|Title:||Compression of high resolution 1D and 2D NMR data sets using JPEG2000|
|Authors:||J. C. Cobas, P. G. Tahoces, I. Iglesias Fernánez|
|Reference:||Chemometrics and Intelligent Laboratory Systems, 91, 141-150 (2008)|
High resolution NMR (HR-NMR) techniques, such as high-throughput NMR and fast multidimensional experiments generate large amounts of data which might result in major problems relating to data storage and communication bandwidth. Data compression techniques reduce data volumes and facilitate the analysis and remote access of HR-NMR information. In the context of high performance image compression algorithms, the Wavelet transform has shown excellent compression performance when dealing with natural images.
In an earlier publication, we have demonstrated that this kind of lossy compression procedures can also be successfully applied to multidimensional HR-NMR data. In this paper we report the results obtained when applying JPEG2000, the latest international standard for image compression, to 1D and 2D HR-NMR spectra. Exceptionally high compression ratios up to 900:1 applied to several NMR spectra have been obtained.
A new index that we called MSE_95 (Mean Squared Error with 95% of confidence) is proposed to set the limits of lossy compression. This index is based in the differences between original and decompressed data sets and the noise of the original data set. MSE_95 establishes objective limits to achieve the maximum compression that preserves the qualitative and quantitative information of 1D or 2D NMR data sets.