|
Reconstruction Challenge!
Three
sets of data are available to download, and reconstruct using your best
algorithms. Images will be compared at the workshop by 4 radiologists,
and common imaging metrics will also be calculated for the submissions.
This will be presented at the workshop, and all but the "best"
submission will remain anonymous.
Images must be submitted by February 16, via an email
jim.pipe@chw.edu which includes:
1. Contact info for questions about your data set.
2. A brief (10 word) qualitative description of your method, e.g.
gridding, then 1D deblurring, then secret nonlinear filter
so that we can generically identify recon algorithms.
3. Provide the total reconstruction time (1 significant digit is fine)
and the platform used (hardware/software).
4. Include the submitted data in the required format; if the data file
is too large for email (e.g. for the PR 3D Angio),
indicate this and instructions will be given for uploading to a web
site.
Participants may choose to reconstruct one, two, or all three of the
following data sets: enclosed PDF files indicate more detailed
instructions.
Some of the data may not be available until Jan. 24; please check
back if it is not there.
1. PR 3D Angio: PR data were simulated using a 64-slice CT data with the
bone removed (collected temporal resolution of 0.8 seconds)
of a contrast bolus passing through intracranial vessels. Original
CT data were 384x384 x 64 slices x 46 time points
(collected over 36.8 seconds). Users will submit MIP time
courses as well as a 3D data set of the arterial phase image.
2. Spiral Blur Simulated Data: Spiral data were simulated using an T2
data set of 19 images and 19 corresponding (collected) field maps.
Users will submit deblurred images.
3. Spiral Coronary Data (actual Data): Spiral MRI data of coronaries are
provided. Users will submit reconstructed images.
|
|
~ 
