We have tested a much more efficient approach that calculates a noise threshold based on statistics of the image intensity distribution, and saves information only on the supra-threshold pixels. This method is no longer lossless, but does reduce file sizes to less than one-quarter their original size for typical brain images, depending on the field of view. Unfortunately, it is not acceptable for functional neuroimaging, as it does not properly handle changes in head position during the scan series. Finally it is worth noting that publicly available lossy compression systems such as MPEG and JPEG are unsuitable for quantitative statistical analyses as commonly performed for functional imaging studies [8].

We believe that a rapid, simple and efficient means of lossless data compression will be of value to the imaging community. We have therefore made the source and executables available at our web site, www.brainmapping.org for public download. At the moment, we have source code available for Macintosh and console driven ANSI C environments that will support a variety of standard image formats including Mayo Clinic "Analyze", Advanced NMR Systems APD2 and the ".bshort" file type developed at Massachusetts General Hospital. We encourage interested users to download, modify, and send back to us any improved versions of the source and any desired extensions to other formats.

UCLA Brain Mapping Center