“The best protection is early detection” Microcalcification identification Microcalcification identification in digital mammogram for early detection of breast cancer Masters -1 Presentation N hid Al Nashid Alam Registration No: 2012321028 [email protected]Supervisor: Dr. Mohammed Jahirul Islam Department of Computer Science And Engineering Shahjalal University of Science and Technology Tuesday, April 29, 2014
172
Embed
MICROCALCIFICATION IDENTIFICATION IN DIGITAL MAMMOGRAM FOR EARLY DETECTION OF BREAST CANCER
Digital mammogram has become the most effective technique for early breast cancer detection modality. Digital mammogram takes an electronic image of the breast and stores it directly in a computer. High quality mammogram images are high resolution and large size images. Processing these images require high computational capabilities. The transmission of these images over the net is sometimes critical especially if the diagnosis of remote radiologists is required. The aim of this study is to develop an automated system for assisting the analysis of digital mammograms. Computer image processing techniques will be applied to enhance images and this is followed by segmentation of the region of interest (ROI). Subsequently, the textural features will be extracted from the ROI. These texture features will be used to classify the microcalcifications as either malignant or benign.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
“The best protection is early detection”
Microcalcification identificationMicrocalcification identificationin digital mammogram for earlydetection of breast cancerMasters -1 Presentation
Department of Computer Science And EngineeringShahjalal University of Science and TechnologyTuesday, April 29, 2014
Introduction
Breast cancer:The most devastating and deadly diseases for women.
Steps to control breast cancer:W ill h i1) Prevention
2) Detection3) Diagnosis4) T t t
We will emphasis on :1) Detection2) Diagnosis
Computerize Breast cancer Detection System:
4) Treatment
o Computer aided detection (CADe) o Computer aided diagnosis (CADx) systems
Micro-calcification
Mammography
Mammogram
Micro-calcificationMicro‐calcifications :
Tiny deposits of calcium
Position:1. Can be scattered throughout the mammary gland orthe mammary gland, or 2. Occur in clusters.
They are caused by a number of reasons:y y
Aging ‐ The majority of diagnoses are made in women over 50
G ti I l i th BRCA1 (b t 1 l t) d BRCA2 (b t 2Genetic ‐ Involving the BRCA1 (breast cancer 1, early onset) and BRCA2 (breast cancer 2, early onset) genes
Depending on what pattern the micro‐calcifications present determines :p g p pThe future course of the action‐
I. Whether it be further investigatory techniques (as part of the triple assessment), or II. More regular screening
Mammography
Mammography :
Process of using low‐energyx‐rays to examine the human breast
Used as a diagnostic and a screening tool.
The goal of mammography :The early detection of breast cancer
Mammography Machine
USE:I. Viewing x‐ray imageII. Manipulate X‐ray image on a computer screen
Mammogram
Mammogram:Amammogram is an x‐ray picture of the breast
Use:Use:To look for changes that are not normal.
Result Archive:mdb226.jpg
Result Archive:The results are recorded on x‐ray film or directly into a computer
Types of mammograms:Types of mammograms:
I. Screening mammograms‐Done for women who have no symptoms of breast cancer.
II. Diagnostic mammograms ‐To check for breast cancer after a lump or other symptom or sign of breast cancer has been found.
III. Digital mammogram‐Uses x‐rays to produce an image of the breast. The image is stored directly on a computer.
Problem Statement
Problem StatementMain challenge :
QUICKLY AND ACCURATELY overcome the development of breast cancerQUICKLY AND ACCURATELY overcome the development of breast cancer
Reason behind the problem:Reason behind the problem:Burdensome Task Of Radiologist :
Eye fatigueHuge volume of images
Detection accuracy rate tends to decreaseNon‐systematic search patterns of humansPerformance gap between :
Specialized breast imagers andgeneral radiologists
Interpretational Errors:Si il h t i tiSimilar characteristics:
Abnormal and normal microcalcification
Problem Statement
The signs of breast cancer are: Masses CalcificationsCalcificationsTumorLesionL
Individual Research Areas
Lump
A k f h i i i lA key area of research activity involves :Developing better ways‐
Mammography Image Analysis Society (MIAS)Mammography Image Analysis Society (MIAS) ‐An organization of UK research groups
• Consists of 322 images‐‐ Contains left and right breast images for 161 patients
• Every image is 1024 X 1024 pixels in size
• Represents each pixel with an 8‐bit word
Internal Breast Structure
Image SegmentationGoal: Removing X‐ray Labeling And Pectoral muscles
Image Segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral musclesg y g
Why Segmentation?
