Progressive medical image compression using a diagnostic quality measure on regions-of-interest

Dealing with lossy compression of medical images requires particular attention whether for still images, video or volumetric slice-sets. In this work we propose an approach based on a selective allocation of coding resources that is directly related to the diagnostic task. We introduce the concepts of Region of Diagnostic Interest (RODI) and Diagnostic Quality as key links between the radiological activities and responsibilities and the functioning of a selective coding algorithm. The coding engine is a modified version of Shapiro's EZW algorithm and the coded bit-stream is fully progressive. The RODI selectivity corresponds to the choice of a set of subband weighting masks that depends on a small set of parameters handled and validated by the radiologist in a very natural manner. In conclusion, we present some experimental results that give interesting insights in favor of using lossy compression in a controlled fashion by a competent physician.