Classifying new image after PCA in SVM

I have 32 breast tissue images where 16 are benign and 16 are malignant. I have applied PCA for dimension reduction. I have applied SVM for classification. Problem arises when I try to find the label of a new input image. I am getting an error message. Please find my code and error message below.

clc;clear;close all

%% Getting Image

for k = 1 : 32
    jpgFileName = strcat('B', num2str(k), '.jpg');
	if isfile(jpgFileName)
		i = imread(jpgFileName);
	else
		fprintf('File %s does not exist.\n', jpgFileName);
    end




% if image is rgb
try
    i=rgb2gray(i);
end

%% Crop The Breast

z=im2bw(i,0.1);
figure(2)
imshow(z);title('Original B&W')
info=regionprops(z);
a=cat(1,info.Area);
[m,l]=max(a);
X=info(l).Centroid;
bw2=bwselect(z,X(1),X(2),8);
i=immultiply(i,bw2);
figure(3)
imshow(i);
title('Getting the Breast and Muscle')

%% Deleting Black Ground

% We will delete the black corners
% So that we can select the muscle
% using bwselect
% convert to B&W first time
[x,y]=size(z);
tst1=zeros(x,y);

% detect empty rows
r1=[];
m=1;
for j=1:x
    if z(j,:)==tst1(j,:)
        r1(m)=j;
        m=m+1;
    end
end

% detect empty columns
r2=[];
m=1;
for j=1:y
    if z(:,j)==tst1(:,j)
        r2(m)=j;
        m=m+1;
    end
end

% Deleting
i(:,r2)=[];
i(r1,:)=[];

figure(4)
imshow(i);title('after deleting background');


%% Deleting the Muscle

if i(1,1)~=0
    c=3;
    r=3;
else
    r=3;
    c=size(i,2)-3;
end

z2=im2bw(i,0.5);
bw3=bwselect(z2,c,r,8);
bw3=~bw3;
ratio=min(sum(bw3)/sum(z2));
if ratio>=1
    i=immultiply(i,bw3);
else
    z2=im2bw(i,0.75);
    bw3=bwselect(z2,c,r,8);
    ratio2=min(sum(bw3)/sum(z2));
    if round(ratio2)==0
        lvl=graythresh(i);
        z2=im2bw(i,1.75*lvl);
        bw3=bwselect(z2,c,r,8);
        bw3=~bw3;
        i=immultiply(i,bw3);
    else
        bw3=~bw3;
        i=immultiply(i,bw3);
    end
end
figure(5)
imshow(i)
title('Getting only the Breast')


%% Adaptive Median Filter
% clc;
% clear;
% close all;
% a=imread('M1.jpg');
% b=rgb2gray(i);
J = imnoise(i,'salt & pepper', 0.02);
NoisyImage=J;
[R C P]=size(NoisyImage);
OutImage=zeros(R,C);
figure;
% imshow(J);
Zmin=[];
Zmax=[];
Zmed=[];

for i=1:R
    for j=1:C
               if (i==1 & j==1)
                   
                   
                        
                        
          % for right top corner[8,7,6]
                elseif (i==1 & j==C)
                        
                        
			% for bottom left corner[2,3,4]
                elseif (i==R & j==1)
                        
                        
        		     % for bottom right corner[8,1,2]

                elseif (i==R & j==C)
                        
                        
        		    %for top edge[8,7,6,5,4]

                elseif (i==1)
		      			 	
		      			 
                     		      			% for right edge[2,1,8,7,6]

                elseif (i==R)
		      			 
		      			% // for bottom edge[8,1,2,3,4]
		      			 
                elseif (j==C)
		      			 
						 
                         %// for left edge[2,3,4,5,6]
						 
                elseif (j==1)
                    
						 
                            
                         
						 
               else
                    
                   
						 	
                             SR1 = NoisyImage((i-1),(j-1));
 							 SR2 = NoisyImage((i-1),(j));
    						 SR3 = NoisyImage((i-1),(j+1));
 							 SR4 = NoisyImage((i),(j-1));
                             SR5 = NoisyImage(i,j);
 							 SR6 = NoisyImage((i),(j+1));
 							 SR7 = NoisyImage((i+1),(j-1));
 							 SR8 = NoisyImage((i+1),(j));
 							 SR9 = NoisyImage((i+1)),((j+1));
                             TempPixel=[SR1,SR2,SR3,SR4,SR5,SR6,SR7,SR8,SR9];
                             Zxy=NoisyImage(i,j);
                             Zmin=min(TempPixel);
                             Zmax=max(TempPixel);
                             Zmed=median(TempPixel);
                             A1 = Zmed - Zmin;
                             A2 = Zmed - Zmax;
                         