Partitioning a digital image into multiple regions (sets of pixels).
GOAL OF SEGMENTATION:T l t bj t d b d i (li t ) i
(C) Final Segmented Image
• To locate objects and boundaries (lines, curves, etc.) in images.
• Result of image segmentationA set of regions that collectively cover the entire image (a)
(intensity <130) (intensity >200)
2
3
‐A set of regions that collectively cover the entire image. (a) ‐A set of contours extracted from the image. (C)
• Each of the pixels in a region(1, 2, 3) are similar with respect to some
1
(a) Segmentation PartEach of the pixels in a region(1, 2, 3) are similar with respect to some characteristic or computed property, such as color, intensity, or texture.
• Adjacent regions(1, 2, 3) are significantly different with respect to the h i i ( )
( ) g
same characteristic(s).
(b)Original image
Image Segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral musclesg y g
Proposed framework for breast profile segmentation
Image Segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral muscles
Plan of Action:
(intensity <130) (intensity >200)
g y g
Separating the Pectoral muscle
1. Original Image 2. Segmentation Part3. Final Segmented Image
Fit All i l i h iFit: All on pixels in the structuring element cover on pixels in the image
AC
g
Hit: Any on pixel in the structuring element covers an on pixel in the image
All morphological processing operations are based on these simpleAll morphological processing operations are based on these simple ideas
Structuring Elements Hits & Fits
Image Morphology Noise Removing
Structuring elements can be any size and make
Structuring Elements, Hits & Fits
Structuring elements can be any size and make any shape
However for simplicity we will use rectangularHowever, for simplicity we will use rectangular structuring elements with their origin at the middle pixelmiddle pixel
•The structuring element is moved across every pixel in the original image to give a pixel in a new processed image(very like spatial filtering)
•The value of this new pixel depends on the operation performed•The value of this new pixel depends on the operation performed
•There are two basic morphological operations:
Erosion and Dilation
Structuring Elements, Hits & Fits
Noise Removing 1. Morphological Analysis
ErosionStructuring Elements, Hits & Fits
Erosion of image f by structuring element s is given by f sThe structuring element s is positioned with its origin at (x, y) and the new pixel value is g ( , y) pdetermined using the rule:
⎧ fiif1 f
⎩⎨⎧
=otherwise0
fitsif1),(
fsyxg
⎩A morphological opening of an image is an erosion followed by a dilation
After Removing Some NoiseImage Containing Noise(mdb041.jpg)(mdb041.jpg)
Noise Removing
Chosen Technique 2D MEDIAN FILTERING FOR SALT AND PEPPER NOISEChosen Technique 2D MEDIAN FILTERING FOR SALT AND PEPPER NOISE
After Removing Some NoiseImage Containing Noise(mdb041.jpg)
I = medfilt2(I, [1 5]); Median filtering is a nonlinear operation often used in image processing to reduce "salt and pepper" noise. A median filter is more effective than convolution when the goal is to simultaneously reduce noise and preserve edges. Since all the mammograms are in high quality images, there is no need to perform median filtering
Why choosing?
