Experimental design and the relative sensitivity of BOLD and perfusion fMRI.

This paper compares the statistical power of BOLD and arterial spin labeling perfusion fMRI for a variety of experimental designs within and across subjects. Based on theory and simulations, we predict that perfusion data are composed of independent observations in time under the null hypothesis, in contrast to BOLD data, which possess marked autocorrelation. We also present a method (sinc subtraction) of generating perfusion data from its raw source signal that minimizes the presence of oxygen-sensitive signal changes and can be used with any experimental design. Empirically, we demonstrate the absence of autocorrelation in perfusion noise, examine the shape of the hemodynamic response function for BOLD and perfusion, and obtain a measure of signal to noise for each method. This information is then used to generate a model of relative sensitivity of the BOLD and perfusion methods for within-subject experimental designs of varying temporal frequency. It is determined that perfusion fMRI provides superior sensitivity for within-subject experimental designs that concentrate their power at or below approximately 0.009 Hz (corresponding to a "blocked" experimental design of 60-s epochs). Additionally, evidence is presented that across-subject hypothesis tests may be more sensitive when conducted using perfusion imaging, despite the better within-subject signal to noise obtained in some cases with BOLD.

A reference effect approach for power analysis in fMRI.

In fMRI, the issues involved in the control of type I error are fairly well understood. In contrast, the control of type II error has received less formal attention. This is perhaps due to the fact that the consideration of type II error requires the specification of an alternative hypothesis/experimental effect. In this paper, we present a method for expressing experimental effects in fMRI in a manner relative to a reference effect. A reference effect is chosen based on its neurophysiological significance to the researcher. This method provides a means to quantitatively express alternative hypotheses for fMRI, thus allowing type II error assessment prior to the collection of fMRI data. The simultaneous control of both type I and type II error should make meaningful interpretations possible from both positive and negative fMRI results.

Spatial localization and resolution of BOLD fMRI.

It has been demonstrated that the blood-oxygenation-level-dependent (BOLD) fMRI initial dip allows us to resolve (without differential subtraction) structures of the order of 0.5 mm. However, recent results support the proposition that even the later, positive BOLD fMRI signal component can allow us to resolve structures less than 1 mm in size by using differential subtraction when the signal-to-noise ratio is high. So, with a sufficient signal-to-noise ratio, the later, positive component should be useable as a probe for testing cognitive neuroscientific hypotheses that predict neuroanatomical dissociations of less than 1mm.

Cortical brain regions engaged by masked emotional faces in adolescents and adults: an fMRI study.

Face-emotion processing has shown signs of developmental change during adolescence. Functional magnetic resonance imaging (fMRI) was used on 10 adolescents and 10 adults to contrast brain regions engaged by a masked emotional-face task (viewing a fixation cross and a series of masked happy and masked fearful faces), while blood oxygen level dependent signal was monitored by a 1.5-T MRI scanner. Brain regions differentially engaged in the 2 age groups were mapped by using statistical parametric mapping. Summed across groups, the contrast of masked face versus fixation-cross viewing generated activations in occipital-temporal regions previously activated in passive face-viewing tasks. Adolescents showed higher maxima for activations in posterior association cortex for 3 of the 4 statistical contrasts. Adolescents and adults differed in the degree to which posterior hemisphere brain areas were engaged by viewing masked facial displays of emotion.

Coupling of neural activation to blood flow in the somatosensory cortex of rats is time-intensity separable, but not linear.

Changes in cerebral blood flow (CBF) because of functional activation are used as a surrogate for neural activity in many functional neuroimaging studies. In these studies, it is often assumed that the CBF response is a linear-time invariant (LTI) transform of the underlying neural activity. By using a previously developed animal model system of electrical forepaw stimulation in rats (n = 11), laser Doppler measurements of CBF, and somatosensory evoked potentials, measurements of neural activity were obtained when the stimulus duration and intensity were separately varied. These two sets of time series data were used to assess the LTI assumption. The CBF data were modeled as a transform of neural activity (N1-P2 amplitude of the somatosensory evoked potential) by using first-order (linear) and second-order (nonlinear) components. Although a pure LTI model explained a large amount of the variance in the data for changes in stimulus duration, our results demonstrated that the second-order kernel (i.e., a nonlinear component) contributed an explanatory component that is both statistically significant and appreciable in magnitude. For variations in stimulus intensity, a pure LTI model explained almost all of the variance in the CBF data. In particular, the shape of the CBF response did not depend on intensity of neural activity when duration was held constant (time-intensity separability). These results have important implications for the analysis and interpretation of neuroimaging data.

