SCE Carleton Logo
  Carleton > Engineering > SCE > Faculty > A. Adler > Courses > BIOM5200 > Assignment 3


BIOM 5200 - Assignment #3

  1. Display of Medical Image Data Software for viewing of medical images typically provides a "windowing" function. Windowing performs a linear stretching of the image histogram, in order to better view pixels with values in the "window". For example, if the screen pixels have intensities in the range 0 to 255, a windowing from wmin to wmax will convert a pixel f to a value of 255 × (f - wmin ) / ( wmaxwmin ) on the output display. Details of f above or below this range are eliminated and f is set to a displayed value of 255 or 0, respectively. Windowing serves two important functions: 1) it enhances the visibility of contrasting tissue regions, 2) it allows selection of the tissue type to be viewed (for example, with CT images, windowing can allow viewing of soft tissue while bone regions are "cropped" to the maximum or minimum value)

    1. Describe how "windowing" allows these goals to be achieved

    2. Describe how the visibility of contrasting regions in an image varies with the size of the contrasting region and the SNR.

    3. Describe how "windowing" interacts with the variation in visibility with SNR. Considering these effects, what image conditions (in terms of contrast size and SNR) benefit most from "windowing"?

  2. Nuclear Medical Imaging
    Consider the SPECT system of the figure below. Six triangular regions are defined (R1 to R6) from which twelve projections measurements are made (P1 to P12) using an Anger-type camera. Each trianglular region will emit the same number of photons in each of the six possible directions. P1 is aligned with collimator hole H1, and P2 with H2, respectively. Holes H0, H1, H2, and H3 are at x,y positions of (0,0.3), (0,0.5), (0,0.7), and (0,0.9), respectively. Photomultiplier tubes PMT1, PMT2, and PMT3, are at x,y positions (0,0.2), (0,0.6), and (0,1.0), respectively.

    Figure: SPECT camera system with object and detector
    1. Why is resolution of Anger camera better than the spacing of the detectors? Can this approach to improve resolution be used in an X-ray system?
    2. For a single SPECT event, the measured signal is
      1 6 mV
      2 14 mV
      3 1 mV
      What is the y position of the event at the detector?
    3. There are 2000 units of activity/second in R1. All regions have an attenuation μ = 0.15/cm. Consider that each region is 3 cm across (independent of the direction of the X-ray beam). Do not consider any attenuation of the X-ray beam in the originating region. Calculate the projection data P1 to P12.
    4. Using the algebraic reconstruction technique (ART), calculate the reconstructed values in each region. You may solve this question using software to calculate ART (ie you don't need to show all the steps)

  3. Grey scale adjustment
      Download this chest X-ray from the Medpix Database. For a complete description of this case (not required), this is medpix case ID #994. This image may be displayed as follows in Octave or Matlab™
          im_in= imread('synpic994.jpg');    % Load image
          im_in=double( mean(im_in,3) ) + 1; % Get grey component
          colormap(gray(256));               % Set Grey colormapping

      Greyscale histogram modification
      In order to better view the greyscale values in the image, we wish to stretch the intensity map to cover 90% of the pixel histogram range.

    1. Develop a Octave or Matlab™ function:
      function im_out= im_normalize(im_in, thresh_l, thresh_h)
      which accepts an image im_in as input and calculates a normalized image im_out for which pixel values below the thresh_l percentile are set to 1, pixel values above the thresh_h percentile are set to 256, and other values are distributed linearly between 1 and 256.

    2. Calculate the histogram of the image using the function hist.
      • Calculate the threshold values tL and tH such that 5% of pixels have values below tL and 5% of pixels have values above tH.
      • Calculate and display the windowed image based on thresholds tL and tH.

    3. Select a region (of size 100×100 pixels) corresponding to the lungs (the two dark triangular regions). Calculate the histogram of this region.
      • Calculate the threshold values tL and tH of the lung region such that 5% of pixels have values below tL and 5% of pixels have values above tH.
      • Calculate and display the windowed image (of the complete image, not the lung region) based on thresholds tL and tH.

    4. Compare the images from the previous two questions. Comment on any features that are more or less easily visible.
    5. Use the function histeq to implement histogram equalization for the source image. Describe the image produced. Does the output appear more "artificial" than the previous images? If so, in what way?

Last Updated: $Date: 2007-03-17 12:56:52 -0400 (Sat, 17 Mar 2007) $