How to split into different tables a multi-level cell?

My final goal is to have a 428x11 table from an image analysis that I ran which final outcome is 'data'. Object 'data' is a 428x1 cell, and that one column is a 1x11 cell that looks like ['name', v1, v2, v3, v4, v5, 6, v7, v8, V9, v10], and again, 428 times.

I have tried using data_table=regexp(data(:), '\s+', 'split'); but get an error using regexp 'All cells must be char row vectors.', and tried with 'cellstr' but didn't work. I have also tried with loops -although I am not proficient with iterations- but I can't index with multi-level cells, and something like data2=data{2:end,1}{1,1}; or data2=data{:,1}{:,1}; hasn't worked. I have seen similar questions in but MATLAB Answers haven't worked for me so far.

I really appreciate any help!

Answer 1

[per isakson]

"final goal is to have a 428x11 table"

Study this

data = {
            {'name1',1,2,3,4,5,6,7,8,9,0}
            {'name2',1,2,3,4,5,6,7,8,9,0}
            {'name3',1,2,3,4,5,6,7,8,9,0}
        };
cac = cat( 1, data{:} );
mytable = cell2table( cac )

it displays in the command window

mytable = 
     cac1      cac2    cac3    cac4    cac5    cac6    cac7    cac8    cac9    cac10    cac11
    _______    ____    ____    ____    ____    ____    ____    ____    ____    _____    _____
    'name1'    1       2       3       4       5       6       7       8       9        0    
    'name2'    1       2       3       4       5       6       7       8       9        0    
    'name3'    1       2       3       4       5       6       7       8       9        0

[JoanManBar]

Thank you so much for your answer. It actually worked with YOUR sample matrix but not with my data. However it's getting close since it creates a 242x11 table. So I wonder what I am doing wrong. I am sharing my cell in case you want to take a closer look. Thank you once again!

[per isakson]

• "it creates a 242x11 table" thus the two lines of code do what they are supposed to do
• "what I am doing wrong" I cannot see any requirement besides "242x11 table" and that is fulfilled. What do you further require? What do you want to achieve? You don't say.

>> load('matlab.mat')
>> cac = cat( 1, data{:} );
mytable = cell2table( cac )
mytable = 
         cac1          cac2     cac3      cac4      cac5     cac6     cac7      cac8      cac9     cac10     cac11 
    _______________    ____    ______    ______    ______    ____    ______    ______    ______    ______    ______
    'sample006.tif'     71     172.31    138.36     112.3    2570    130.46     25.35    172.31    138.36     112.3
    'sample026.tif'    116      168.1    134.01    106.86     952    68.517     17.91     168.1    134.01    106.86
    'sample223.tif'     96     163.64    137.07    115.81    1293    83.787    20.473    163.64    137.07    115.81
    'sample231.tif'    107     178.36    150.75    127.69    2283    115.56    25.323    178.36    150.75    127.69
    'sample006.tif'     71     172.31    138.36     112.3    2570    130.46     25.35    172.31    138.36     112.3
    'sample223.tif'     96     163.64    137.07    115.81    1293    83.787    20.473    163.64    137.07    115.81
    'sample006.tif'     71     172.31    138.36     112.3    2570    130.46     25.35    172.31    138.36     112.3
    'sample231.tif'    107     178.36    150.75    127.69    2283    115.56    25.323    178.36    150.75    127.69
    'sample223.tif'     96     163.64    137.07    115.81    1293    83.787    20.473    163.64    137.07    115.81
    'sample026.tif'    116      168.1    134.01    106.86     952    68.517     17.91     168.1    134.01    106.86
...

