Meeting Banner
Abstract #2089

Physiological Noise Covariance Across Receiver Channels Explains Time-Series SNR Model for RF Coil Array FMRI Data

Jonathan R. Polimeni1, Christina Triantafyllou1, 2, Lawrence L. Wald1, 3

1Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Harvard Medical School, Massachusetts General Hospital, Charlestown, MA, United States; 2A. A. Martinos Imaging Center, McGovern Institute for Brain Research, Massachusetts Institute of Technology, Cambridge, MA, United States; 3Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States

The original model for relating image SNR to temporal SNR demonstrated that physiological noise scales linearly with signal level, providing a quantitative relationship between image SNR and temporal SNR. Recently it has been demonstrated that fMRI data acquired with multi-channel array coils deviates from the standard model and the deviation appears to increase with element count. Here we consider the effects of physiological noise correlations across coil channels. Extending the physiological noise model to include a physiological covariance matrix can explain the observed deviation of the data from the Kruger model and provides an interpretation of the recently proposed models.

Abstract PDF

Keywords

ability account acquisition activations additional alone applied applies array assessing assuming better biomedical body brain built calculating capture captures channel channels class coil coils combination common consider constant correlation correlations cortical cost count coupling covariance custom degree dependent describe described deviates deviation diagonal direct displayed distant division element elements entries expected experimental explain expressed extended extending fixed fluctuations freedom fully general gray head health hospital impact include increasing independent influence institute interpretation intuitively investigate isotropic linear linearly little locations many match matrices matrix measured medical mismatch model models much must neuronal noise note obeys observation opposite outer pair particular percentage perhaps physiological positive potential predicted presence product proportional protocols quantitative radiology receiver recently reconstructed reduce relationship relevant removed represents revisit scale scales scaling scanner school semidefinite series sides since statistically structure subjects subtracting supported technology temporal theory thermal tissue trend unclear variance vector vectors vendor violated whether whole whose yielded zero