1. Morphological AnalysisOVER
2 Image Smoothing/Filtering(Low pass):
OVER
2.Image Smoothing/Filtering(Low pass):
‐Does not work will on all the image [I = medfilt2(I, [1 5]);] •No effect most of the time•Absence of salt and peeper noise
‐Tendency of loosing interesting detailsTendency of loosing interesting details
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral muscles
Class: Benign
g y g
mdb212 150 200mdb214 150 200 00
) (a)Original image
mdb214 150 200mdb218 150 210mdb219 150 210db222 150 210 in
tensity
>20
2
3
mdb222 150 210mdb223 150 210mdb226 150 210
sity <130) (i
1
(b) Segmentation Part
mdb227 150 210mdb236 150 210mdb240 150 210
(inten
mdb248 150 210mdb252 140 210
(C) Final Segmented Image
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral muscles
(e)Image containing
g y g
Achievement: X‐Ray Label removedClass: Benign
(b)Segmentation Part (c) Final Segmented Image(a)Main Image (e)Image containing Duct, Lobules, Sinus
(d)Binary Image
mdb236.jpg
mdb001.jpg
mdb254.jpg
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral musclesg y g
Issues 1.Biggest Cluster Does Not Contain BreastProduce Artifacts In Pectoral muscle And Breast Region
Class: Benign(b)Segmentation Part (c) Final Segmented Image(a)Main Image (e)Image containing
only Pectoral muscle(d)Binary Image
Class: Benign
mdb212 jpg
What we need
mdb212.jpg
What we have
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral muscles
( )I C i i
g y g
Issues 1.Biggest Cluster Does Not Contain BreastProduce Artifacts In Pectoral muscle And Breast Region
Class: Benign(b)Segmentation Part (c) Final Segmented Image(a)Main Image
(e)Image Containing Only Pectoral muscle(d)Binary Image
Class: Benign
mdb214.jpg
mdb001.jpg
mdb218.jpg
What We Need What We Have
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral muscles
(e)Image containing duct, lobules,
g y g
(d)Binary Image
Issues 1.Biggest Cluster Does Not Contain BreastProduce Artifacts In Pectoral muscle And Breast Region
Class: Benign
(b)Segmentation Part (c) Final Segmented Image(a)Main Image(
sinus & Pectoral muscle(d)Binary Image
mdb222.jpg
ntW
mdb001.jpg
What w
e wa hat w
e Hav
mdb223jpg Wve
mdb226jpg
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral musclesg y g
(e)Image containing duct lobules
Issues 1.Biggest Cluster Does Not Contain BreastProduce Artifacts In Pectoral muscle And Breast Region
Class: Benign
(b)Segmentation Part (c) Final Segmented Image(a)Main Image (d)Binary Image(e)Image containing duct, lobules,
sinus & Pectoral muscle
mdb240.jpg
mdb001.jpg
mdb248 jpg at we want W
hat we H
mdb248.jpg
Wha
Have
mdb252.jpg
Image segmentation K‐means Clustering
Goal: Removing X‐ray Labeling And Pectoral muscles
Class: Malignant
g y g
mdb209 140 210mdb211 140 210 10
) (a)Original imagemdb209.jpg
(a) Original imagemdb213 140 210mdb216 140 210mdb231 140 210
X-ray Label Removing Finding The Big BLOBPl f A ti
1.Binarizatin of original image.Plan of Action:
(threshold luminance level‐=0.5)
Original image Binary Image
2.Find the biggest blob.
( )
mdb219.jpg
(a) Artifacts (Hole) in ROI
Original image Binary Image
Label successfully removed Issues
y g
(b)Absence of Ligaments and fatty tissue
mdb231.jpgmdb253.jpg
(c) Absence of pectoral muscles(c) Absence of pectoral muscles
X-ray Label Removing Finding The Big BLOBClass: Benign Issue with fatty tissues and ligaments existenceOriginal image Binary Image
(threshold luminance level‐=0.5) Original image Binary Image(threshold luminance level‐=0.5)
db212 j mdb219.jpgmdb212.jpg mdb219.jpg
mdb214.jpgmdb222.jpg
mdb218.jpg mdb223.jpg
X-ray Label Removing Finding The Big BLOBClass: Benign Issue with fatty tissues and ligaments existenceOriginal image Binary Image
(threshold luminance level‐=0.5) Original image Binary Image(threshold luminance level‐=0.5)
mdb226.jpg mdb240.jpg
mdb227 jpg mdb248 jpgmdb227.jpg mdb248.jpg
mdb236.jpg mdb252.jpg
X-ray Label Removing Finding The Big BLOBIssue with fatty tissues and ligaments existenceClass: Malignant
Original imageBinary Image
(threshold luminance level‐=0.5) Original image Binary Image(threshold luminance level‐=0.5)
mdb209.jpg mdb216.jpg
mdb211.jpg mdb231.jpg
mdb213.jpgmdb233.jpg
X-ray Label Removing Finding The Big BLOBIssue with fatty tissues and ligaments existenceClass: Malignant
Original image Binary Image(threshold luminance level‐=0.5) Original image Binary Image
(threshold luminance level‐=0.5)
mdb245.jpg
mdb238.jpg
mdb249.jpg
mdb239.jpg
mdb253.jpg
mdb241.jpg mdb256.jpg
X-ray Label Removing
Issue With Fatty Tissues And Ligaments Existence
Moving towards solution
y g
X-ray Label RemovingPl f A tiPlan of Action:
1.Binarize the image
2.Fill inside the hole region of the binary image
3.Finding the largest Blob:function [outim] = bwlargestblob( im,connectivity)if size(im,3)>1,error('bwlargestblob accepts only 2 dimensional images');dend
