Proposal: About:
The field of radiology is based entirely on human judgement. As such, the diagnosis of medical images is highly subject to human error and inability. A particular area of difficulty is the analysis of MRIs taken of the knee. Problem areas my be missed by the radiologist, or a radiologist may not be able to distinguish different areas of the MRI. Also, problem areas may be invisible to the human eye.

Solution
The "Giraffe" Project aims to solve these problems by automating the diagnosis of knee MRIs through computer-aided image analysis. The program will scan the many MRI images simultaneously, linking subject-matter of the images, and identifying the potential problem areas. The problem areas will then be "highlighted" and outputted to the user. Severity prediction will be provided using several "discrepancy" algorithms. "Giraffe" Project also provides users with a "manual scan" function, allowing radiologists to distinguish subtle brightness differences in the MRI image. Subject estimation will also be outputted to the user, pointing out the different bones, tendons, etc. based on gradient, given the subject-matter is indistinguishable.

The program will be written in Java.
Scanning procedure will consist of using a pixel-by-pixel scan in combination with identification algorithms. Basic scanning will be completed witht he Java 2D API. Given that 3D scanning is needed (when images are linked), the program will switch to the 3D API class.

mentor is Dr. Puneet Ghei and Martin.

Team Members:

  Alan Hshieh
  Lydia Hammond
  Julia Silva

Sponsoring Teacher: Gregory Marez

Mail the entire Team