Nonlinear coupling of neural activity and CBF in rodent barrel cortex Myles Jones, * Nicola Hewson-Stoate, John Martindale, Peter Redgrave, and John Mayhew Neural Imaging Research Unit, Department of Psychology, University of Sheffield, Western Bank, Sheffield S10 2TP, UK Received 20 November 2003; revised 26 January 2004; accepted 2 February 2004 Available online 30 April 2004 The relationship between neural activity and accompanying changes in cerebral blood flow (CBF) and oxygenation must be fully understood before data from brain imaging techniques can be correctly interpreted. Whether signals in fMRI reflect the neural input or output of an activated region is still unclear. Similarly, quantitative relationships between neural activity and changes in CBF are not well understood. The present study addresses these issues by using simultaneous laser Doppler flowmetry (LDF) to measure CBF and multichannel electro- physiology to record neural activity in the form of field potentials and multiunit spiking. We demonstrate that CBF-activation coupling is a nonlinear inverse sigmoid function. Comparing the data with previous work suggests that within a cortical model, CBF shows greatest spatial correlation with a current sink 500 Am below the surface corresponding to sensory input. These results show that care must be exercised when interpreting imaging data elicited by particularly strong or weak stimuli and that hemodynamic changes may better reflect the input to a region rather than its spiking output. D 2004 Elsevier Inc. All rights reserved. Keywords: Cerebral blood flow; Laser Doppler flowmetry; fMRI; Hemodynamic changes Introduction Blood oxygen level-dependent (BOLD) fMRI (Kwong et al., 1992; Ogawa et al., 1992) is based on local changes in blood flow, blood volume, and oxygenation, which accompany neural activity. The increase in cerebral blood flow (CBF) into activated regions is the major factor in producing the positive BOLD response com- monly used in human brain mapping studies (Buxton and Frank, 1997). A good understanding of the relationship between neural activity and CBF, therefore, is fundamentally important for the interpretation of functional imaging data (Logothetis, 2002). The nature of the relationship between neural activity and its accompa- nying hemodynamics is still largely unresolved. First, it is uncertain whether hemodynamic changes are associated more closely with input to an activated region, as reflected by the local field potential (LFP), or with the spiking output. Some studies have found that LFP is a better predictor of the CBF and BOLD response (Caesar et al., 2003; Logothetis et al., 2001), while others have found strong correlations with spiking activity (Heeger et al., 2000; Smith et al., 2002). If imaging signals result from excitatory postsynaptic potentials (EPSP) rather than intrinsic spiking, then different functions may be attributed to a brain region on the basis of imaging (Tolias et al., 2001) and electrophysiological experiments (Desi- mone and Schein, 1987). Secondly, it is unclear what the precise relationship is between the magnitudes of the hemodynamic response and the changes in neural activity. An a priori assumption in BOLD imaging studies is that the change in signal is linearly related to the amount of neural activity. However, if the relationship is nonlinear, then differences in imaging signals detected during different task conditions (Chawla et al., 1999) or between subjects with genetic variations (Hariri et al., 2002) could reflect disproportionate differences in the underlying amount of neural activity. Although linear coupling between neural activity and CBF (Ngai et al., 1999) and BOLD has been reported by some authors (Arthurs et al., 2000; Brinker et al., 1999; Heeger et al., 2000; Sheth et al., 2003; Smith et al., 2002 ), a recent study by Devor et al. (2003) found that for brief stimuli, the relationship between neural activity and hemoglobin concentration and satura- tion was highly nonlinear and was better fitted by a power law. Their use of subtle stimuli and a short interstimulus interval (ISI) allowed neurovascular coupling to be investigated on a spatially discrete scale with good statistical power. However, the magnitude of the resultant hemodynamic changes was small (approximately 0.1%) and as such may not be detectable in the BOLD signal. Thus, the value of their findings to biophysical models of BOLD fMRI may be limited. In the present study, we have employed a similar animal model but used comparable stimulus conditions to fMRI studies (Mandeville et al., 1999; Marota et al., 1999; Ogawa et al., 2000; Silva and Koretsky, 2002; Silva et al., 2000), thus allowing neuro- vascular coupling to be investigated across a range of neural activities and CBF changes relevant to fMRI. A third potential problem for previous studies investigating neurovascular coupling is the intrinsic ambiguity inherent in LFP measurements of neural activity (e.g., see Nielsen and Lauritzen, 2001). LFPs are generated by current sinks (inward flow of ions into the cell responsible for generating EPSPs) and current sources (passive outflow of current from nonactivated regions of the cell). As a consequence of superposition, the LFPs produced in an activated region by multiple sinks and sources add algebraically. The interpretation of data from a single electrode, therefore, is inherently ambiguous. Current source density (CSD) analysis of LFPs from multiple cortical depths allows deconvolution of the data 1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2004.02.007 * Corresponding author. Fax: +44-114-276-6515. E-mail address: m.jones@sheffield.ac.uk (M. Jones). Available online on ScienceDirect (www.sciencedirect.com.) www.elsevier.com/locate/ynimg NeuroImage 22 (2004) 956 – 965