Speculations: The cell array data isn't what you think it is. There are only four unique row headers, i.e. file names in data

>> col1 = cac(:,1);
>> length(col1)
ans =
   242
>> [ col1_unique, ixa, ixc ] = unique( col1 );
>> col1_unique
col1_unique = 
    'sample006.tif'
    'sample026.tif'
    'sample223.tif'
    'sample231.tif'
>>

Actually, there are only four unique rows

>> [ cac_unique_rows, ixa, ixc ] = unique_rows( cac, [1:size(cac,2)] );
>> cac_unique_rows
cac_unique_rows = 
    'sample006.tif'    [ 71]    [172.3078]    [138.3623]    [112.3047]    [2570]    [130.4569]    [25.3503]    [172.3078]    [138.3623]    [112.3047]
    'sample026.tif'    [116]    [168.1040]    [134.0053]    [106.8592]    [ 952]    [ 68.5168]    [17.9097]    [168.1040]    [134.0053]    [106.8592]
    'sample223.tif'    [ 96]    [163.6419]    [137.0719]    [115.8105]    [1293]    [ 83.7872]    [20.4729]    [163.6419]    [137.0719]    [115.8105]
    'sample231.tif'    [107]    [178.3574]    [150.7464]    [127.6881]    [2283]    [115.5610]    [25.3232]    [178.3574]    [150.7464]    [127.6881]

unique_rows is a wrapper to Unique Rows for a cell array by Jim Hokanson

[JoanManBar]

More than just an answer, this was a meaningful lesson! You are right, my data cell was completely wrong, and although I still don't accomplish my objective (sheet with RGB and size of my seeds), you answered wisely the question I asked. I am still struggling really bad to learn the basics of Matlab.

I know I have already asked a lot, but if you would still like to help me, all that I want is my files' names in each corresponding row, i.e. 'sample..' instead of numbers from 1 to number of files -I know that's a different question.

I am attaching my code and files anyways.

.

.

Two copies removed

.

close all; clear all; clc;
imagefiles = dir('*.tif');
filenames = {imagefiles.name};
nfiles = length(imagefiles);

    for i=1:nfiles
      currentfilename = imagefiles(i).name;
      currentimage = imread(currentfilename);
      images = currentimage;

      images = imcrop(images, [0 0 1557 1844]);

      Red_ch = images(:,:,1);
      Green_ch = images(:,:,2);
      Blue_ch = images(:,:,3);

      binary = Red_ch > 90;
      binary = bwareafilt(binary, [100, inf]);
      se = strel('disk',7,0);
      binary = imerode(binary, se);

      [labeledI, numSeeds] = bwlabel(binary);

      props = regionprops(labeledI, 'Area', 'MajorAxisLength', 
         'MinorAxisLength');

      for k = 1:numSeeds
          Areas(k) = props(k).Area;
          Length(k) = props(k).MajorAxisLength;
          Width(k) = props(k).MinorAxisLength;      
      end

      propsR = regionprops(labeledI, Red_ch, 'MeanIntensity', 
        'PixelValues');
      propsG = regionprops(labeledI, Green_ch, 'MeanIntensity', 
        'PixelValues');
      propsB = regionprops(labeledI, Blue_ch, 'MeanIntensity', 
        'PixelValues');

      MeansR = [propsR.MeanIntensity];
      MeansG = [propsG.MeanIntensity];
      MeansB = [propsB.MeanIntensity];

      for j = 1:numSeeds
          PixMeansR(j) = mean(propsR(j).PixelValues);
          PixMeansR = PixMeansR';
          PixMeansG(j) = mean(propsG(j).PixelValues);
          PixMeansG = PixMeansG';
          PixMeansB(j) = mean(propsB(j).PixelValues);
          PixMeansB = PixMeansB';
      end

      if exist('data','var')~=1 
         data = [repmat(i,1,numSeeds)', (1:numSeeds)', MeansR', MeansG', 
         MeansB', Areas', Length', Width', filenames];
      else
         data = vertcat(data,[repmat(i,1,numSeeds)', (1:numSeeds)', 
         MeansR', MeansG', MeansB', Areas', Length', Width']);
     end 
 end

THANK YOU SO MUCH!

[per isakson]

I assume your question is about

if exist('data','var')~=1
    data = [repmat(i,1,numSeeds)', (1:numSeeds)', MeansR', MeansG', ...
        MeansB', Areas', Length', Width', filenames];
else
    data = vertcat(data,[repmat(i,1,numSeeds)', (1:numSeeds)', ...
        MeansR', MeansG', MeansB', Areas', Length', Width']);
end

In the second iteration of the loop, when the else-clause is executed for the first time, I get the error

Error using vertcat
Dimensions of matrices being concatenated are not consistent.
Error in cssm (line 57)
            data = vertcat(data,[repmat(i,1,numSeeds)', (1:numSeeds)', ...

which shouldn't surprise.