                             if A1 > 0 && A2 < 0
                             
                                  %   go to level B
                                  B1 = Zxy - Zmin;
                                  B2 = Zxy - Zmax;
                                  if B1 > 0 && B2 < 0
                                      OutImage(i,j)= Zxy;
                                  else
                                      OutImage(i,j)= Zmed;

                                  end
                             else
                                 
                                 if ((R > 4 && R < R-5) && (C > 4 && C < C-5))
                                 
                                 S1 = NoisyImage((i-1),(j-1));
                                 S2 = NoisyImage((i-2),(j-2));
                                 S3 = NoisyImage((i-1),(j));
                                 S4 = NoisyImage((i-2),(j));
                                 S5 = NoisyImage((i-1),(j+1));
                                 S6 = NoisyImage((i-2),(j+2));
                                 S7 = NoisyImage((i),(j-1));
                                 S8 = NoisyImage((i),(j-2));

                                 S9 = NoisyImage(i,j);
                                 S10 = NoisyImage((i),(j+1));
                                 S11 = NoisyImage((i),(j+2));
                                 S12 = NoisyImage((i+1),(j-1));
                                 S13 = NoisyImage((i+2),(j-2));
                                 S14 = NoisyImage((i+1),(j));
                                 S15 = NoisyImage((i+2),(j));
                                 S16 = NoisyImage((i+1)),((j+1));
                                 S17 = NoisyImage((i+2)),((j+2));
                                 TempPixel2=[S1,S2,S3,S4,S5,S6,S7,S8,S9,S10,S11,S12,S13,S14,S15,S16,S17];
                                 Zmed2=median(TempPixel2);
                                 OutImage(i,j)= Zmed2;
                                 else
                                     
                                 OutImage(i,j)= Zmed;

                                 end

                             end         
                     
          end    
      
    end
end
imshow(OutImage,[]);
title('Adaptive Median Filter')
disp('exit');

%%GLCM Feature Extraction
% Y=rgb2gray(OutImage);
Y=double(OutImage);


%     statsArray=[];
%     Y=Y{k};
    glcm2=graycomatrix(Y);
    stats = GLCM_Features1(glcm2,0);
    ExtractedFeaturesm16=stats;
% save('ExtractedFeaturesm16.mat')
    statsTable = struct2table(stats);
%     statsArray = table2array(statsTable);
 statsArray(k,:) = table2array(statsTable);
% statsArray'


end
%     coeff = pca(statsArray);
    [coeff,score,latent,~,explained] = pca(statsArray);
    covarianceMatrix = cov(statsArray);
    [V,D] = eig(covarianceMatrix);
    coeff
    V
    dataInPrincipalComponentSpace = statsArray*coeff
    score2=score(:,1:2);
%     coeff2 = coeff(:,1:2);
    corrcoef(dataInPrincipalComponentSpace)
    
var(dataInPrincipalComponentSpace)'
latent
sort(diag(D),'descend')
    

    inp=input('Enter Image:');
    i=imread(inp);
    try
    i=rgb2gray(i);
end
    z=im2bw(i,0.1);
figure(2)
imshow(z);title('Original B&W')
info=regionprops(z);
a=cat(1,info.Area);
[m,l]=max(a);
X=info(l).Centroid;
bw2=bwselect(z,X(1),X(2),8);
i=immultiply(i,bw2);
figure(3)
imshow(i);
title('Getting the Breast and Muscle')

%% Deleting Black Ground

% We will delete the black corners
% So that we can select the muscle
% using bwselect
% convert to B&W first time
[x,y]=size(z);
tst1=zeros(x,y);

% detect empty rows
r1=[];
m=1;
for j=1:x
    if z(j,:)==tst1(j,:)
        r1(m)=j;
        m=m+1;
    end
end

% detect empty columns
r2=[];
m=1;
for j=1:y
    if z(:,j)==tst1(:,j)
        r2(m)=j;
        m=m+1;
    end
end

% Deleting
i(:,r2)=[];
i(r1,:)=[];

figure(4)
imshow(i);title('after deleting background');


