Predictive Coding Integer-Based Wavelet Transform Approach to Diagnostically Lossless Medical Image Volume Compression

Homogeneous data in radiological image databases consumes an extraordinary amount of storage space. An example of this is the multiple voxels representing different cross sections of a body part being imaged, which are referred to as volumetric data sets. Lossless compression algorithms are imperative for efficient storage and transmission of volumetric data sets. This paper presents a diagnostically lossless medical image volume compression algorithm that achieves better compression than the other compression algorithms typically available on computer systems. Our algorithm is a combination of predictive coding and integer wavelet transforms plus a special preprocessing procedure to remove noise outside the diagnostic region in medical image volumes. This volume compression improvement is demonstrated with MR (magnetic resonance) brain images. Our algorithm results in a compression ratio of around 3:1 for a set of medical images.