CS6825: Computer Vision
Instructor Outline Syllabus Projects Links  
Main
Data Input
Research
Getting Started
Blob Detecton
Proposal
Implementation
Requirements
Data Output
Class and Menu Specifications
Other
Deliverables

Project 3: Feature Detection

    Blob-detection (2 cases ) and Features for Blobs..... to Get you Started (but you must find evidence in literature for them) .....WE ARE ONLY DOING CASE 1 (MASSES)

    Some of the literature on Breast Cancer paritions finding cancer into the 2 following cases which we will concentrate on:

     

    case 1 (masses): the detection of masses (blobs)

    breast massSome general reasoning in the Masses (case 1) case is that somehow (you need to read the literature) the size, location, shape, density, etc. of the Mass can say if cancer and what kind (something again I want you to read about...there are different stages of cancer as well as saying no cancer-benign). You may find some other unique research looking at some very different feauters.

     

     

    case 2 (micro-calcifications): detecting small (much smaller than the masses) micro-calcifications (mini-blobs).

    microcalcificaton imageSome general reasoning in the micro-calcifications (case 2) case is that somehow the number, distirubtion, density, location, etc. of the micro-calcifications can say if it is cancer and what kind. However, you may find other theories looking at shape, etc of these tiny blobs...you do the research.

     

     

    Given our 2 cases ....here are some features you might think about. IMPORTANT: note that with each you must first find a potential mass or the microcalcifications. So, that blob detection comes first before you extract the features/attributes of each blob (or mini-blob). Here is a directory http://algebra.sci.csueastbay.edu/~grewe/CS6825/Projects/Project3_resources/ with some papers I am giving you to get you started. You CAN NOT use these as your 5 references you post.

    Here are some ideas you might see (and hopefully you will see much more) in your 5 research references and each of your classmates 5 references(remember we are sharing and commenting on ).

    For Cancer Represented as Masses (blobs)

    First must do Blob detection (via thresholding or segmentation or ....) then for each Blob collect following attributes/features of the blob

    • shape of blob - taken from boundary of blob
    • texture inside blob
    • size of blob
    • boundary shape - smooth versus spurious
    • density (intensity)

    For Cancer Represented as Microcalcifications (small tiny blobs)

    Note: not doing this for this project...only for your information

    First must detect and label mini-blobs (via thresholdig or segmentation or...) then collect the following feature information (note in this case the blobs may be too small for shape analysis done with masses)

    • number in image
    • distribution of in ROI
    • shape
    • size
    • density/variance



    NOTE: WE ARE ONLY GOING TO LOOK AT CASE 1 - detction and feature extraction of Masses

     

 
© Lynne Grewe