Adaptively Optimized Combination of Phased-Array MR Spectroscopy Data Without Noise Correlation Estimation: A Joint Diagonalization Method.

The adaptively optimized combination (AOC) method optimally combines multichannel magnetic resonance spectroscopy (MRS) data by incorporating noise correlations among receive coil elements/channels. These correlations are typically estimated from noise samples extracted from the fully decayed tails of acquired free induction decays (FIDs). To eliminate the need for FID noise tails and the corresponding interchannel noise correlation estimation, we propose a novel coil-combination method, termed AOC-JD, that exploits the intrinsic multichannel noise information contained in the differences between individual repetitions and the repetition-averaged spectroscopic data. The proposed method is based on joint diagonalization (JD) of two correlation matrices constructed from the individual repetitions and the repetition-averaged multichannel MRS data, thereby circumventing the conventional interchannel noise correlation estimation procedure. The performance of AOC-JD was evaluated using both computer simulations and in vivo human brain experiments. The results demonstrate that, when multichannel FIDs do not contain valid or sufficiently long noise tails, AOC-JD significantly improves the signal-to-noise ratio (SNR) of the combined spectra compared with the conventional singular value decomposition (SVD) method. Moreover, its performance is comparable to that of the original AOC method, which is optimized to maximize the SNR of the combined spectrum, when applicable. These findings demonstrate that AOC-JD provides a theoretically grounded and reliable approach for combining multichannel MRS data when FID noise tails are unavailable or insufficient.