Introduction to Imaging in Bioinformatics
Processing Images to Detect Breast Cancer
Prep readings:
In this assignment, we wish to:
In this assignment, you will be developing a program for a real-world bioinformatics application, where data storage and speed of performance issues are critical. The particular application you will study with this assignment is the detection of breast cancer. The images above were derived from the on-line Digital Database for Screening Mammography, located at the University of South Florida. These images were scanned from actual X-ray films taken of women being screened for breast cancer. Each patient has four images taken, providing two different views of each breast.
For example, consider the image and link below which indicate a case profile where a patient's right breast tissue contains a spiculated lesion. Click on the case profile and you can review where a radiologist has marked the extent of this abnormality in two different views (you are viewing the RIGHT_CC view below). The radiologist's assessment and any markings on the images are known as the "ground truth" for each case. The DDSM terminology page explains the terms a radiologist uses when he/she finds abnormalities such as this one.
Now let's think about how images are stored in a computer. If we considered a small section of the image above, say 5 rows and 8 columns and examined the data stored in an area of this size, it might look as follows:
160 106 100 102 111 100 127 120
107 110 110 110 127 242 107 100
150 110 106 132 213 213 120 148
100 117 115 108 100 234 210 254
120 106 111 105 154 222 148 170
For this example, the 5 x 8 subimage contains 40 different
locations with various intensity values.
The values at each location indicate the effect of the
X-rays at each underlying location in the breast tissue. With 8-bit
images such as this we have a 0..255 range of possible values at each
location.
In terms of analyzing X-ray films, intensity values close
to 0 appear dark and those closest to 255 appear bright. As you can
see from the images above, finding breast cancer means locating areas in
the image that are closer to the bright end of the intensity range.
In the example subimage provided above, let's assume that
the pixels with intensity
values greater than 175 represent suspicious regions which the radiologist
should examine further.
Across the country, thousands of images just like this are evaluated by clinicians each day, but with image sizes in the range of 5000 x 3000 pixels. Your goal in this assignment is to see how you can use the MATLAB code provided in the file above to develop a tool to help these radiologists. For example, one way you could help the radiologist to quickly locate the suspicious areas, would be to produce an image array containing just two shades of intensities, 0 and 255, where 0 represents a normal area and 255 represents a possible suspicious area. In real applications, this type of output is considered a "region-of-interest" (ROI) or "prompting" image, in that it directs the attention of the radiologists to the areas of the image requiring more careful analysis. This is because 255 stands out better against a bunch of 0 values, compared to the full 0..255 range. Given these objectives for the example subimage provided above, a new subimage could be the following:
0 0 0 0 0 0 0 0
0 0 0 0 0 255 0 0
0 0 0 0 255 255 0 0
0 0 0 0 0 255 255 255
0 0 0 0 0 255 0 0
How easy is it for you to spot the abnormality in this subimage, compared to
the original subimage above?
Getting Started with Bioimaging
Move the figure windows to different parts of the screen and identify the titles at the top of each figure which are labeled as follows:
% ??? Error using ==> edge (parse_inputs) % RGB images are not supported. Call RGB2GRAY first.Add the following MATLAB code after the call to the imread subroutine:
OriginalImage = rgb2gray(OriginalImage);
To get credit for this assignment, you need to complete these additional steps:
Wondering how far are you away from the "real world" with this assignment? Check out these products, approved by the FDA in your lifetime (within the last several years, in fact):
Feel free to check out how the other students
in the class are doing on the assignments.