NMR data compression by Principle Component Analysis method

One basic problem encountered in Nuclear magnetic resonance (NMR) logging is the vast amount of data that has to be analyzed, which requires plenty of computer memory and is also very time consuming. Hence, it becomes indispensable to develop stable and fast NMR data compression methods of high compression ratio which could improve the NMR data inversion velocity without losing the accuracy. This paper adopted Principle Component Analysis (PCA) to compress 1D and 2D NMR data under a high compression ratio [1] , then inversed the 1D original and compressed NMR data, as well as the 2D compressed NMR data with normal smoothing inversion method [2] . The T2 distributions inversed from 1D original and compressed NMR data were compared with the modeled T2 distribution, as was the D-T2 map inversed from 2D compressed NMR data with the modeled D-T2 map. We also recorded the time cost by 1D and 2D NMR data compression respectively. The results showed that the compressed NMR data got from PCA compression method under high compression ratio was almost lossless and could be inversed to a satisfied T2 distribution as well as D-T2 map, besides, this method was much less time consuming and would perform more prominent advantages in multi-dimensional NMR data compression.
Acknowledgements
This project was funded by the National Natural Science Foundation of China (41272163) and the National Natural Science Foundation of China-China National Petroleum Corporation Petrochemical Engineering United Fund (U1262114).

Figure 1: T2 distributions from modeled, original and compressed 1D NMR data.

Figure 2: modeled D-T2 map and D-T2 map from compressed NMR data.