calculate the volume from slices of image

I have segmented 13 images using for loop in the listbox. How can I calculate the volume of the segmented image by adding up the area of each individual segmented image?

Answer 1

[Walter Roberson]

If you stored all of the data into one 3D array, then use nnz() to get the number of cubic pixels. Multiply that by the x resolution times the y resolution times the z resolution to get real-world volumes; the x and y resolution are usually the same but the z is often higher. (I am using resolution in the sense of distance between centers.)

[Pui Choo Wong]

how do I save into the 3D array?

[Walter Roberson]

%distance between pixel centers
x_dist = 0.3;
y_dist = 0.3;
z_dist = 0.75;

projectdir = 'the/directory_that/the/images/are_in';

selected_image_idx = get(handles.listbox1, 'Value');
all_image_names = cellstr( get(handles.listbox1, 'String') );
selected_image_names = all_image_names(selected_image_idx);

num_selected = length(selected_image_names);

for K = 1 : num_selected
  this_name = selected_image_names{K};
  this_image_file = fullfile( projectdir, this_name );   %presuming the stored name includes the extension
  this_image = imread(this_image_file);
  imshow(handles.axes1, this_image);      %original image
  title(handles.axes1, ['Original image: ', this_name ];
  title(handles.axes2, 'Segmentation being computed');
  imshow(handles.axes2, nan(1,1,3));      %blank output while we compute segmentation
  drawnow();
  segmented_image = segment_one_image(this_image);  %a routine you provide
  imshow(handles.axes2, segmented_image);
  title(handles.axes2, ['Segmented image: ', this_name ];
  drawnow();
  image_stack(:,:,K) = segmented_image;
end

volume_of_image = nnz(image_stack) * x_dist * y_dist * z_dist;

set(handles.text2, sprintf('%.2f', volume_of_image) );

[Pui Choo Wong]

Thank you!

[George Papas]

Can I ask a question? Do you need to multiply by the pixel centers or the pixel size instead?

If you will multiply by the pixel center then you get half the pixel size, is that right or I have not understood something correctly?

[Walter Roberson]

You should use the distance between pixel centers in each direction.

Pixel "size" could in theory be smaller than the distance between pixel centers. In theory a pixel could be (for example) 1/10 mm wide sensor area with the sensors set (for example) 3/10 mm between centers.

   -   -   -   -
  |+| |+| |+| |+|
   -   -   -   -

Sensors are seldom literally edge to edge -- too much risk of charge leak for photograph purposes. So you might sometimes see information about how big each individual sensor is, but each sensor "represents" the area out to half way to the next center, so you need to know the distance to the next sensor rather than the individual surface area.

... More to the point, though, is that DICOM header files for MRI images (and other pixel-oriented images) are coded in terms of the distance between centers. In the above diagram, that would be the distance between the center of the + symbols. Whereas if you talked about the "distance between pixels" then you could be referring to the size of the gap between boxes.

[George Papas]

Brilliant response, many thanks. So that means that the DICOM header, when it refers to pixel size, it gives the distance between centers. So, what I understand is that I can just use the pixel size for x, y and z as it is. Thank you indeed!

[Walter Roberson]

http://dicom.nema.org/medical/Dicom/2016b/output/chtml/part03/sect_10.7.html

Pixel Spacing; Tag (0028,0030); type 1C

Physical distance in the patient between the center of each pixel, specified by a numeric pair - adjacent row spacing (delimiter) adjacent column spacing in mm. See Section 10.7.1.1 and Section 10.7.1.3. Required if the image has been calibrated. May be present otherwise.

[George Papas]

Thank you, interesting information. I guess one should really take care of what dicom header can contain for different data.

Answer 2

[abdey munaf]

can I know how the distance between pixel center is measured?

[Walter Roberson]

DICOM images support multiple image types .

In some of the cases there are physical sensors at fixed distances (that might differ between manufacturers and models ) such as in the case of ccd arrays used for photography . Xray photograph are example .

In other cases there might only be one fixed detector but there might be some process going on that has the effect of selecting a particular 2d area or 3d volume, in which case the distance recorded reflects how finely controlled the selection was . MRI is an example of time based signals converted to spatial by knowing how signals were controlled in continuous space.

In other cases there might have been raw data taken at multiple overlapping angles and what is recorded in the image file might reflect how finely grained the angled measurements were converted to rectangular coordinates . CT is an example of angle to rectangular reconstruction .