% this is a sample processing program demonstrates the two modes of
% segmentation explained in the paper (one by one - Batch)
%" Automatic Semantic Segmentation of Breast Tumors in Ultrasound  ...
% Images Based on Combining Fuzzy Logic and Deep Learning – A Feasibility Study "
% by: Samir M. Badawy 1, Abd El-Naser A. Mohamed 2, Alaa A. Hefnawy 3, ...
% Hassan E. Zidan 3, Mohammed T. GadAllah 3,*, Ghada M. El-Banby 1

% first you have to download MT_Small_Dataset from S1 in the paper or from
%  https://www.kaggle.com/mohammedtgadallah/mt-small-dataset 
% second you must download the eight CNNs used here from Ref. No. 8:
% https://doi.org/10.1016/j.compbiomed.2020.104036
%========================================================================

clear ; close all; clc;
cd('C:\Program Files\Polyspace\R2020b\bin\Processing_modes_Explanation');

% Select a CNN model

nets = {'alexnet','unet','vgg16','vgg19','resnet18','resnet50','mobilenet','xception'};
choice = menu('Select a CNN model:',nets);
% Load CNN for semantic segmentation
load(fullfile('CNNs', sprintf("cnn_%s.mat",nets{choice})));
%=========================================================================
%=========================================================================
%            Define reading and write locations for input images
%-------------------------------------------------------------------------

read_200_benign = '';
read_200_malignant = '';
read_200_fuzzy_benign = '';
read_200_fuzzy_malignant = '';
%=========================================================================
%            Define reading and write locations for Output images
%-------------------------------------------------------------------------
write_200_benign = '';   
write_200_malignant = '';
write_200_fuzzy_benign = '';
write_200_fuzzy_malignant = '';

%=========================================================================
%  Now starting One by One processing mode for each 200 images as follow:
%-------------------------------------------------------------------------

% Benign 200 before fuzzy (original)
    for gg = 1:200
        
        cd (read_200_benign);
        input = strcat ('benign_Adjusted (', num2str(gg),').png');
        eval ('im = imread (input);');
        imseg = semanticseg(im, net);
        imseg = double (imseg);
        imseg = mat2gray (imseg , [0 255]);
        
        output = strcat ('pixelLabel_', num2str (gg, '%03d'),'.png');
        
        cd(write_200_benign);
        eval ('imwrite (imseg, output)');
    end
    
% Malignant 200 before fuzzy (original)
 for gg = 1:200       
        cd (read_200_malignant);
        input = strcat ('malignant_Adjusted (', num2str(gg),').png');
        eval ('im = imread (input);');
        imseg = semanticseg(im, net);
        imseg = double (imseg);
        imseg = mat2gray (imseg , [0 255]);
        
        output = strcat ('pixelLabel_', num2str (gg, '%03d'),'.png');
        
        cd(write_200_malignant);
        eval ('imwrite (imseg, output)');
 end
   
 % Benign 200 After fuzzy
    for gg = 1:200
        
        cd (read_200_fuzzy_benign);
        input = strcat ('benign_Fuzzy_', num2str(gg, '%03d'),'.png');
        eval ('im = imread (input);');
        imseg = semanticseg(im, net);
        imseg = double (imseg);
        imseg = mat2gray (imseg , [0 255]);
        
        output = strcat ('pixelLabel_', num2str (gg, '%03d'),'.png');
        
        cd(write_200_fuzzy_benign);
        eval ('imwrite (imseg, output)');
    end
    
% Malignant 200 After fuzzy
 for gg = 1:200     
        cd (read_200_fuzzy_malignant);
        input = strcat ('malignant_Fuzzy_', num2str(gg, '%03d'),'.png');
        eval ('im = imread (input);');
        imseg = semanticseg(im, net);
        imseg = double (imseg);
        imseg = mat2gray (imseg , [0 255]);
        
        output = strcat ('pixelLabel_', num2str (gg, '%03d'),'.png');
        