To smooth or not to smooth? Bias and efficiency in fMRI time-series analysis.

This paper concerns temporal filtering in fMRI time-series analysis. Whitening serially correlated data is the most efficient approach to parameter estimation. However, if there is a discrepancy between the assumed and the actual correlations, whitening can render the analysis exquisitely sensitive to bias when estimating the standard error of the ensuing parameter estimates. This bias, although not expressed in terms of the estimated responses, has profound effects on any statistic used for inference. The special constraints of fMRI analysis ensure that there will always be a misspecification of the assumed serial correlations. One resolution of this problem is to filter the data to minimize bias, while maintaining a reasonable degree of efficiency. In this paper we present expressions for efficiency (of parameter estimation) and bias (in estimating standard error) in terms of assumed and actual correlation structures in the context of the general linear model. We show that: (i) Whitening strategies can result in profound bias and are therefore probably precluded in parametric fMRI data analyses. (ii) Band-pass filtering, and implicitly smoothing, has an important role in protecting against inferential bias.

Testing for neural responses during temporal components of trials with BOLD fMRI.

Some cognitive neuroscientific hypotheses might concern neural responses occurring during particular periods of time in a behavioral trial. Here, these particular periods of time are referred to as temporal components of the trial. A difficulty in using BOLD fMRI to test hypotheses about neural responses during temporal components is that some information is irretrievably lost when neural responses are hemodynamically transformed. As a result, one cannot in general use the fMRI signal to unambiguously specify if there was a neural response during a given temporal component. However, adoption of a linear-time invariant model for the transform from neural signal to fMRI signal and constraint of the space of underlying neural waveforms might allow one to ask such questions. Here, the basic theory relevant to this issue and a corresponding method are discussed. The application of this method to fMRI time series data collected during the performance of a delayed-response trial is provided as an illustrative example.

The role of prefrontal cortex in sensory memory and motor preparation: an event-related fMRI study.

Delayed-response tasks are behavioral paradigms in which subjects must remember stimulus attributes across a delay to subsequently perform the appropriate motor response. Quintana and Fuster (1992), reported that there exist subpopulations of neurons in monkey lateral prefrontal cortex (PFC) whose firing rates during the delay are tuned to either sensorial attributes of the stimulus (i.e., involved in sensory memory) or the direction of a postdelay motor response associated with the stimulus (i.e., involved in motor preparation). We studied human subjects with an event-related fMRI method that would allow us to test the hypothesis that there are regions within the PFC that are recruited during both motor preparation and sensory memory. Subjects performed a delayed-response task with two types of trials that either (1) allowed subjects to prepare during a delay period for a specific motor response or (2) required that subjects maintain a sensory attribute (specifically, color) during a delay period for correct performance postdelay. It was assumed that during the delay periods, the delayed-response trials would engage motor preparation while delayed-match trials would engage sensory memory. Behavioral data supported this assumption. Imaging results support the hypothesis that the PFC is involved in both motor preparation and sensory memory. Furthermore, no selectivity (in terms of intensity of neural representation on the spatial scale of the voxel size <5 mm(3)) for motor preparation over sensory memory (or vice-versa) was detected within the PFC. This latter result fails to support a gross anatomical segregation within the PFC with respect to involvement in these two cognitive processes.

Using event-related fMRI to assess delay-period activity during performance of spatial and nonspatial working memory tasks.