%% Deleting the Muscle

if i(1,1)~=0
    c=3;
    r=3;
else
    r=3;
    c=size(i,2)-3;
end

z2=im2bw(i,0.5);
bw3=bwselect(z2,c,r,8);
bw3=~bw3;
ratio=min(sum(bw3)/sum(z2));
if ratio>=1
    i=immultiply(i,bw3);
else
    z2=im2bw(i,0.75);
    bw3=bwselect(z2,c,r,8);
    ratio2=min(sum(bw3)/sum(z2));
    if round(ratio2)==0
        lvl=graythresh(i);
        z2=im2bw(i,1.75*lvl);
        bw3=bwselect(z2,c,r,8);
        bw3=~bw3;
        i=immultiply(i,bw3);
    else
        bw3=~bw3;
        i=immultiply(i,bw3);
    end
end
figure(5)
imshow(i)
title('Getting only the Breast')


J = imnoise(i,'salt & pepper', 0.02);
NoisyImage=J;
[R C P]=size(NoisyImage);
OutImage=zeros(R,C);
figure;
% imshow(J);
Zmin=[];
Zmax=[];
Zmed=[];

for i=1:R
    for j=1:C
               if (i==1 & j==1)
                   
                   
                        
                        
          % for right top corner[8,7,6]
                elseif (i==1 & j==C)
                        
                        
			% for bottom left corner[2,3,4]
                elseif (i==R & j==1)
                        
                        
        		     % for bottom right corner[8,1,2]

                elseif (i==R & j==C)
                        
                        
        		    %for top edge[8,7,6,5,4]

                elseif (i==1)
		      			 	
		      			 
                     		      			% for right edge[2,1,8,7,6]

                elseif (i==R)
		      			 
		      			% // for bottom edge[8,1,2,3,4]
		      			 
                elseif (j==C)
		      			 
						 
                         %// for left edge[2,3,4,5,6]
						 
                elseif (j==1)
                    
						 
                            
                         
						 
               else
                    
                   
						 	
                             SR1 = NoisyImage((i-1),(j-1));
 							 SR2 = NoisyImage((i-1),(j));
    						 SR3 = NoisyImage((i-1),(j+1));
 							 SR4 = NoisyImage((i),(j-1));
                             SR5 = NoisyImage(i,j);
 							 SR6 = NoisyImage((i),(j+1));
 							 SR7 = NoisyImage((i+1),(j-1));
 							 SR8 = NoisyImage((i+1),(j));
 							 SR9 = NoisyImage((i+1)),((j+1));
                             TempPixel=[SR1,SR2,SR3,SR4,SR5,SR6,SR7,SR8,SR9];
                             Zxy=NoisyImage(i,j);
                             Zmin=min(TempPixel);
                             Zmax=max(TempPixel);
                             Zmed=median(TempPixel);
                             A1 = Zmed - Zmin;
                             A2 = Zmed - Zmax;
                         
                             if A1 > 0 && A2 < 0
                             
                                  %   go to level B
                                  B1 = Zxy - Zmin;
                                  B2 = Zxy - Zmax;
                                  if B1 > 0 && B2 < 0
                                      OutImage(i,j)= Zxy;
                                  else
                                      OutImage(i,j)= Zmed;

                                  end
                             else
                                 
                                 if ((R > 4 && R < R-5) && (C > 4 && C < C-5))
                                 
                                 S1 = NoisyImage((i-1),(j-1));
                                 S2 = NoisyImage((i-2),(j-2));
                                 S3 = NoisyImage((i-1),(j));
                                 S4 = NoisyImage((i-2),(j));
                                 S5 = NoisyImage((i-1),(j+1));
                                 S6 = NoisyImage((i-2),(j+2));
                                 S7 = NoisyImage((i),(j-1));
                                 S8 = NoisyImage((i),(j-2));

                                 S9 = NoisyImage(i,j);
                                 S10 = NoisyImage((i),(j+1));
                                 S11 = NoisyImage((i),(j+2));
                                 S12 = NoisyImage((i+1),(j-1));
                                 S13 = NoisyImage((i+2),(j-2));
                                 S14 = NoisyImage((i+1),(j));
                                 S15 = NoisyImage((i+2),(j));
                                 S16 = NoisyImage((i+1)),((j+1));
                                 S17 = NoisyImage((i+2)),((j+2));
                                 TempPixel2=[S1,S2,S3,S4,S5,S6,S7,S8,S9,S10,S11,S12,S13,S14,S15,S16,S17];
                                 Zmed2=median(TempPixel2);
                                 OutImage(i,j)= Zmed2;
                                 else
                                     