4.Keep the Largest Blob and discard other blobs(to remove X-ray level)
Image MorphologyGoal: Region filling(Region inside the blob)
X-ray Label Removing
Experimental results:
g f g( g )
Original imageFinding biggest blob(Level removed)
Hole fillingInside the blob(dialation)
Result image(Label Removed)Binary image
Direct Binarisation Without Image enhancement
g g (Level removed) Inside the blob(dialation) ( )y g
mdb240.jpg
mdb219.jpg
mdb231.jpg
Image Morphology X-ray Label RemovingGoal: Region filling(Region inside the blob)g f g( g )
Original imageResult image
(Label Removed)Binary image
Direct Binarisation Without Image enhancement Experimental results:
g g ( )y g
mdb240.jpg
Issues
1.Does not always produce appealing output
mdb219.jpgappealing output
2.Some details are missing(Details around Edge region )
mdb231.jpg
Image Morphology X-ray Label RemovingGoal: Region filling(Region inside the blob)g f g( g )
Original image Result image (Label Removed)
Direct Binarisation Without Image enhancement Experimental results:
g g
mdb240.jpg
mdb219 jpg
Issues
1 Does not always produce
mdb212.jpg
mdb219.jpg 1.Does not always produce appealing output
2.Some details are missing
mdb214.jpg
mdb231.jpg
(Details around Edge region )mdb219.jpg
mdb226.jpg
X-ray Label Removing
To Achieve The Desired Final Result:
-ApplyA Range Of Techniques on original image
-To find largest blobUse -Otsu’s thresholding technique (graytrash) [9]
-Finding Bi-level the image(im2bw)
[9] Otsu, N., "A Threshold Selection Method from Gray‐Level Histograms," IEEE Transactions on Systems, Man, and Cybernetics, Vol. 9, No. 1, 1979, pp. 62‐66.
X-ray Label Removing Plan of Action
1. Histogram equalization of the original X-ray image2 Adjust image contrast2. Adjust image contrast3. Apply Otsu's Thresholding Method [9] and
fi d bi l l th i hi h h l bl b i itfind bi-level the image which has several blobs in it. 4. Finding the Largest blob (Bwlargest.bolb)5. Hole filling within the blob region6. Keep the true pixel value covering only the area of largest
blob and discard other features from the original image7. X-ray label is successfully removed
[9] Otsu, N., "A Threshold Selection Method from Gray‐Level Histograms," IEEE Transactions on Systems, Man, and Cybernetics, Vol. 9, No. 1, 1979, pp. 62‐66.
X-ray Label Removing Combining Range of techniquesJ = histeq(I); %histogram equalizationJ = histeq(I); %histogram equalization
Removing pectoral muscleKeeping fatty tissues and ligaments
mdb214.jpg
Main ImageMain Image
Result Image
Extraction of ROIRemoving pectoral muscle
Why removing pectoral muscle?
o Pectoral muscle will never contain micro‐calcification
o Less Computational Time And Costo Less Computational Time And Cost‐Operation on small image area
o Fat ty t i s sue areao D u c to Fat ty t i s sue areao D u c t
Existence of micro‐calcification:
o D u c to Lobule so Sinus
l i t
o D u c to Lobule so Sinus
l i to l i g a m e nt s o l i g a m e nt s ROI
Edge Detection of pectoral muscleRemoving pectoral muscle
P ibl A h T Ed d iPossible Approach To Edge‐detection:
1.Scanning pixel value intensity at each points2.find out the sudden big intensity change at the edge location
Approach‐01:
3.Mark the pixels at edge location4.Estimate a straight line depending on the marked edge points
Problem faced in Approach‐01:
‐Finding appropriate Thresholding value in an unsupervised method, which will work on every image ‐The threshold value must be found in an unsupervised mannerA d fi d h h ld l ill d d i d f ll i‐Any predefined threshold value will not produce desired output for all image
Edge Detection of pectoral muscleRemoving pectoral muscle
P bl f d ( h )Problems faced in (Approach‐02):
3 S h h ldi l (i 130 210 ) d k ll ll h i d3.Same thresholding value(i.e.,130‐210,) does not work well on all the images and Produce improper output(complete black image as output)
Edge Detection of pectoral muscleRemoving pectoral muscle
Points to be noted from approach-2:(2)Binary Image(1)Original Image
‐Pectoral muscle a Triangular areamdb212.jpg
Based on this point: M i t h 03
mdb214.jpg
Moving on to approach ‐03
mdb209.jpg
Triangle Detection of pectoral muscleRemoving pectoral muscle
1 Fi h i l f h l l i
Approach‐03(Triangle Detection of pectoral muscle):
1.Fing the triangular area of the pectoral muscle region
I. Finding white seeding pointII Finding the 1st black point of 1st row after getting a white seeding pointII. Finding the 1st black point of 1st row after getting a white seeding pointIII. Draw a horizontal line in these two points.IV. finding the 1st black point of 1st column after getting a white seeding pointV Draw a vertical line and angular lineV. Draw a vertical line and angular line.