Event-related experimental design and analysis techniques for functional magnetic resonance imaging (fMRI) take advantage of the intrinsic temporal resolution of fMRI to permit investigation of complex human behaviors on the time scale over which they can occur. The protocol described in this report permits the effective isolation and assessment of variance in the fMRI signal that is attributable solely to the delay portion of delayed-response tasks. It permits, therefore, evaluation of the purely mnemonic portions of working memory tasks without requiring the "cognitive subtraction" of nonmnemonic components of such tasks, such as visual processing and motor output. Features of this event-related fMRI technique include the empirical derivation of an impulse response function (IRF) from each subject participating in the experiment, single-subject and random effects group analyses, use of t-values of dependent measures, and the use of regions of interest (ROI) to improve the sensitivity of a priori contrasts. This report provides a detailed exposition of the research methodology of our event-related fMRI technique, the rationale behind many of its critical features, and examples of its application to two empirical datasets.

Replication and further studies of neural mechanisms of spatial mnemonic processing in humans.

Changes in neuronal firing rates during periods of time when subjects are required to remember information (retention delays) have been reported in non-human primates. In humans, tests for such functional changes using hemodynamic markers of neural activity have typically relied on cognitive subtraction. However, the temporal resolution of fMRI allows a more direct test than that afforded by cognitive subtraction of the idea that certain brain regions may increase their neural activity during retention delays in humans. Using a method that exploits this temporal resolution, increased functional activity attributable to a retention delay for spatial information in regions proximate to/within the right frontal eye field and the right superior parietal lobule were detected (in four out of four and three out of four subjects, respectively; this is an internal replication of the results of [E. Zarahn, G.K. Aguirre, M. D'Esposito, Temporal isolation of the neural correlates of spatial mnemonic processing with fMRI, Cognit. Brain Res., 7 (1999) 255-268. ]). Second, a model in which ventral and not dorsal prefrontal cortex in humans is involved in simply maintaining spatial information was tested. The results disputed this model as increases in fMRI signal attributable to the retention delay were detected more frequently in dorsal than ventral prefrontal cortex. Third, a model which posited that the intensity of neural activity is causally related to the accuracy of spatial mnemonic representation was tested by comparing retention delay signal between correct and incorrect trials. The results did not support this model in any of the regions tested.

Stochastic designs in event-related fMRI.

This article considers the efficiency of event-related fMRI designs in terms of the optimum temporal pattern of stimulus or trial presentations. The distinction between "stochastic" and "deterministic" is used to distinguish between designs that are specified in terms of the probability that an event will occur at a series of time points (stochastic) and those in which events always occur at prespecified time (deterministic). Stochastic designs may be "stationary," in which the probability is constant, or nonstationary, in which the probabilities change with time. All these designs can be parameterized in terms of a vector of occurrence probabilities and a prototypic design matrix that embodies constraints (such as the minimum stimulus onset asynchrony) and the model of hemodynamic responses. A simple function of these parameters is presented and used to compare the relative efficiency of different designs. Designs with slow modulation of occurrence probabilities are generally more efficient than stationary designs. Interestingly the most efficient design is a conventional block design. A critical point, made in this article, is that the most efficient design for one effect may not be the most efficient for another. This is particularly important when considering evoked responses and the differences among responses. The most efficient designs for evoked responses, as opposed to differential responses, require trial-free periods during which baseline levels can be attained. In the context of stochastic, rapid-presentation designs this is equivalent to the inclusion of "null events."

The effect of normal aging on the coupling of neural activity to the bold hemodynamic response.

The use of functional neuroimaging to test hypotheses regarding age-related changes in the neural substrates of cognitive processes relies on assumptions regarding the coupling of neural activity to neuroimaging signal. Differences in neuroimaging signal response between young and elderly subjects can be mapped directly to differences in neural response only if such coupling does not change with age. Here we examined spatial and temporal characteristics of the BOLD fMRI hemodynamic response in primary sensorimotor cortex in young and elderly subjects during the performance of a simple reaction time task. We found that 75% of elderly subjects (n = 20) exhibited a detectable voxel-wise relationship with the behavioral paradigm in this region as compared to 100% young subjects (n = 32). The median number of suprathreshold voxels in the young subjects was greater than four times that of the elderly subjects. Young subjects had a slightly greater signal:noise per voxel than the elderly subjects that was attributed to a greater level of noise per voxel in the elderly subjects. The evidence did not support the idea that the greater head motion observed in the elderly was the cause of this greater voxel-wise noise. There were no significant differences between groups in either the shape of the hemodynamic response or in its the within-group variability, although the former evidenced a near significant trend. The overall finding that some aspects of the hemodynamic coupling between neural activity and BOLD fMRI signal change with age cautions against simple interpretations of the results of imaging studies that compare young and elderly subjects.