                                 OutImage(i,j)= Zmed;

                                 end

                             end         
                     
          end    
      
    end
end
imshow(OutImage,[]);
title('Adaptive Median Filter')
disp('exit');

%%GLCM Feature Extraction
% Y=rgb2gray(OutImage);
Y=double(OutImage);



%     statsArray=[];
%     Y=Y{k};
    glcm2=graycomatrix(Y);
    stats = GLCM_Features1(glcm2,0);
    ExtractedFeaturesm16=stats;
% save('ExtractedFeaturesm16.mat')
    statsTable = struct2table(stats);
%     statsArray = table2array(statsTable);
 statsArray1 = table2array(statsTable);
% statsArray'


% [coeff3,score3,latent,~,explained] = pca(statsArray1);
%     coeff4 = coeff(:,1:2);
[coeff3,score3,latent,~,explained] = pca(statsArray1);
   TrainingSet=[score2(1,:);score2(2,:);score2(3,:);score2(4,:);score2(5,:);score2(6,:);score2(7,:);score2(8,:);
       score2(9,:);score2(10,:);score2(11,:);score2(12,:);score2(13,:);score2(14,:);score2(15,:);score2(16,:);
       score2(17,:);score2(18,:);score2(19,:);score2(20,:);score2(21,:);score2(22,:);score2(23,:);score2(24,:);
       score2(25,:);score2(26,:);score2(27,:);score2(28,:);score2(29,:);score2(30,:);score2(31,:);score2(32,:)]
   GroupTrain={'1','1','1','1','1','1','1','1','1','1','1','1','1','1','1','1','2','2','2','2','2','2','2','2','2','2','2','2','2','2','2','2',}
   
   TestSet=score3;
   
   SVMModels=cell(2,1);
   S=GroupTrain;
   classes =unique(S);
   rng(1);
   
%     test = TestSet;
%         result = multisvm(TrainingSet,classes,test);
   for j=1:numel(classes)
       indx=strcmp(S',classes(j));
       SVMModels{j}=fitcsvm(score2,indx,'ClassNames',[false true],'Standardize',true,'KernelFunction','rbf','BoxConstraint',1);
   end
   
   xGrid=TestSet;
   for j= 1:numel(classes)
   [~,score]=predict(SVMModels{j},xGrid);
   Scores(:,j)=score(:,2);
   end
   
   [~,maxScore]=max(Scores,[],2);
   
   result=maxScore;
   figure,imshow(i)
   
   if result==1
       msgbox('Benign')
   elseif result==2
       msgbox('Malignant')
   end
   

The error message is below:

   Error using classreg.learning.classif.ClassificationModel/predictEmptyX (line 203)
X data must have 2 columns.

Error in classreg.learning.classif.ClassificationModel/predict (line 406)
                [labels,scores,cost] = predictEmptyX(this,X);

Error in classreg.learning.classif.CompactClassificationSVM/predict (line 433)
                [e-mail](this,X,varargin{:});

Answer 1

[Mahesh Taparia]

Hi

From your code and the error message, it seems the variable xGrid did not have the 2 columns.

[~,score]=predict(SVMModels{j},xGrid);

Check the dimension of xGrid, there may be dimensional mismatch/ not properly arranges/you might have missed the data.

[Warid Islam]

Hi @Mahesh,

Thank you for the response. You are correct. xGrid didn't have two columns. It is actually the features extracted from the test image. It has 22 columns and one row. I am actually confused on how to apply PCA on a single row. Do you recommend taking just the first two columns only? do you think the result would be valid then?

[Mahesh Taparia]

Hi Warid

Number of columns will be same as the number of columns you have used to train the model using fitcsvm. Regarding PCA, you can reduce to any number of dimension. It depends how much variance you want in data. For more information on PCA, you can refer to this documentation.

[Warid Islam]

Hi @Mahesh,

I have applied PCA on a 1*22 array(1 row and 22 column).

[coeff,score,latent,explained]=pca(statsArray1)

However, all the values of the variables are empty. I was wondering what could be the cause for this issue. Thank you.

[Mahesh Taparia]

Hi

Atleast take the number of data to be 2 or more. For example, consider below:

a=randn(2,20);
[coeff,score,latent]=pca(a,'NumComponents',10);

[Warid Islam]

Thank you Mahesh..