how to count number of elements greater than ?

i'am new in matlab i want to the ways to count the number of formfactor in this code is greater than (e.g. 0.8)then this is normal Cell for example and print the number of normal and the number(abnormal) it is between (0.31-0.5)

 hold on;
 S =regionprops(Label,'area','perimeter');
 for i=1:num
    x(i)=S(i).Area; 
    y(i)=S(i).Perimeter;
    %Form factor=4* pi * area/ (perimeter* perimeter) 
    form(i)=4.*pi.*x(i)./(y(i).^2);

 end

i think to make array and if statement which save the true from if statement in array , and i try to make this :

 NormalCount=sum(form(i)==1 || form(i)>=0.8);
 AbnormalCount=sum(form(i)<0.8);
 message1 = sprintf('Number of normal cells: %d %d',NormalCount);
 message2 = sprintf('Number of Abnormal cells: %d %d',AbnormalCount);

 msgbox({message1,message2});

But the result from this is wrong and i don't know this in for loop or i put it put ?

thanks.

Answer 1

[Geoff Hayes]

Menna - rather than using find just use the greater than operator to determine which elements of form are greater than 0.8 and then sum the result. Note that

 form>0.8

will return an array of 0s and 1s where each zero indicates that the element in form at the index where the zero appears is not greater than 0.8, and each one indicates that the element in form at the index where the one appears is greater than 0.8.

For example, if

 form = [0.1 0.9 0.83 0.33 0.22 0.44 0.99];

then

 form>0.8

returns

 0     1     1     0     0     0     1

as expected. Then just sum the result as

 NormalCount = sum(form>0.8);

For your abnormal count, just do

 AbnormalCount = sum(form>=0.31 & form<=0.5);

[menna gamal]

no, i need to take the value of form factor from the equation which i wrote in code i solve it , i make 1-D array then save form factor values in the array, and then use Sam () to check the ranges , Thanks.

[Image Analyst]

Not sure why you say "no" - that's exactly what he did.

Anyway, it looks like you have not seen my Image Segmentation Tutorial where I show you how to get circularities and do other things like size filtering. Your first loop can be done without loops like this

allAreas = [S.Area];
allPerimeters = [S.Perimeter];
formFactors = (4*pi*allAreas) ./ allPerimeters .^ 2;
numRound = sum(formFactors >= 0.8);
numNonRound = sum(formFactors < 0.8);
message1 = sprintf('Number of normal cells: %d\n', numRound );
message2 = sprintf('%s\nNumber of Abnormal cells: %d', ...
message1, numNonRound);
uiwait(helpdlg(message2));

[menna gamal]

i say no because he create the array and put the numbers in it and this not i want , i do the same you do and work with the loop and give me the same result from your code .. i do this :

S =regionprops(Label,'area','perimeter');

data = zeros(num, 1);

for i=1:num x(i)=S(i).Area;

    y(i)=S(i).Perimeter;

    %Form factor=4* pi * area/ (perimeter* perimeter) 
    form(i)=4.*pi.*x(i)./(y(i).^2);

    data(i)=form(i);

end

count1 = sum(data(:) >=0.8);

count2=sum(data(:)<0.8);

message1 = sprintf('Number of Normal cells: %d',count1);

message2 = sprintf('Number of Abnormal cells: %d ',count2);

 msgbox({message1,message2});

[Image Analyst]

You said "i do the same you do" - however I did not do the loop and in fact I showed you the vectorized way to do it. If you'd rather do it your looping way, that's fine - it doesn't matter to me. I was just showing you the more "MATLAB-ish" way that most experienced MATLAB programmers would do it.

At least you took a little advice from Geoff on how to do the sum (so you can go ahead and officially "Accept" his Answer, and vote for it, to give him reputation points.) By the way, you're creating an array too, just like Geoff said and which you said you did not want. What do you think "form" and "form>0.8" are? They are arrays! Both of them!

[Nicholas Prebble]

When I do this I recieve the error message "The logical indices contain a true value outside of the array bounds."

Error in practice (line 3)

NormalCount = sum(form>0.8);

[Image Analyst]

Nicholas, attach your m-file and image file so we can reproduce it and fix it. Preferably in a new thread.