Event-related functional MRI: implications for cognitive psychology.

Functional magnetic resonance imaging (fMRI) has rapidly emerged as a powerful technique in cognitive neuroscience. We describe and critique a new class of imaging experimental designs called event-related fMRI that exploit the temporal resolution of fMRI by modeling fMRI signal changes associated with behavioral trials as opposed to blocks of behavioral trials. Advantages of this method over block designs include the ability to (a) randomize trial presentations, (b) test for functional correlates of behavioral measures with greater power, (c) directly examine the neural correlates of temporally dissociable components of behavioral trials (e.g., the delay period of a working memory task), and (d) test for differences in the onset time of neural activity evoked by different trial types. Consequently, event-related fMRI has the potential to address a number of cognitive psychology questions with a degree of inferential and statistical power not previously available.

Temporal isolation of the neural correlates of spatial mnemonic processing with fMRI.

The use of cognitive subtraction to study the neural substrates of the maintenance component of spatial working memory in humans relies upon the assumptions of the pure insertion of cognitive processes and a linear transform of neural activity to neuroimaging signal. Here, functional changes attributable to the memory requiring phase (referred to as the retention delay) of a spatial working memory task were temporally discriminated from those attributable to other behavioral subcomponents within trials using an experimental design that is argued to obviate these assumptions, as well as permit a joint test of their validity. The hypothesis that the assumptions of cognitive subtraction (as applied to neuroimaging) hold in general was not supported. Functional changes attributable to the retention delay were detected in the dorsolateral prefrontal cortex as well as in other cortical regions in a subset of the subjects, and in the right frontal eye field and right superior parietal lobule of all subjects (n=5). These results support models in which these regions are involved in maintaining spatial representations in humans. In addition, nearly all regions that evidenced such functional changes during the retention delay also evidenced functional changes during behaviors that did not require spatial working memory. This result tends to dispute models which posit the existence of gross neuroanatomical regions involved in solely mnemonic function.

The variability of human, BOLD hemodynamic responses.

Cerebral hemodynamic responses to brief periods of neural activity are delayed and dispersed in time. The specific shape of these responses is of some importance to the design and analysis of blood oxygenation level-dependent (BOLD), functional magnetic resonance imaging (fMRI) experiments. Using fMRI scanning, we examine here the characteristics and variability of hemodynamic responses from the central sulcus in human subjects during an event-related, simple reaction time task. Specifically, we determine the contribution of subject, day, and scanning session (within a day) to variability in the shape of evoked hemodynamic response. We find that while there is significant and substantial variability in the shape of responses collected across subjects, responses collected during multiple scans within a single subject are less variable. The results are discussed in terms of the impact of response variability upon sensitivity and specificity of analyses of event-related fMRI designs.

Human prefrontal cortex is not specific for working memory: a functional MRI study.

Lesion studies in monkeys have provided evidence that lateral prefrontal cortex is necessary for working memory, the cognitive processes involved in the temporary maintenance and manipulation of information. Monkey electrophysiological studies, however, have also observed prefrontal neuronal activity associated with cognitive processes that are nonmnemonic. We tested the hypothesis that the same regions of human prefrontal cortex that demonstrate activity during working memory tasks would also demonstrate activity during tasks without working memory demands. During echoplanar fMRI imaging, subjects performed a three-condition experiment (working memory task, nonworking memory task, rest). In the working memory task, subjects observed serially presented stimuli and determined if each stimulus was the same as that presented two stimuli back. The nonworking memory task in Experiment 1 required subjects to identify a single predetermined stimulus; in Experiment 2, subjects were required to make a button press to every stimulus. In all subjects in both experiments, the working memory task exhibited greater prefrontal cortical activity compared to either nonworking memory task. In these same prefrontal regions, greater activation was also observed during both nonworking memory tasks compared to rest. We conclude that human lateral prefrontal cortex supports processes in addition to working memory. Thus, reverse inference of the form "if prefrontal cortex is active, working memory is engaged" is not supported.

