Isolating physiologic noise sources with independently determined spatial measures

• Published in: NeuroImage (Orlando, Fla.) Vol. 37; no. 4; pp. 1286 - 1300
• Main Authors: Beall, Erik B., Lowe, Mark J.
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.10.2007; Elsevier Limited
• Subjects: Algorithms; Aliasing; Brain Mapping; Brain research; Data Interpretation, Statistical; Functional connectivity; Functional magnetic resonance imaging; Humans; Image Processing, Computer-Assisted - statistics & numerical data; Independent component analysis; Magnetic Resonance Imaging - statistics & numerical data; NMR; Nuclear magnetic resonance; Oxygen - blood; Physiologic monitoring; Physiologic noise; Plethysmography; Principal Component Analysis; Pulse; Reproducibility of Results; Respiratory Mechanics - physiology; Resting state; Signal processing; Studies; Functional connectivity; Physiologic noise; Resting state; Aliasing; Physiologic monitoring; Independent component analysis; Functional magnetic resonance imaging; Signal processing
• ISSN: 1053-8119; 1095-9572
• DOI: 10.1016/j.neuroimage.2007.07.004

To properly account for the presence of physiologic noise in fMRI data, parallel measurement of pulse and respiratory data is necessary. In some cases, this parallel measurement is difficult or impossible due to the experimental paradigm or lack of available monitoring equipment. We present a robust method for determining the direct-sampled pulse and respiratory data for a subject from the fMRI data itself, utilizing an independently determined spatial weighting matrix. It is shown that temporal independent component analysis can reliably separate the spatial and temporal patterns of physiologic noise through correlation if the parallel measurement is made. The spatial patterns thus determined can be applied to a separate scan of the same subject to produce the temporal pattern specific to this independent scan. The robustness of this method leads to the more general method of creating spatial weight matrices in standard brain space averaged over multiple subjects in order to acquire the physiologic signals without the necessity of any (further) parallel measurements. The resulting cardiac and respiratory estimators can effectively be used in a manner similar to that of a direct-sampled physiologic signal, e.g., direct input to retrospective correction methods, evaluation of cardiac and respiratory effects of tasks, etc. Spatial mixing matrices for estimating cardiac and respiratory sources for the acquisition protocols described here (and others as they are developed) are offered to investigators and can be obtained through e-mail from the corresponding author.