Transcript caddlab - the UNC Department of Computer Science

CADDLAB
Medical Imaging on
Remote Compute Servers
Introduction to the Project
The Department of Radiology has servers
that store 3D medical images (124,000!)
Web-based system known as PACS
Project designed to view images over web,
point and click processing, command line
programs
Authorization and security already handled
eXtreme Programming!
Domain Problems
In past, clinicians had to download whole
3D image before being able to process
Excessive download times and wasted
resources
Because of the large file sizes, clinicians
need an efficient way to view and process
the images online
Key Use Cases
After logging in, the user will see the main page.
Here he/she will choose an image and select
one of two commands
The View Command will allow the user to view
and/or select points on the image for
processing.
The other command is the Blur Command that
allows the user to run a process on their
selected image. This processing is done on the
server.
Future commands will be added, such as
registering and segmentation
Interfaces
A Web site to choose images and the
command to run
A viewer (that we will demo)
A Blur command page
A Login Page (for now)
Architecture
Repository architecture built around a
database of 3D medical images
Allows for the
addition of future
commands
through the
main page
Key Technical Problems
Slow connections for downloading 40+ MB
files
Figuring out the technology to use

We searched on google.com, found an applet
called JIV
We are not allowed access to their servers
because they were just hacked
Clinicians’ older computers
Team’s slow laptops
Technologies
The viewer is a Java applet based on a
program we found called JIV
The website will use html and php
(tentative)
Image converting, connecting with the
database
Client will take care of security