Mammography Image Feature extraction

I'm trying to extract texture features of a tumor (ROI) for mammography classification.

The features I'm trying GLCM features. I tried two codes, but the results of classification is not good. The classifiers are working with other mammography data, so the problem is the extracted features.

So, i need to know what is wrong through these steps?

The first code i tried is the code in this link:

https://www.mathworks.com/matlabcentral/fileexchange/22187-glcm-texture-features?focused=5107518&tab=function

and called the function by the following:

    GLCM = graycomatrix(ROI,'Offset',[0 1;-1 1;-1 0;-1 -1]);
    [stats,feat] = GLCM_Features1(GLCM,0);
    GLCM_Diff_Cal1(l,:)=feat;

Another code I tried is:

GLCM = graycomatrix(ROI,'Offset',[0 1;-1 1;-1 0;-1 -1]);

stats=graycoprops(GLCM2,{'Contrast','Homogeneity','Correlation','Energy'});

contrast=(stats.Contrast);

en=(stats.Energy);

co=(stats.Correlation);

hom=(stats.Homogeneity);

m=mean(mean(ROI));

s=std2((ROI));

f1=[m s hom co en contrast];

J2=SegmentedROI;

I=J2;

J2 = uint8(255 * mat2gray(J2));

J3=edge(J2,'log');

bw=bwareaopen(J2,150);                          

bwp=edge(bw,'sobel');

geometric.area=sum(sum(bw));

geometric.peri=sum(sum(bwp));

geometric.compact=geometric.peri^2/geometric.area;

count=1;

roih=0;

for i=1:size(bw,1);

      for j=1:size(bw,2);

          if bw(i,j)

              roih(count,1)=double(SegI(i,j)); 

              count=count+1;
          end
      end
  end

nh=roih/geometric.area;

texture.mean=mean(roih);

texture.globalmean=texture.mean/mean(mean(double(I)));

texture.std=std(roih);

texture.smoothness=1/(1+texture.std^2);

texture.uniformity=sum(nh.^2);

texture.entropy=sum(nh.*log10(nh));

texture.skewness=skewness(roih);

texture.correlation=sum(nh'*roih)-texture.mean/texture.std;

x1=[];

y1=[];

z1=[];

[gradI3, Gdir] = imgradient(I,'sobel');

gradJ3=medfilt2(gradI3);

gradJ2=uint8(255 * mat2gray(gradJ3));

gradJ3=edge(gradJ2,'log');

gradbw=bwareaopen(gradJ2,250);

gradbwp=edge(gradbw,'sobel');

count=1;

for i=1:size(gradbw,1);

      for j=1:size(gradbw,2);

          if gradbw(i,j)

              gradroih(count,1)=double(gradI3(i,j)); 

              count=count+1;
          end
      end
  end

gradroih(gradroih==0)=[];

gradnh=gradroih/geometric.area;

gradient.mean=mean(gradroih);

gradient.globalmean=texture.mean/mean(mean(double(gradI3)));

gradient.std=std(roih);

gradient.smoothness=1/(1+gradient.std^2);

gradient.uniformity=sum(gradnh.^2);

gradient.entropy=sum(gradnh.*log10(gradnh));

gradient.skewness=skewness(gradroih);

gradient.correlation=sum(gradnh'*gradroih)-gradient.mean/gradient.std;

geometric

texture

gradient

xy=[geometric.area,geometric.peri,geometric.compact];

x1=[x1;xy];

yy=[texture.mean,texture.globalmean,texture.std,texture.smoothness,texture.uniformity,...
    texture.entropy,texture.skewness,texture.correlation];
y1=[y1;yy];

zy=[gradient.mean,gradient.globalmean,gradient.std,gradient.smoothness,gradient.uniformity,...
    gradient.entropy,gradient.skewness,gradient.correlation];

z1=[z1;zy];

f2=[x1 y1 z1];

[Image Analyst]

http://www.mathworks.com/matlabcentral/answers/13205#answer_18099

[Mohamed Ahmed]

I applied the question providing the

1- All used images if needed for trials and illustrated what each image refers to and which one is used for each step.

2- I provided a picture for my main requirements from the uploaded picture

This is the link of the question https://www.mathworks.com/matlabcentral/answers/350752-texture-feature-extraction-from-a-mammography-image

[Ali Zulfikaroglu]

Hi I am stuying same subject. I am using some feature extraction codes too. Resemble yours.

But my ANN classification results is so bad . Can you help me if you find how to solve it

Thank you from now.

[AHT.fatima]

Hi I am stuying same subject. Can you please share the code for pectoral muscle removal from the image please. It would be of great help to me.

Answer 1

[AHT.fatima]

Hi I am stuying same subject. Can you please share the code for pectoral muscle removal from the image please. It would be of great help to me