Biology Computational 03-310/510/710, Spring 2004
R. F. Murphy
Problem D1 Finding objects in images Due: April 30, 2004 (accepted without penalty until May 11)
You should use NIH Image or ImageJ public domain image analysis packages, and/or Matlab, for this assignment. You will also need to download file(s) from the Homework Problems web page.
Extra Credit Questions (10 points each)
1. Open file giantin2.tif. This is an image from a fluorescence microscope where light regions (high pixel values) indicate the presence of fluorescence and the background is dark (low pixel values). The image is of cells that have been stained for the Golgi protein giantin. Invert the image so that the background is white and the regions with fluorescent dye are black. Threshold the image so that only the bright major objects are visible. Now use the Analyze Particles option to identify individual objects in the image that have a minimum size of 2 pixels and a maximum size of 10,000 pixels (check the box for Label Objects). You should obtain 7 objects if your settings are correct. (a) What threshold value did you use? (b) Print the thresholded, analyzed image. (c) Print the list of objects. 2. You are given files E1first.tif and E1second.tif and told that they contain fluorescence for images different proteins. (a) Print an inverted image of just those organelles that contain both proteins (i.e., have pixels that are above threshold in both images.) (b) Describe how you created it. 3. Use the Analyze Particles option to find objects in the E1first.tif image. Save the list of object measurements to disk. Repeat for E1second.tif. Use a spreadsheet program such as Excel to import the object measurements. Print a scatter plot of mean intensity vs. object area for both images on the same plot, using different colors or symbols for the objects from the different images. Do not print the lists of objects! 4. Train and test a classifier to distinguish objects like those in E1first.tif from those in E1second.tif. 5. Design a kernel that can be used to detect diagonal edges of slope -1 (i.e., running from upper left to lower right). Apply it to the Angles.tif image. Can you use it in combination with thresholding to make an image that shows (in black against a white background) only the edges of those objects in the image whose slope matches? Submit a printout of the kernel, a printout of the black on white image, and a description of the operations you used to create it.
Acknowledgments
Thanks to Michael Boland for providing images.
Page 1 of 1