BraTS2017特征提取.zip_BraTS 2017_BraTS 2017数据_BraTS2017_BraTS2017数据集

% Function: % % 《数字图像技术处理及应用》课程大作业预处理程序及特征提取程序 % Writer: Wu Chengpeng % Time: 2017/12/12 clear;clc;close all %% ====================Set the Input and Output Data Path============ % 该路径为测试数据的输入路径，下面包括HGG和LGG两个子文件夹 test_path = 'D:\Data\MICCAI_BraTS17_Data_Training\'; % 设置输出路径 feature_name = {'Intensity','Texture','Gradient','GradientMean','GradientVariance','SingleRay','DoubleRay','Symmetry'}; % 特征名称，建议每种特征单独处理，以便后期选择使用哪些特征 % feature_length = 3; % 特征长度，也就是每个点的该特征需要用几个数存储，如方差用3个（3^3,5^3,7^3）就行 output_path = strcat(test_path,'Features\',feature_name,'\'); HGG_path = strcat(test_path,'HGG\'); LGG_path = strcat(test_path,'LGG\'); % 列出HGG子文件夹下的所有子文件夹，也就是每个病人一个文件夹 HGG_dir = fullfile(HGG_path); HGG_dir_out = dir(fullfile(HGG_dir,'*')); HGG_dir_name = {HGG_dir_out(3:end).name}'; % HGG_name是后面需要用的一个路径前缀 HGG_name = HGG_dir_name; for ii = 1:length(HGG_dir_name) HGG_name{ii} = strcat(HGG_path,HGG_dir_name{ii},'\',HGG_dir_name{ii}); end % 列出LGG子文件夹下的所有子文件夹，也就是每个病人一个文件夹 LGG_dir = fullfile(LGG_path); LGG_dir_out = dir(fullfile(LGG_dir,'*')); LGG_dir_name = {LGG_dir_out(3:end).name}'; % LGG_name是后面需要用的一个路径前缀 LGG_name = LGG_dir_name; for ii = 1:length(LGG_dir_name) LGG_name{ii} = strcat(LGG_path,LGG_dir_name{ii},'\',LGG_dir_name{ii}); end % 将HGG和LGG的两部分病人合并，也就是全部的病人文件夹路径 total_dir_name = {HGG_name{:},LGG_name{:}}; total_output_dir = cell(1,length(feature_name)); for ii = 1:length(feature_name) if ~exist(output_path{ii},'dir') mkdir(output_path{ii}); end if ~exist(strcat(output_path{ii},'HGG'),'dir') mkdir(strcat(output_path{ii},'HGG')); end if ~exist(strcat(output_path{ii},'LGG'),'dir') mkdir(strcat(output_path{ii},'LGG')); end output_HGG_path = HGG_dir_name; for jj = 1:length(HGG_dir_name) output_HGG_path{jj} = strcat(output_path{ii},'HGG\',HGG_dir_name{jj},'\'); if ~exist(output_HGG_path{jj},'dir') mkdir(output_HGG_path{jj}); end end output_LGG_path = LGG_dir_name; for jj = 1:length(LGG_dir_name) output_LGG_path{jj} = strcat(output_path{ii},'LGG\',LGG_dir_name{jj},'\'); if ~exist(output_LGG_path{jj},'dir') mkdir(output_LGG_path{jj}); end end % 将HGG和LGG的两部分病人合并，也就是全部的病人输出文件夹路径 total_output_dir{ii} = {output_HGG_path{:},output_LGG_path{:}}; end %% ====================Loading Data================================== sigma = [2.5,3.0,4.5]; % sigma = 3.0; % 通过parfor对每个病例进行循环，并行处理加快速度 parfor kk = 61:70 tic nii_path = total_dir_name{kk}; t1_struct = load_nii(strcat(nii_path,'_t1.nii.gz')); t2_struct = load_nii(strcat(nii_path,'_t2.nii.gz')); t1ce_struct = load_nii(strcat(nii_path,'_t1ce.nii.gz')); flair_struct = load_nii(strcat(nii_path,'_flair.nii.gz')); [img_w,img_h,img_d] = size(t1_struct.img); % 三维数据大小 toc disp(strcat('Finish Loading Data!',nii_path)); tic %% Feature 1: Intensity intensity_feature = IntensityFeature(t1_struct,t2_struct,t1ce_struct,flair_struct); output_matname = strcat(total_output_dir{1}{kk},'intensity.mat'); name_t1 = 'intensity_feature'; parsave(output_matname,name_t1,intensity_feature); toc disp(strcat('Finish Extract Intensity Features!',nii_path)); %% ====================Part 1: Smooth and Normalized================= % 高斯平滑 tic % 下面XX_smooth是一个n*1的cell，n是sigma的个数，此处n=3 t1_smooth = SmoothGauss(t1_struct,sigma); % 