Title:  Compression of Real-Time Cardiac MRI Video Sequences

Students:  	Julie Sabataitis		Neal Bangerter
		julie@quench.stanford.edu	nealb@stanford.edu
		725-5638			725-7005

Description:

Recent advances in MRI technology have allowed real-time acquisition
of cardiac MRI images at a rate yielding up to 18 reconstructed
frames per second.  There is some interest in real-time compression
of these image sequences for transmission to a remote viewer.  A
standard cardiac study typically produces 128x128 pixel 8-bit
images at a rate of 16 frames per second. [1]  We propose to implement
a simple motion compensated prediction compression algorithm, and
then attempt to optimize the motion prediction for these
cardiac sequences by using a variable block size for the motion
prediction.  We wish to generate rate-distortion curves for both the
original and optimized motion compensated compression algorithms.
We then propose to investigate above which point on the
PSNR curve we need to be in order for the diagnostic utility of
the image sequences to be preserved.  We will determine this point
experimentally by showing reconstructed image sequences at various
points on the PSNR curve to three cardiologists.  Comparing the
position of this point on each of the two PSNR curves will give us
an idea of whether we have reduced or increased the diagnostic
utility of the image sequence at a given PSNR as a result of our
optimizations.


1. "Real-Time Color Flow MRI", Nayak, K.S., Pauly, J.M., Kerr, A.B.,
Hu, B.S., and Nishimura, D.G., Magnetic Resonance in Medicine, 
February 2000 43:251-258