2.Making all the pixels black(zero)resides in the pectoral muscle area
Visualization in next slide
Triangle Detection of pectoral muscleRemoving pectoral muscle
Approach 03(Triangle Detection of pectoral muscle):Approach‐03(Triangle Detection of pectoral muscle):stratching_in_range=uint8(imadjust(I,[0.01 0.7],[1 0]));
BW=~stratching_in_range;
mdb212.jpg1.Original image
2.Contrast stretching
3.Binary of contrast image
Triangle Detection of pectoral muscleRemoving pectoral muscle
Approach 03(Triangle Detection of pectoral muscle):Approach‐03(Triangle Detection of pectoral muscle):
6.muscle removed
5.Triangle Filled
4.Triangle
Removing pectoral muscleApproach 03(Triangle Detection of pectoral muscle):
Triangle Detection of pectoral muscle
Approach‐03(Triangle Detection of pectoral muscle):
E l lExperimental results
Triangle Detection of pectoral muscleRemoving pectoral muscle
Extraction of ROIRemoving pectoral muscleA BP lA BPectoral
muscle
Eliminate:‐Unexposed X‐ray portion (left side)‐top left most pixel is a pectoral muscle
Determine:Skin air boundary
hhRegion of Interest
‐Skin‐air boundary‐ Region of interest(ROI)
Result:Result:‐ROI(ABCDA) includes complete pectoral muscle
D C ROI(ABCDA)
Removing pectoral muscle Extraction of ROI
Image # mdb226.jpg1. ORIGINAL IMAGE 2.BINARY IMAGE
3 Blue Circle indicates the3. Blue Circle indicates the True ‘Skin‐air boundary’ in
the binary image
4. True ‘Skin‐air boundary
5. Extraction of ROI
Removing pectoral muscle Extraction of ROI
E l lExperimental results
Removing pectoral muscle Benign Extraction of ROI
1 Original image 2 Binary image 3 Blue Circle indicates the True ‘Skin‐air 4. True ‘Skin‐air boundary 5. Extraction of ROI
mdb212 jpg
1.Original image 2.Binary image 3. Blue Circle indicates the True Skin air boundary in the binary image
y
mdb212.jpg
mdb214 jpgmdb214.jpg
db218 jmdb218.jpg
mdb219 jpgmdb219.jpg
mdb222.jpgmdb222.jpg
mdb223.jpg
Removing pectoral muscle Extraction of ROIBenign
3 Blue Circle indicates the True ‘Skin air 4 Tr e ‘Skin air bo ndar 5 E t ti f ROI1.Original image 2.Binary image3. Blue Circle indicates the True Skin‐air
boundary in the binary image4. True ‘Skin‐air boundary 5. Extraction of ROI
mdb226.jpg
mdb227.jpg
mdb240.jpg
mdb248.jpg
mdb252.jpg
Removing pectoral muscle Malignant Extraction of ROI
1 Original image 2 Binary image3. Blue Circle indicates the True ‘Skin‐air 4. True ‘Skin‐air boundary 5. Extraction of ROI
mdb209.jpg
1.Original image 2.Binary image boundary in the binary image
db 09.jpg
mdb211.jpgjpg
mdb213.jpgjpg
mdb216.jpg
mdb231.jpg
mdb233.jp
Removing pectoral muscle Extraction of ROI
1 Original image 2 Binary image3. Blue Circle indicates the True ‘Skin‐air
4 True ‘Skin air boundary 5 Extraction of ROI
Malignant
1.Original image 2.Binary image boundary in the binary image 4. True Skin‐air boundary 5. Extraction of ROI
mdb238.jpg
mdb239.jpg
mdb241.jpg
mdb245.jpg
Removing pectoral muscle Extraction of ROIMalignant
1.Original image 2.Binary image3. Blue Circle indicates the True ‘Skin‐air
boundary in the binary image4. True ‘Skin‐air boundary 5. Extraction of ROI
mdb249.jpg
mdb253.jpg
mdb256.jpg
Removing pectoral muscle Extraction of ROI
Extraction of ROISuccessful
Extraction of ROImdb212.jpg
mdb214.jpg
mdb222.jpg
mdb218.jpgmdb223.jpg
Removing pectoral muscleOverview of the proposed method
%#Sobel mask for x‐direction:Gx=((C(i+2,j+2)+2*C(i+1,j+2)+C(i,j+2))‐(C(i+2,j)+2*C(i+1,j)+C(i,j)));%#Sobel mask for y‐direction:Gy=((C(i+2,j)+2*C(i+2,j+1)+C(i+2,j+2))‐(C(i,j)+2*C(i,j+1)+C(i,j+2)));
Define SOBEL MASK PAIR (Gx, Gy)to convolve with the original image
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI GradientB(i,j)=sqrt(Gx.^2+Gy.^2);……………………………..%#mean filtering to smooth ROI gradient(B)kernel = ones(3, 3) / 9; %# 3x3 mean kernel
