Min-max compression methods for medical image databases

The volume of medical imaging data produced per year is rapidly increasing, overtaxing the capabilities of Picture Archival and Communication (PACS) systems. Image compression methods can lessen the problem by encoding digital images into more space-efficient forms. Image compression is achieved by reducing redundancy in the imaging data. Existing methods reduce redundancy in individual images. However, these methods ignore an additional source of redundancy, which is based on the common information stored in more than one image in a set of similar images. We use the term "set redundancy" to describe this type of redundancy. Medical image databases contain large sets of similar images, therefore they also contain significant amounts of set redundancy.This paper presents two methods that extract set redundancy from medical imaging data: the Min-Max Differential (MMD), and the Min-Max Predictive (MMP) methods. These methods can improve compression of standard image compression techniques for sets of medical images. Our tests compressing CT brain scans have shown an average of as much as 129% improvement for Huffman encoding, 93% for Arithmetic Coding, and 37% for Lempel-Ziv compression when they are combined with Min-Max methods. Both MMD and MMP are based on reversible operations, hence they provide lossless compression.