaussian ', hsize, sigma) is a MATLAB code that returns a rotationally symmetric Gaussian lowpass filter of size hsize with standard deviation sigma (positive). hsize can be a vector specifying the number of rows and columns in h, or it can be a scalar, in which case h is a square matrix. The default value for hsize is [3 3]; the default value for sigma is 0.5.

• imfilter(i.e.'replicate') is a MATLAB code such that Input array values outside the bounds of the array are computed by mirror-reflecting the array across the array border.

• graythresh: MATLAB code to find out Global image threshold using Otsu's method.

• bwmorph: MATLAB code for morphological operations on binary images.

TRANSFORMS AND ENHANCEMENT USED(cont.)

• skel: MATLAB code such that with n = Inf, removes pixels on the boundaries of objects but does not allow objects to break apart. The pixels remaining make up the image skeleton. This option preserves the Euler number.

• spur: MATLAB code that removes spur elements from the image matrix.

• bwarea(BW) is a MATLAB code that estimates the area of the objects in binary image BW. total is a scalar whose value corresponds roughly to the total number of on pixels in the image, but might not be exactly the same because different patterns of pixels are weighted differently.

• ComponentMeasurements[ Morphological Components[image] ,"Elongation"] is a Mathametica code to find out the length of individual component in an image.

FLOW PROCESS

Gaussian lowpass filtering & mirror-reflecting the array across the array border

graythresh

Skeleton transform

Spur reduction

FLOW PROCESS(cont.)

Component Measurements[Morphological Components[],"Elongation"]

bwperim

bwarea

Gives perimeter of binary image

Gives area of binary image

Gives length of binary image

CONCLUTION: Using software like MATLAB and Mathametica we have

successfully performed image enhancement and feature extraction of chromosomes like it’s medial axis construction and length measurement using medial axis, area and perimeter measurement have also been performed and the corresponding data obtained from individual images are normalized using maximum limit.

Now we are looking forward to next advanced stages of finding density profile, complete data set production and their normalization. After which we can train a ANN classifier to recognize each individual chromosome and classify them accordingly to a karyotype image.

On successful completion of project we can take forward the legacy of Alfredo Ruggeri which he had pioneered in the field of Bio-medical Imaging.

THANK YOU