Comments on this if-else-end construct

• it attempts to create a cell array, data, which differs from data that you uploaded in the mat-file. E.g. MeanR ... in the uploaded data are scalars, here they are vectors with one element per seed.
• In the uploaded data the format of the first row didn't differ from the rest. It might be a good idea to show column headers in the first row, e.g. ...,'MeansR','MeansG','MeansB','Areas','Length','Width',...
• this construct is inefficient and hard to read. The numbers of files is known, thus pre-allocating the cell array, data = cell( nfiles, 12 );, and use assignments to fill it with data.

I'm lost. Show with an example what data shall look like.

[JoanManBar]

I am so sorry, I feel like I'm wasting your time and that's because I haven't been clear at all. The previous code was a little different that the second one since I realized -thank to you- that the array was not what I thought it was. Additionally, the second I uploaded was had mistakes as well. I honestly don't know what to do!

Dear Per Isakson, I want to have a spread sheet that shows these columns and their corresponding values:

 Image_Name, Seed, Mean_Red, Mean_Green, Mean_Blue, Area, Length, Width

Where:

 Image_Name: Name of each photo
 Seed: Number of seeds per photo
 Mean_Red: Mean red value of each seed
 Mean_Green: Mean green value of each seed
 Mean_Blue: Mean blue value of each seed
 Area: Area of each seed
 Length: Length of each seed
 Width: Width of each seed

I'd like to have other variables, but that would be more than enough. How can I do that?

Thank you once again, I really appreciate all the time and help you have provided.

[per isakson]

There is a serious problem with this loop and a couple of minor.

for k = 1:numSeeds
    Areas(k) = props(k).Area;
    Length(k) = props(k).MajorAxisLength;
    Width(k) = props(k).MinorAxisLength;
end

1. When the outer loop encounters a picture with fewer seeds than the previous the values at the end of Areas etc. will not be overwritten. The length of Areas etc. will increase down the table. Pictures with few seeds will give entries in the table with data on seeds of previous pictures.
2. pre-allocation of memory is a good habit.
3. The generic names, e.g. Length, make me nervous. There is a risk, they are in conflict with names used by Matlab.

[per isakson]

First I downloaded the picture and made an extra copy of sample006.tif, which I called sample.306.tif. (I was surprised that three pictures contained exactly 112 seeds. Ok, it's copies of the same picture.)

Then I modified your code.

Now the question is whether the result is what you expected.

>> data = cssm()
data = 
    'Image_Name'       'Seed'    'Mean_Red'        'Mean_Green'      'Mean_Blue'       'Area'            'Length'          'Width'       
    'sample006.tif'    [  69]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]
    'sample026.tif'    [ 112]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]
    'sample223.tif'    [ 112]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]
    'sample231.tif'    [ 112]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]
    'sample306.tif'    [  69]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]

where

function    data = cssm()
    %#ok<*AGROW>
    %#ok<*NASGU>
    %#ok<*ASGLU>
    imagefiles = dir('sample*.tif');
    filenames = {imagefiles.name};
    nfiles = length(imagefiles);
    %
    data = cell( nfiles+1, 8 );
    data(1,:) = { 'Image_Name','Seed','Mean_Red','Mean_Green' ...
                , 'Mean_Blue' ,'Area','Length'  ,'Width'      };
    %
    for i=1:nfiles
        currentimage = imread( imagefiles(i).name );
        %
        Red_ch   = currentimage(:,:,1);
        Green_ch = currentimage(:,:,2);
        Blue_ch  = currentimage(:,:,3);
        %
        binary = Red_ch > 90;
        binary = bwareafilt(binary, [100, inf]);
        se     = strel('disk',7,0);
        binary = imerode(binary, se);
        %
        [labeledI, numSeeds] = bwlabel(binary);
        %
        props = regionprops(labeledI,'Area','MajorAxisLength','MinorAxisLength');
        %
        Areas  = reshape([props.Area]           , [], 1 );
        Length = reshape([props.MajorAxisLength], [], 1 );
        Width  = reshape([props.MinorAxisLength], [], 1 );
        %
        propsR = regionprops(labeledI, Red_ch, 'MeanIntensity', 'PixelValues');
        propsG = regionprops(labeledI, Green_ch, 'MeanIntensity', 'PixelValues');
        propsB = regionprops(labeledI, Blue_ch, 'MeanIntensity', 'PixelValues');
        %
        MeansR = reshape([propsR.MeanIntensity], [], 1 );
        MeansG = reshape([propsG.MeanIntensity], [], 1 );
        MeansB = reshape([propsB.MeanIntensity], [], 1 );
        %
        data(i+1,:)  = { imagefiles(i).name, numSeeds ...
                     ,   MeansR, MeansG, MeansB       ...
                     ,   Length, Areas, Width         };
    end
end