( )Filtered_ROI_Gradient = conv2(B, kernel, 'same'); % Convolve keeping size of I;Mean filtering is usually thought of as a convolution filter.
3.Smoothing SOBEL ROI(ROI Gradient):
Mean filter 3*3 is usedMean filter :
Smoothes Local VariationReduce Noise
The FILTERED image is called ROI Gradient.
ROI Gradient image is used as input toROI Gradient image is used as input to The Watershed Transform.
(d)Sobel ROI (d)Filtered ROI Gradient
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Gradient Operation
%#The gradient of the image/magnitude%#B(i,j)=abs(Gx)+abs(Gy); %#To avoid complex computation, the gradient can also be computed using
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
(b)Horizontal Gradient
mdb212.jpg.
(a)Original Image (c)Vertical Gradient (d)ROI Gradient(b+c) (d)Filtered ROI Gradient
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
E l lExperimental results
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Benign Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image 2. Extraction of ROI (b)Horizontal Gradient (c)Vertical Gradient (d)ROI Gradient(b+c) (d)Filtered ROI Gradient
mdb212.jpg
db214 jmdb214.jpg
mdb218.jpg
mdb219.jpg
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Benign Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image 2. Extraction of ROI (b)Horizontal Gradient (c)Vertical Gradient (d)ROI Gradient(b+c)(d)Filtered ROI
Gradient
mdb222.jpg
mdb223.jpg
mdb226.jpg
mdb227.jpg
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Benign Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image 2. Extraction of ROI (b)Horizontal Gradient (c)Vertical Gradient (d)ROI Gradient(b+c)(d)Filtered ROI
Gradientg g 2. Extraction of ROI ( ) ( ) ( ) ( )
mdb240.jpg
mdb248.jpg
mdb252.jpgmdb252.jpg
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Malignant Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image 2. Extraction of ROI (b)Horizontal Gradient (c)Vertical Gradient (d)ROI Gradient(b+c)(d)Filtered ROI
Gradient
mdb209.jpg
mdb211 jpgmdb211.jpg
mdb213.jpg
mdb216.jpg
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Malignant Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image 2. Extraction of ROI (b)Horizontal Gradient (c)Vertical Gradient (d)ROI Gradient(b+c)(d)Filtered ROI
Gradient
mdb231.jpg
mdb233.jpg
mdb238.jpg
mdb239.jpg
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Malignant Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image 2. Extraction of ROI (b)Horizontal Gradient (c)Vertical Gradient (d)ROI Gradient(b+c)(d)Filtered ROI
Gradient
mdb241.jpg
mdb245.jpg
mdb249.jpg
mdb253.jpg
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Malignant Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
1.Original image
2 Extraction of ROI2. Extraction of ROI
3.Horizontal Gradient
4.Vertical Gradient
5.ROI Gradient(b+c)
6.Filtered ROI Gradient
Removing pectoral muscleGradient Operation SOBEL GRADIENT ROI Gradient
Malignant Gradient Operation
Gradient Operation SOBEL GRADIENT, ROI Gradient
Watershed Transformation
Gradient Operation
S thi Filt
Gradient Image
Smoothing Filter
Watershed Transformation
Filtered Image
Watershed Transformation
Removing pectoral muscle WatershedTransformation
Removing pectoral muscle WatershedTransformation
Th t h d t f fi t d b Di b l d L t j l ([10]The watershed transform was first proposed by Diagabel and Lantuejoul.([10] A region‐based segmentation approach from the field of mathematical morphology, and a well‐organized survey of its different definitions and algorithms can be found in the work of Roerdink and Meijster[11]j [ ]
The concept of watershed transform can be realized by visualizing the ROI gradient as a topographic surface, such that the gray value of each pixel defines its altitude.