为了避免高斯平滑有问题，同时保留一个原始数据，此处XX_smooth变成(n+1)*1的cell % t1_smooth = {t1_struct,t1_smooth{:}}; t2_smooth = SmoothGauss(t2_struct,sigma); % t2_smooth = {t2_struct,t2_smooth{:}}; t1ce_smooth = SmoothGauss(t1ce_struct,sigma); % t1ce_smooth = {t1ce_struct,t1ce_smooth{:}}; flair_smooth = SmoothGauss(flair_struct,sigma); % flair_smooth = {flair_struct,flair_smooth{:}}; toc disp(strcat('Finish Smoothing Data!',nii_path)); % 归一化 tic t1_smooth = ThreeDNorm(t1_smooth); % 为了避免三维归一化有问题，同时保留一个原始数据，此处XX_smooth变成(n+2)*1的cell % t1_smooth = {t1_struct,t1_smooth{:}}; t2_smooth = ThreeDNorm(t2_smooth); % t2_smooth = {t2_struct,t2_smooth{:}}; t1ce_smooth = ThreeDNorm(t1ce_smooth); % t1ce_smooth = {t1ce_struct,t1ce_smooth{:}}; flair_smooth = ThreeDNorm(flair_smooth); % flair_smooth = {flair_struct,flair_smooth{:}}; toc disp(strcat('Finish Normalizing Data!',nii_path)); %% ====================Part 2: Extract Features====================== for nn = 1:length(t1_smooth) t1_nii_struct = t1_smooth{nn}; t2_nii_struct = t2_smooth{nn}; t1ce_nii_struct = t1ce_smooth{nn}; flair_nii_struct = flair_smooth{nn}; t1_smooth{nn} = []; t2_smooth{nn} = []; t1ce_smooth{nn} = []; flair_smooth{nn} = []; if nn==2 %% T1 NII Struct tic % Feature 2: Texture t1_texture = Extract3DTexture(t1_nii_struct); output_matname_part1 = strcat(total_output_dir{2}{kk},'texture_','sigma',num2str(sigma(nn)),'_t1_3x3.mat'); output_matname_part2 = strcat(total_output_dir{2}{kk},'texture_','sigma',num2str(sigma(nn)),'_t1_5x5.mat'); output_matname_part3 = strcat(total_output_dir{2}{kk},'texture_','sigma',num2str(sigma(nn)),'_t1_7x7.mat'); name_t1_part1 = 't1_texture_3x3'; name_t1_part2 = 't1_texture_5x3'; name_t1_part3 = 't1_texture_7x7'; t1_texture_3x3 = squeeze(t1_texture(1,:,:,:,:)); parsave(output_matname_part1,name_t1_part1,t1_texture_3x3); t1_texture_5x5 = squeeze(t1_texture(2,:,:,:,:)); parsave(output_matname_part2,name_t1_part2,t1_texture_5x5); t1_texture_7x7 = squeeze(t1_texture(3,:,:,:,:)); parsave(output_matname_part3,name_t1_part3,t1_texture_7x7); t1_texture = []; t1_texture_3x3 = []; t1_texture_5x5 = []; t1_texture_7x7 = []; toc disp(strcat('Finish Extract T1 Texture Feature!',nii_path)); % Feature 3: Gradient tic img_tep=double(t1_nii_struct.img); [px,py,pz]=gradient(img_tep); gra=sqrt((px.*px)+(py.*py)+(pz.*pz)); t1_gradient=gra;%gradiant output_matname = strcat(total_output_dir{3}{kk},'gradient_','sigma',num2str(sigma(nn)),'_t1.mat'); name_t1 = 't1_gradient'; parsave(output_matname,name_t1,t1_gradient); t1_gradient = []; toc disp(strcat('Finish Extract T1 Gradient Feature!',nii_path)); % Feature 4: Gradient Mean tic t1_gradientmean_3x3 = grad_mean( gra,3,3,3 ); %grad_mean3 t1_gradientmean_5x5 = grad_mean( gra,5,5,5 ); %grad_mean5 t1_gradientmean_7x7 = grad_mean( gra,7,7,7 ); %grad_mean7 output_matname_part1 = strcat(total_output_dir{4}{kk},'gradientmean_','sigma',num2str(sigma(nn)),'_t1_3x3.mat'); output_matname_part2 = strcat(total_output_dir{4}{kk},'gradientmean_','sigma',num2str(sigma(nn)),'_t1_5x5.mat'); output_matname_part3 = strcat(total_output_dir{4}{kk},'gradientmean_','sigma',num2str(sigma(nn)),'_t1_7x7.mat'); name_t1_part1 = 't1_gradientmean_3x3'; name_t1_part2 = 't1_gradientmean_5x5'; name_t1_part3 = 't1_gradientmean_7x7';