Comments:

1. I have problems with the blip as a shortcut for transpose. I prefer reshape, because the name is good and reshape outputs the desired shape regardless of whether the input is row or column.
2. I removed some help-variables, which I didn't think added any value.
3. I vectorized the loops over numSeeds.
4. I removed the crop-command to simplify testing.

[per isakson]

And finally

>> mytable = cell2table( data(2:end,:), 'VariableNames',  data(1,:) )
mytable = 
      Image_Name       Seed       Mean_Red         Mean_Green        Mean_Blue            Area             Length            Width     
    _______________    ____    ______________    ______________    ______________    ______________    ______________    ______________
    'sample006.tif'     69     [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]
    'sample026.tif'    112     [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]
    'sample223.tif'    112     [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]
    'sample231.tif'    112     [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]    [112x1 double]
    'sample306.tif'     69     [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]    [ 69x1 double]

[JoanManBar]

Today is one of the best days of my life thanks to you! I can't believe it finally works! And I can't tell with words how grateful I feel with you!

I realized that I uploaded the wrong images and that's why three of them have the same results. I apologize!

Would you mind doing one last thing for me? Could you please tell me how the first column (Image_Name) can have the sample name repeated in rows as many times as there is seeds?

If there is anything I can do for you, please let me know!

THANKS A LOT!

[per isakson]

Here is a function, which takes the cell array data from above and returns the table you want (I think). In this table there is one row per seed. There are no row names.

I find it most rewarding to learn that I really helped you.

My functions lack both comments and error checking. I encourage you to try to understand the code and add some.

>> out_tbl = cssm2( data )
out_tbl = 
      Image_Name       Seed    Mean_Red    Mean_Green    Mean_Blue     Area     Length    Width 
    _______________    ____    ________    __________    _________    ______    ______    ______
    'sample006.tif'      1     171.83      137.49        111.47       122.38    2117      22.299
    'sample006.tif'      2     170.82      134.89        109.36       117.08    1800      20.107
    'sample006.tif'      3     178.06      145.43        115.19       129.65    5976      82.874
    'sample006.tif'      4     153.78      118.21        89.604       128.57    3121       31.13
    'sample006.tif'      5     171.72      141.22        116.84       124.48    1920      20.122
    ...
    ...
>> whos out_tbl data
  Name           Size            Bytes  Class    Attributes
  data           6x8             28410  cell               
  out_tbl      474x8             95114  table

where

function    out_tbl = cssm2( data )
    %
    colhead = data( 1, : );
    out_tbl = one_batch_of_seeds_to_table( data(2,:), colhead );
    %
    for jj = 3 : size( data, 1 )
        tbl = one_batch_of_seeds_to_table( data(jj,:), colhead );
        out_tbl = vertcat( out_tbl, tbl );  %#ok<AGROW>
    end
end

and where

function    tbl = one_batch_of_seeds_to_table( body, colhead )
    %
    len = length( body{3} );
    pic = repmat( body(1), len, 1 );
    sno = reshape( (1:len), [], 1 );
    tbl = table( pic, sno, body{3:end}, 'VariableNames', colhead );
end

[JoanManBar]

Once again, thank you very much! I'm having some trouble with that last part since 'body' is not defined but I think it's time for me to stop bothering you and do the work myself.

I have learned more from your answers than from my experience reading the MATLAB documentation and I really appreciate that.

[JoanManBar]

I got it. you are brilliant, Per Isakson!