Temporal SNR optimization through RF coil combination in fMRI: The more, the better?

PLoS One. 2021 Nov 8;16(11):e0259592. doi: 10.1371/journal.pone.0259592. eCollection 2021.

Abstract

For functional MRI with a multi-channel receiver RF coil, images are often reconstructed channel by channel, resulting into multiple images per time frame. The final image to analyze usually is the result of the covariance Sum-of-Squares (covSoS) combination across these channels. Although this reconstruction is quasi-optimal in SNR, it is not necessarily the case in terms of temporal SNR (tSNR) of the time series, which is yet a more relevant metric for fMRI data quality. In this work, we investigated tSNR optimality through voxel-wise RF coil combination and its effects on BOLD sensitivity. An analytical solution for an optimal RF coil combination is described, which is somewhat tied to the extended Krueger-Glover model involving both thermal and physiological noise covariance matrices. Compared experimentally to covSOS on four volunteers at 7T, the method yielded great improvement of tSNR but, surprisingly, did not result into higher BOLD sensitivity. Solutions to improve the method such as for example the t-score for the mean recently proposed are also explored, but result into similar observations once the statistics are corrected properly. Overall, the work shows that data-driven RF coil combinations based on tSNR considerations alone should be avoided unless additional and unbiased assumptions can be made.

Grant support

Leducq foundation large equipment ERPT program,NEUROVASC7T project https://www.fondationleducq.org/ European Union’s Horizon 2020 research and innovation program (AROMA project) Grant no 885876 https://aroma-h2020.com/ NO - The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.