Improving backup strategies in large DICOM databases based on weighted image compression

Abstract

In this work, a new way of organizing medical images based on adaptive compression algorithms is presented. The main purpose is to define a workflow that optimizes image storage, which is a highly significant process in securing eHealth systems. A set of existent lossy and lossless compression algorithms have been chosen and adapted for supporting the DICOM standard; one of them has been developed by this group. Said algorithms were chosen and parameterized according to a combination of compression rate and critical information that an image could contain. Such information was computed either manually or through CNN trained models and stored as metadata. Several tests were carried out, mainly showing the compression rate compression error—measured using PSNR—and the performance of different proposed algorithms. A combination of such algorithms was run considering the inferred metadata in order to get the final metric. Finally, findings and considerations were summarized for being used in real backup systems.