Objectives

The huge amount of data involved by medical imaging systems needs a compression step before transmission, storage or manipulation. The task "Medical data management" targets:

1. compression algorithms
2. protocols

Our objective in subtask (1) is to provide the "best" compression algorithm with respect to the medical request and its constraints.

Subtask (2) is dedicated to medical streaming through two aspects:

• coupling compression and transfer allowing a dynamic adaptation of the compression ratio depending on the machine speed and the application usage;
• progressivity factory, i.e. adaptative access to invidual medical images in the context of interactivity.

First results

Several studies have been done in subtask (1) and have yielded the following issues and results for volumetric images:

• the best lossy compression scheme in terms of rate-distortion trade-off is based on 3D discrete wavelet transform;
• 3D SPIHT is one of the most efficient algorithms;
• CAD* (Computer aided detection) detection performance of solid lung nodules in 3D images is robust across all compression levels up to 96:1 (using 3D SPIHT) [1]. Furthermore, no significant effect due to compression has been detected on lung nodule volumetry [2]. This study is a collaboration with R2Tech.