A hole is pierced in each regional minimum which allows water to gradually rise in catchment basins;
Each basin is evolved from a regional minimum.
When any two catchment basins are about to merge, a dam is built between them to prevent them from mergingprevent them from merging.
When water reaches the highest peak of the landscape, the flooding process is stopped.
Finally, several catchment basins divided by dams (otherwise called watersheds) are evolved. In terms of image segmentation, these catchment basins represent different regions, and watersheds are the boundaries between these regions.
Removing pectoral muscle WatershedTransformation
1 Filtered ROI Gradient
2.Watershed linesobtained from image SOBEL_gradient (wr)
1.Filtered ROI Gradient
3 internal markers3.internal markers(light grayshed regions inside catchment basins)
Our proposed method consists of three main modules:) i f l f1) Basic of wavelet transform
2) Micro‐calcifications detection 3) Evaluation of detection methods)
1 Basic of wavelet transform:
Future plan1.Basic of wavelet transform:
-Wavelet analysis is an extremely powerful data representation method
-Allows :The separation of images into frequency bands without affecting the spatiale sep o o ges o eque cy b ds w ou ec g e splocality Bouyahia et al. [12].
-Makes use of :Makes use of :two separate bases for analysis and synthesis.
Information Extraction:-Information Extraction:localized high frequency signals such as micro-calcifications could be extracted
Th t di i l l t t f ill b hi d-The two dimensional wavelet transform will be achieved:By implementing a bank of one-dimensional low-pass and high-pass analysis filters.
1 wavelet transform:
Future plan1.wavelet transform:
For one level of decomposition:-For one level of decomposition:The image will be decomposed into four orthogonal sub-bands:LL, HL, LH, and HH.
ill d di i f b dwill correspond to distinct frequency bands.
1 B i f l t t f
Future plan1.Basic of wavelet transform:
The HL sub band will contain :-The HL sub-band will contain :Horizontal oriented features.
-The LH sub-band will contain:Vertically oriented structures
-The HH sub-band will contains :Diagonal structures.
-The LL sub-band will be :The low-pass filtered version of the imageThe low pass filtered version of the image
and will further be decomposed
1 Basic of wavelet transform:
Future plan1.Basic of wavelet transform:
-Collection of sub-images will form:A multi resolution representationp
Multi resolution representation will organize:-Multi resolution representation will organize:The image into a set of details appearing
at different resolutions.
2 Micro calcifications detection:
Future plan2. Micro‐calcifications detection:‐Full resolution will be maintained
during the multi‐resolution analysis by using wavelet transform.
‐The wavelet transform will be operated :without down‐sampling and up‐sampling in respectively the analysis and synthesis
computations.computations.
‐This will ensure:translation invariance and
implies –a finer sampling rate of the wavelet decompositiona vital requirement during small object detection such as micro‐calcifications.
‐The redundant transform will be applied:In each pixel of the image.
The size of each sub‐band will be:The same as the original image.
2 Micro calcifications detection:
Future plan2. Micro‐calcifications detection:
Three levels redundant wavelet decomposition of the image will be performed with bi‐orthogonal daubechies wavelet , daubechies et al. [13]
The wavelet decomposition is performed after an enhancement Step Bouyahia et al. [12]
First level detail coefficients will contain mostly noise.First level detail coefficients will contain mostly noise.
Detail coefficients in level 2 and 3 will contain fine breast structure and micro‐calcifications.
Af d i i f h i h l f b b d ill b ( hAfter decomposition of the image, the low‐frequency sub‐band will be set to zero (the micro‐calcifications will appear in the high‐frequency sub‐bands).
An adaptive thresholding will be performed to detect micro‐calcifications. p g p
After wavelet Decomposition, we will determine the maximum value in each sub‐band.