An area within human ventral cortex sensitive to "building" stimuli: evidence and implications.

Isolated, ventral brain lesions in humans occasionally produce specific impairments in the ability to use landmarks, particularly buildings, for way-finding. Using functional MRI, we tested the hypothesis that there exists a cortical region specialized for the perception of buildings. Across subjects, a region straddling the right lingual sulcus was identified that possessed the functional correlates predicted for a specialized building area. A series of experiments discounted several alternative explanations for the behavior of this site. These results are discussed in terms of their impact upon our understanding of the functional structure of visual processing, disorders of topographical disorientation, and the influence of environmental conditions upon neural organization.

Functional MRI studies of spatial and nonspatial working memory.

Single-unit recordings in monkeys have revealed neurons in the lateral prefrontal cortex that increase their firing during a delay between the presentation of information and its later use in behavior. Based on monkey lesion and neurophysiology studies, it has been proposed that a dorsal region of lateral prefrontal cortex is necessary for temporary storage of spatial information whereas a more ventral region is necessary for the maintenance of nonspatial information. Functional neuroimaging studies, however, have not clearly demonstrated such a division in humans. We present here an analysis of all reported human functional neuroimaging studies plotted onto a standardized brain. This analysis did not find evidence for a dorsal/ventral subdivision of prefrontal cortex depending on the type of material held in working memory, but a hemispheric organization was suggested (i.e., left-nonspatial; right-spatial). We also performed functional MRI studies in 16 normal subjects during two tasks designed to probe either nonspatial or spatial working memory, respectively. A group and subgroup analysis revealed similarly located activation in right middle frontal gyrus (Brodmann's area 46) in both spatial and nonspatial [working memory-control] subtractions. Based on another model of prefrontal organization [M. Petrides, Frontal lobes and behavior, Cur. Opin. Neurobiol., 4 (1994) 207-211], a reconsideration of the previous imaging literature data suggested that a dorsal/ventral subdivision of prefrontal cortex may depend upon the type of processing performed upon the information held in working memory.

Functional MRI lateralization of memory in temporal lobe epilepsy.

OBJECTIVE: To determine the feasibility of using functional magnetic resonance imaging (fMRI) to detect asymmetries in the lateralization of memory activation in patients with temporal lobe epilepsy (TLE). BACKGROUND: Assessment of mesial temporal lobe function is a critical aspect of the preoperative evaluation for epilepsy surgery, both for predicting postoperative memory deficits and for seizure lateralization. fMRI offers several potential advantages over the current gold standard, intracarotid amobarbital testing (IAT). fMRI has already been successfully applied to language lateralization in TLE. METHODS: fMRI was carried out in eight normal subjects and 10 consecutively recruited patients with TLE undergoing preoperative evaluation for epilepsy surgery. A complex visual scene encoding task known to activate mesial temporal structures was used during fMRI. Asymmetry ratios for mesial temporal activation were calculated, using regions of interest defined in normals. Patient findings were compared with the results of IAT performed as part of routine clinical evaluation. RESULTS: Task activation was nearly symmetric in normal subjects, whereas in patients with TLE, significant asymmetries were observed. In all nine patients in whom the IAT result was interpretable, memory asymmetry by fMRI concurred with the findings of IAT including two patients with paradoxical IAT memory lateralization ipsilateral to seizure focus. CONCLUSIONS: fMRI can be used to detect asymmetries in memory activation in patients with TLE. Because fMRI studies are noninvasive and provide excellent spatial resolution for functional activation, these preliminary results suggest a promising role for fMRI in improving the preoperative evaluation for epilepsy surgery.