A clinically relevant approach to medical image compression.

摘要

As medical1biological imaging facilities move toward complete filmless imaging, compression will play a key role. Although lossy compression techniques yield high compression rates, the medical community has been reluctant to adopt these methods, largely for legal reasons, and have instead relied on lossless compression techniques which yield low compression rates. The true goal is to maximize compression while maintaining clinical relevance and balancing legal risk. It is important to figure out how to preserve the quality of the image at a level that is necessary for both what is clinically relevant and what doctors and lawyers feel is necessary for liability reasons, a criterion that may transcend what is visually perceptible. This dissertation proposes a novel model-based compression technique that makes use of clinically relevant regions as defined by radiologists. Lossless compression is used in these clinically relevant regions with high quality, and lossy compression is used everywhere else (that is, in the areas that are not clinically relevant). The clinically relevant regions are predefined on standard film types in an atlas.