        cd(write_200_fuzzy_malignant);
        eval ('imwrite (imseg, output)');
 end
%=======================================================================================
%====================================================================================
% % Define the class names and their associated label IDs - for both 
% ground truth and predected output
classNames = ["Cancer","Normal_Tissue"];
labelIDs   = [2 1];
%=======================================================================
% creating pixelLabelDatastores holding ground truth lables to starte
% evaluating process
%---------------------------------------------------------------
cd('C:\Program Files\Polyspace\R2020b\bin\Processing_modes_Explanation');
ground_Truth_benign_Dir = fullfile('GroungTruth','benign_ground');
ground_Truth_malignant_Dir = fullfile('GroungTruth','malignant_ground');

ground_Truth_benign = pixelLabelDatastore(ground_Truth_benign_Dir,classNames,labelIDs);
ground_Truth_malignant = pixelLabelDatastore(ground_Truth_malignant_Dir,classNames,labelIDs);

%=========================================================================
% creating datastore holding for predected output lables from the semantic
% segmentation process 
%-------------------------------------------------------------------------

cd(''); % define the output location you have determined before for original benign and malignant segmented images 
original_benign_labels = pixelLabelDatastore('benign',classNames,labelIDs);
original_malignant_labels = pixelLabelDatastore('malignant',classNames,labelIDs);

cd('');% define the output location you have determined before for fuzzy benign and malignant segmented images 
fuzzy_benign_labels = pixelLabelDatastore('benign',classNames,labelIDs);
fuzzy_malignant_labels = pixelLabelDatastore('malignant',classNames,labelIDs);

%=======================================================================================
% Quantitative evaluation of the prediction (segmented) results against the ground truth:
%=======================================================================================

original_benign = evaluateSemanticSegmentation(original_benign_labels,ground_Truth_benign);
disp('this is for segmented benign -original ');
fuzzy_benign =  evaluateSemanticSegmentation(fuzzy_benign_labels,ground_Truth_benign);
disp('this is for segmented benign by fuzzy');
original_malignant = evaluateSemanticSegmentation(original_malignant_labels,ground_Truth_malignant);
disp('this is for segmented malignant -original ');
fuzzy_malignant = evaluateSemanticSegmentation(fuzzy_malignant_labels,ground_Truth_malignant);
disp('this is for segmented malignant by fuzzy');

%=========================================================================
%=========================================================================
%=========================================================================
%=========================================================================
%=========================================================================
%=========================================================================
% In case of Batch processing mode  :
% All the program above must be repeated except lines [ 40 to 105 ] 
% it will be replaced by the follow:
%-------------------------------------------------------------------------

% Benign 200 before fuzzy (original)
        cd (read_200_benign);
        % creating a datastore containing 200 images (Batch)
        imds_original_benign = imageDatastore(read_200_benign);
 % Excution of the semantic segmentation process on the Batch
 % datastore at one time
 Segmented = semanticseg(imds_original_benign,net,"WriteLocation",'write_200_benign'); 
 % segmented original-benign       
        
 
 % Malignant 200 before fuzzy (original)
        cd (read_200_malignant);
        % creating a datastore containing 200 images (Batch)
        imds_original_malignant = imageDatastore(read_200_malignant);
 % Excution of the semantic segmentation process on the Batch
 % datastore at one time
 Segmented = semanticseg(imds_original_malignant,net,"WriteLocation",'write_200_malignant'); 
 
 
 
 % Benign 200 After fuzzy
         cd (read_200_fuzzy_benign);
        % creating a datastore containing 200 images (Batch)
        imds_fuzzy_benign = imageDatastore(read_200_fuzzy_benign);
 % Excution of the semantic segmentation process on the Batch
 % datastore at one time
 Segmented = semanticseg(imds_fuzzy_benign,net,"WriteLocation",'write_200_fuzzy_benign'); 

 
 
 % Malignant 200 After fuzzy
         cd (read_200_fuzzy_malignant);
        % creating a datastore containing 200 images (Batch)
        imds_fuzzy_malignant = imageDatastore(read_200_fuzzy_malignant);
 % Excution of the semantic segmentation process on the Batch
 % datastore at one time
 Segmented = semanticseg(imds_fuzzy_malignant,net,"WriteLocation",'write_200_fuzzy_malignant');