CS6825: Computer Vision word cloud

Project 3: Feature Detection

Breast Cancer Feature Detection

Due  Nov. 11
(points 200)

Grading: 50 points = ave from survey, 150 points using Leveled Grading Fomula


    As a class you will collaborate to create a singe breast cancer feature detection system. Specifically, each person (or sub-group) will be responsible for detecting specific features, the class will also elect one or more people to integrate all of these feature detection components into a single GUI applicaiton.

    Each person/subgroup will be given their own unique survey score and the rest of the points will be evaluated by the instructor via binary grading formula and will be the same for everyone participating in the project.

     

    Data Input

    JPG images from the USF database. You should be able to load each image and display it and then process it. The user of the system can then select a region of interest and request the system to process it. It is only in the ROI that features will be detect. Note:

    • ROI = Region of Interest. Regions of Interest are sub-images that can be selected by the user by clicking the upper-left hand point and lower-right hand point of a rectangular area that is the sub-image. A GUI to capture this must be selected. A unique, sequently ROI # is assigned to this Region of Interest. It is intended that an ROI will be then analyzed by the system and a Data XML Output file for this ROI will be produced.

    Data Analysis (Feature Extraction) and Data Output

    You are to produce a single XML ascii data file for each ROI. If the ROI has a number of N, the filename will be INPUT_FILE_NAME_ROI_N.xml. This file will represent all of the features detected in a given ROI. The format of this data will be determined as part of your design process and described in an XML DTD file as well as in your web documentation. You will need to specify the image file name, the ROI, number and type of each feature and details of each feature. When calculating each feature, you will not only specify it but, also give an certainty metric on a scale of 0 to 100% indicating how true this feature is. Finally, for each feature you will give a metric on a scale of 0 to 1 indicating the importance of this feature. Note:

    • Feature Certainty = this metric should be calcuated automatically by the program and a function of the certainty of the pixels used in its calculation. The certainty of a pixel represents the certainty that a pixel belongs to the tumor under consideration instead of belonging to the non-tumor or background part of the ROI.
    • Feature Importance = this metric may be calculated automatically but, when this is not feasible can be determined by a human operator. In the later case, the feature, its location, its value should be given in a pop-up box and a prompt for a "feature importance" value from 0 to 1 requested of the operator as their input.

    Features

    Masses

    • shape
    • texture
    • size
    • lessions - stellate pattern (linear edges)
    • location?

    Microcalcifications

    • number of in ROI
    • distribution of in ROI
    • shape
    • size

    SPECIALIST:    Lynne Grewe

    LABS

     

    PAPERS

    CONSIDERATONS

    1. Convert images to JPEG (use batch program)
    2. Design Architecture/ Class structure
    3. Develop ROI extrator GUI
    4. Feature 1: Area
    5. Feature 2: Shape
    6. Feature 3: Spiculation- Stellate Pattern?
    7. Feature 4: Texture
    8. Feature 5: Boundary Sharpness (Margin Sharpness)
    9. Feature 6: Boundary Irregularity-Measure of Smoothness (same as feature 5???....related to feature 3)?

     

    Implementation Details

    You will select a supervisor from your group to monitor progress on the project. Everyone must partcipate in all phases of this projects. Also, EVERYONE must be able to understand and reproduce the code created if necessary. If you do not individually complete the portion of the work allocated to you as stipulated by the supervisor, you will not be given credit on that phase of the project AND this work must then be redistributed by the supervisor to the rest of the class.

    Attendance throughout the class periods we will work on this project is CRITICAL and NECESSARY for credit on this project. ONLY absences excused by written doctor reports or other accepted excuses (see instructor) will be allowed and in this case a different individual project will be assigned to you.


Deliverables

  1. HTML paper on how you designed your Breast Cancer Feature Detection Program.  Describe your approach and why you decided to take it.

    Upload HTML w/images, etc. to one person's server account. Print out this HTML paper and write on top the URL of its location. YOU MUST create your own test sequence, in addition you must test using at least 4 cases from the database.
  2. Fully comment and test out program.
  3. Turn in USB keyfob/CDwith all java code and compiled class files needed to run the program. IF YOU use additional classes not in standard Java, you need to have all of the files here...make a JAR file.
  4. Printout of code, fully commented, with YOUR NAME ON TOP, and INSTRUCTIONS ON HOW TO RUN THE CODE FROM YOUR USB keyfob/CD
  5. A one-page description of how code is structured and the state of how it works.
  6. A one-page description of each persons contribution to the project. If you have 3 people in your group this page should be formated by having 3 sections, each with a different person's name as the header. Under each section, describe that person's contributions to the design, code, and documentation development.
  7. Print outs of screen shots of program working showing the results of: image loading and display of image sequence before and after processing.

© Lynne Grewe