Parsing a sequence of brain activations at psychological times using fMRI
Introduction
Mental chronometry seeks to measure the time course of mental operations in the nervous system. (Posner, 1978, Sternberg, 1969). A recent challenge has been to integrate psychological chronometric methods and imaging techniques, to observe the sequence of brain activations of a complex cognitive task. A series of previous studies have shown that fine temporal fMRI measurements (∼ 100 ms) while not necessarily easy to achieve, are feasible (Buckner et al., 1996, Dale et al., 2000, Formisano et al., 2002, Henson et al., 2002, Kim et al., 1997, Lange and Zeger, 1997, Lee et al., 2005, Liao et al., 2002, Menon et al., 1998, Purdon et al., 2001, Richter et al., 1997a, Richter et al., 1997b, Richter et al., 2000, Saad et al., 2003, Schacter et al., 1997, Thierry et al., 1999, Weilke et al., 2001, Wildgruber et al., 1999 for reviews, see Buckner, 2003, Formisano and Goebel, 2003, Menon and Kim, 1999, Rosen et al., 1998). Yet at the moment, they have not been applied to the full decomposition of a cognitive task at the whole-brain level. Part of the difficulty to achieve this is that three fundamental methodological issues have to be simultaneously resolved: (1) estimate timing information at a slow sampling rate, thus allowing to provide a whole-brain measure; (2) estimate timing parameters invariantly across different brain regions, independently of possible variations in the shape of the hemodynamic response function (HRF); (3) distinguish changes in onset latency and in duration (Bellgowan et al., 2003).
In addition to the solution of these methodological issues, a full parsing of a task requires a model of the expected sequence, to decompose the observed delays and durations in terms of contributions of the successive stages. In a serial scheme, the onset of the stage {i} can be calculated as the sum of the onset and durations of stage {i − 1}. This can be solved recursively to obtain the total organization of the task into subcomponents. However, questions such as the parallel or serial nature of the stages of a task are not fully understood and in a general cognitive task, the organization of processing stages may be highly complex, variable and unknown (Sigman and Dehaene, 2005).
Here, we introduce a novel strategy which solves the previously stated methodological difficulties and use it in a sequential task, where onset and duration of the processing stages are under experimental control. The strategy is based on previous Fourier-based methods (Lee et al., 1995, Menon et al., 1998, Rajapakse et al., 1998), and crucially, on the experimental manipulation of the task to differentially affect the timing of its successive stages. By studying the covariation of the timing parameters with the experimental manipulation, we achieve a measure which is insensitive to the particular shape of the HRF and thus invariant across the brain. With an algebraic estimation of these obtained parameters, we are able to recover the precise timing of all the stages which compose the task, which involves a large variety of cognitive processes. The duration of the whole task, composed of five stages, is less than one TR (the sampling time).
Section snippets
Simulations and calculation of predicted values of the phase
Simulations were performed by convolving square pulses of duration D and delay Δ with the HRF described in SPM99, using the default parameters: delay of response (relative to onset) = 6 s, delay of undershoot (relative to onset) = 16 s, dispersion of response = 1 s, dispersion of undershoot = 1 s, ratio of response to undershoot = 6 s. We show here the analytic calculation of the value of the phase and the amplitude as a function of the critical parameters. The BOLD time-series is given by the
An analytic method to obtain timing information from the fMRI signal and to identify sequential stages of processing
In this section, we describe the basic aspects of our analytic method, as well as simulations designed to estimate the optimal parameters. We extract the spectral parameters, amplitude and phase, from the BOLD signal resulting from a slow periodic stimulation set by the inter trial interval (ITI). While the absolute value of the phase and spectrum depends on various parameters such as the HRF, or the repetition frequency, their relative value (i.e. the comparison of these values in two
Discussion
We have shown that we can parse a sequence of brain activations, using whole brain fMRI at a resolution of a few hundredth milliseconds, about 10 times better than the sampling time. While here we used the phase information in a periodic signal to obtain timing information from the fMRI (Menon et al., 1998), this does not seem to be essential and other proposed methods, which achieve comparable resolution could have been used (Bellgowan et al., 2003, Formisano and Goebel, 2003, Henson et al.,
Acknowledgments
We thank Ghislaine Dehaene-Lambertz for substantial experimental help. MS is supported by a Human Frontiers Science Program Fellowship and SD by a centennial fellowship of the McDonnell Foundation.
References (58)
- et al.
The variability of human, BOLD hemodynamic responses
NeuroImage
(1998) - et al.
Specialization within the ventral stream: the case for the visual word form area
NeuroImage
(2004) - et al.
Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity
Neuron
(2000) - et al.
Tracking cognitive processes with functional MRI mental chronometry
Curr. Opin. Neurobiol.
(2003) - et al.
Tracking the mind's image in the brain: I. Time-resolved fMRI during visuospatial mental imagery
Neuron
(2002) - et al.
Disturbance of rhythm sense following right hemisphere damage
Neuropsychologia
(1990) - et al.
Detecting latency differences in event-related BOLD responses: application to words versus nonwords and initial versus repeated face presentations
NeuroImage
(2002) - et al.
Limitations of temporal resolution in functional MRI
Magn. Reson. Med.
(1997) - et al.
Medial prefrontal and subcortical mechanisms underlying the acquisition of motor and cognitive action sequences in humans
Neuron
(2002) - et al.
Nonlinear regression of functional MRI data: an item recognition task study
NeuroImage
(2000)