Title : Compression for 3-D Medical Images, with special concern on
Virtual Colonoscory

Name : Salih Burak Gokturk, gokturkb@stanford.edu

Description : 
The medical images consist of three dimensional volumes. An abdomen CT
scan contains 512x512x300 voxels in average, where each voxel is coded by
16 bits. In other words, the medical information for one individual is
150MB. This fact shows the obvious necessity of compression in medical
images. On the other hand, the number of recent papers in literature is
too few, and a generic compression method has not been set up. 
   The main property of medical images is smoothness. The amount of change
inter and intra slices is really small. On the other hand, this change
spreads to the whole image, as opposed to a little portion in movies, so
MPEG is probably not the right choice for medical images.  In this
project, I will be concentrating on applying the common compression
methods to 3-D medical volumes. These will include lossless coding[3],
vector quantization, transform coding and if time permits wavelet
coding[2].

I would like to focus mainly on colon images. Virtual colonoscopy is a new
branch of science for the computer aided detection of colon cancer. This
branch of science started first  by visualization. The next and current
step is automated detection of colon cancer[1]. I feel that in the near
future, a need for compression will occur. The final step of my project
will be a knowledge based compression scheme. For this purpose, a
radiologist will be contacted and asked about the redundant portions of CT
abdomen volumes for virtual colonoscopy. Next, a compression scheme that
uses this information will be applied.

Tentative Schedule :
Nov.6 - Nov.13 : Gathering the medical data. Formating the data into a
matlab readable format. Application of lossless coding, Vector Coding, DCT
based transform coding.
Nov.13-Nov.20 : Meeting with the radiologist. Design of of knowledge based
transformation. Applying wavelet based transform (if time permits)
Nov.20-Nov.27 : Application of the knowledge based compression scheme. 
Nov 27 Dec1 : Report and presentation generation.

References:
[1] A graph Memhod for conservative detection of polyps by Salih
Burak Gokturk, Carlo Tomasi, Proceedings of 2nd International Symposium on
Virtual Colonoscopy, Boston, 2000,page 105.
[2] Coding of 3D medical images using 3D wavelet decompositions by
Baskurt, A.; Peyrin, F.; Benoit-Cattin, H.; Goutte, R. Acoustics,
Speech, and Signal Processing, 1993. ICASSP-93., 1993 IEEE
International Conference on Volume: 5 , 1993 , Page(s): 562 -565
vol.5
[3] Information preserved guided scan pixel difference coding for
medical images by Takaya, K.; Tannous, C.G.; Li Yuan WESCANEX
95. Communications, Power, and Computing. Conference Proceedings., IEEE
Volume: 1 , 1995 , Page(s): 238 -243 vol.1