We will threshold the detail coefficients of each sub band with the corresponding thresholdWe will threshold the detail coefficients of each sub‐band with the corresponding threshold and perform the reconstruction of the image.
The process will be iterated by varying the thresholds with logarithmic way.
Future plan3) Evaluation of detection:
Simulations will be operated on Mini‐Mammographic Image Analysis Society (MIAS) database
The results will be presented and compared to some relative works.
We will show that the proposed approach will competitive with the best of the state of the artthe best of the state of the art.
REFERENCES
[1]K. Santle Camilus , V. K. Govindan, P.S. Sathidevi,” Pectoral muscle identification inmammograms”, Journal of Applied Clinical Medical Physics , Vol. 12 , Issue No. 3 , 2011
[2]D.Narain Ponraj, M.Evangelin Jenifer, P. Poongodi, J.Samuel Manoharan “A Survey on the Preprocessing Techniques of Mammogram for the Detection of Breast Cancer”, Journal ofEmerging Trends in Computing and Information Sciences, Volume 2, Issue 12, pp. 656‐664
[3]Arnau Oliver, Albert Torrent, Meritxell Tortajada, Xavier Llad´o,Marta Peracaula,d d l ´ d d ” l f d lLidia Tortajada, Melcior Sent´ıs, and Jordi Freixenet,” Automatic microcalcification and clusterdetection for digital and digitised mammograms”, Springer‐Verlag Berlin Heidelberg, 36,pp. 251–258, 2010
REFERENCES
[4]Arnau Olivera, Albert Torrenta , Xavier Lladóa , Meritxell Tortajada, Lidia Tortajadab,Melcior Sentísb, Jordi Freixeneta, Reyer Zwiggelaarc,” Automatic microcalcification and clusterdetection for digital and digitised mammograms”,Elsevier:Knowledge‐Based Systems, Volume 28,pp. 68–75, April 2012.
[5]A.Papadopoulos, D.I . Fotiadis, L.Costrrido,” Improvement of microcalcification clusterdetection in mammogaphy utilizing image enhancement techniques”.Comput.Bio.Med.10,Vol 38,Issue 38,pp.1045‐1055,2008
[6]N.R.Pal,B.Bhowmik, S.K.Patel, S.Pal, J.Das,”A multi‐stage nural network aided system fordetection of microcalcification in digitized mammogeams”,Neurocomputing, Vol 11,pp.2625‐2634,2008
[7]M.Rizzi, M.D’Aloia, B.Castagnolo,” Computer aided detection of microcalcification in digitalM d i l d i i ” I C Aid d E V l 16 I 2Mammograms adopting a wavelet decomposition ”,Integr.Comput.‐Aided Eng.,Vol 16,Issue 2,pp.91‐103,2009
REFERENCES
[8]S.N.Yu, Y.K. Huang,” Detection of microcalcifications on digital mammograms using combinedModel based and statistical textural features” Expert Syst Appl Vol 37 Issue 7 pp 5461 5469Model‐based and statistical textural features”, Expert Syst.Appl. , Vol 37,Issue 7,pp.5461‐5469,2010
[9] Otsu N "A Threshold Selection Method from Gray Level Histograms " IEEE Transactions on[9] Otsu, N., A Threshold Selection Method from Gray‐Level Histograms, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 9, No. 1, 1979, pp. 62‐66.
[10] Digabel H, Lantuejoul C. Iterative algorithms. In: Proceedings Actes du Second Symposium Europeen d’Analyse Quantitative des Microstructures en Sciences des Materiaux, Biologie et Medecine. Stuttgart: Riederer Verlag; 1977. p. 85–99.
[11]Roerdink JBTM and Meijster A. The watershed transform: definitions, algorithms and parallelization strategies. Fundamenta Informaticae. 2000;41:187–228.
REFERENCES
[12]Bouyahia S, Mbainaibeye J, Ellouze N. 2005: Wavelets and Wavelet packets for mammography 3rd International Conference: Sciences of Electronic Technologies of
[13]Daubechies I 1992: Ten lectures on wavelets SIAM Philadelphia
mammography. 3rd International Conference: Sciences of Electronic, Technologies of Information, and Telecommunications, SETIT 2005, Sousse, Tunisia, March, 27‐31.
[13]Daubechies I 1992: Ten lectures on wavelets, SIAM, Philadelphia.