Select specific area of a DICOM file

I am trying to select the area between the two circles, everything inside the inner circle and outside the outer is not of interest. This is how I am trying to "activate" and "deactivate" the pixels I want. By doing it twice, I am able to select part of the left and part of the right side, and then just add them up. I would like to finish with the whole area between the circle with the pixels "on". Probably this is not even the way that I should be doing it but if anybody has a recommendation I would be happy to hear! (cant upload dicom file so I will compress in zip)

also, I will take the absolute value of the selected area so no worries about it changing color from the first one to the last one.

clear; clc; clf; close all;

I(:, :, 1) = dicomread('MRIm1.dcm');

P = I(:, :, 1);

figure

subplot(2,2,1), imshow(P, [])

%%

Pb = imbinarize(P, graythresh(P));

se = strel('line', 4, 4);

Pb = imerode(Pb, se);

props = regionprops(Pb, 'Area', 'PixelIdxList');

[m, index] = max([props.Area]);

Pbm = zeros(size(P,1), size(P,2));

Pbm(props(index).PixelIdxList) = 1;

P(~Pbm) = 0;

%same thing again but with absolute value so that it inverts and takes right side

A = I(:, :, 1);

A = abs(A);

Ab = imbinarize(A, graythresh(A));

% Ab = bwareaopen(Ab, 250);

% Ab = imfill(Ab, 'holes');

se = strel('line', 4, 4);

Ab = imerode(Ab, se);

props1 = regionprops(Ab, 'Area', 'PixelIdxList');

[n, index1] = max([props1.Area]);

Abm = zeros(size(A,1), size(A,2));

Abm(props1(index1).PixelIdxList) = 1;

A(~Abm) = 0;

C = P + A; %just adds them

subplot(2,2,2), imshow(P, [])

subplot(2,2,3), imshow(A, [])

subplot(2,2,4), imshow(C, [])

The goal is to complete the bottom right one.

[Rik]

Are you allowed to assume this is a perfect circle?

[Alfredo Scigliani]

Let's say yes, I am

Answer 1

[prasanth s]

remove noise using median filter

IM=medfilt2(P);

find gradient

[Gmag,Gdir] = imgradient(IM);

then apply circle detection or any other methds to separate the circle

[Alfredo Scigliani]

Thanks for the suggestion, but it affects the pixel intensity of the selected area. I would like to keep them as initial.

This is how it looks after filter.

See the difference:

[Rik]

Why does that matter? You only need to use the filtered image to detect the circle. Once you have a mask for the circle you can select the correct pixels from the original image.

[Alfredo Scigliani]

How do I select the pixels after mask?

[Rik]

Logical indexing:

% load a builtin example image
load mri
im=squeeze(D(:,:,1,ceil(end/2)));
im=im2double(im);

subplot(1,3,1)
imshow(im)
title('original image')

[X,Y]=ndgrid(linspace(-1,1,size(im,1)),linspace(-1,1,size(im,2)));
R=sqrt(X.^2+Y.^2);
CircleMask = R>0.6 & R<0.8;
subplot(1,3,2)
imshow(CircleMask)
title('mask')

NewImage= 0.5*ones(size(im));
% copy data over with logical indexing
NewImage(CircleMask) = im(CircleMask);
subplot(1,3,3)
imshow(NewImage)
title('image with old values in mask')