Mapping Cognitive Function




Cognitive functions are fundamental to being human. Although tremendous progress has been made in the science of cognition using neuroimaging, the clinical applications of neuroimaging are just beginning to be realized. This article focuses on selected technologies, analysis techniques, and applications that have, or will soon have, direct clinical impact. The authors discuss how cognition can be imaged using MR imaging, functional MR imaging, positron emission tomography, magnetoencephalography and electroencephalography, and MR imaging diffusion tensor imaging. A unifying theme of this article is the concept that a more complete understanding of cognition only comes through integration of multimodal structural and functional imaging technologies.


Cognitive functions are fundamental to being human. Although tremendous progress has been made in the science of cognition using neuroimaging, the clinical applications of neuroimaging are just beginning to be realized. Rather than give an exhaustive review of the neuroimaging of cognition, this article focuses on selected technologies, analysis techniques, and applications that have, or will soon have, direct clinical impact. The authors discuss how cognition can be imaged using MR imaging, functional MR (fMR) imaging, positron emission tomography (PET), magnetoencephalography and electroencephalography (MEG/EEG), and MR imaging diffusion tensor (DT) imaging. Many of these technologies have been topics in other articles, so the authors briefly introduce a few new ones in terms of image integration across the more familiar modalities.


Technologies for functional imaging


Cognition is a large topic and the authors focus on cognition of primary sensory areas, language, and memory. A unifying theme of this article is the concept that a more complete understanding of cognition only comes through integration of imaging technologies. The imaging technologies are structural (MR imaging, DT imaging, CT) and functional (PET, fMR imaging, MEG/EEG, optical) techniques, and both have important roles in imaging cognition. Further, functional techniques can be thought of as originating from two sources: neuroelectric (EEG and MEG) and hemodynamic (PET, single photon computed tomography (SPECT), fMR imaging, optical). Exploring the structure-function relationships found with neuroimaging is an area of intense research. Understanding the structure-function relationship requires the integration across imaging modalities, and even by integrating behavioral metrics .


Each imaging technology has its own individual strengths and weaknesses, typically with some trade-off between temporal and spatial resolution. The electromagnetic imaging techniques include MEG and EEG, which may not be thought of as imaging at all, yet do have excellent temporal resolution because they measure the underlying neural currents directly, on the order of milliseconds, although with limited spatial localization. fMR imaging has excellent spatial resolution, on the order of millimeters, but it depends on the hemodynamic response, which limits the temporal resolution. By combining imaging technologies, it is possible to optimize the temporal and spatial scale, to localize accurately and characterize cognition noninvasively.




Using sensory stimulation to probe for disorders of higher cognitive function


Although cognitive studies often use tasks that involve higher cognitive function, using sensory stimuli that have a cognitive correlate, such as the mismatch negativity and steady-state responses, has a long history .


Mismatch negativity


The auditory mismatch response is a component of the MEG/EEG response elicited by any change (“deviant”) in auditory stimulation (“standard”), usually peaking at 80 to 200 milliseconds from change onset . The mismatch response is elicited, for example, when a sound changes in frequency, duration, or intensity. The mismatch response is elicited in the absence of attention (ie, when no task performance is required), although clear modulation by attention of the mismatch responses has been described . MEG/EEG source analysis , and depth recordings in humans and animals , suggest that the mismatch response is generated in the posterior superior temporal gyrus (STG), in or near Heschl’s gyrus. Prior work has suggested that the mismatch response is a composite of several different neural generators, including parietal and frontal generators, in addition to STG sources. Others suggest that the STG source of the mismatch response results from the anterior and posterior components of the N1 response, known as N1a and N1p, respectively . The N1p may be generated in the associative auditory cortex and the planum temporale (earlier latency than the anterior component) and the N1a may have a genesis just anterior and lateral to the primary auditory cortex ( Fig. 1 ). The mismatch response may originate in or near the N1a generator or slightly anterior to it .




Fig. 1


Auditory mismatch response and combining MEG with fMR imaging (fMRI) data. Upper panel shows a three-dimensional (3-D) rendering of a brain computed from the subject’s MR image (MRI), with the corresponding inflated cortex patch around the auditory areas. Labeled areas include Heschl’s gyrus (HG), planum temporale (PT), STG, superior temporal sulcus (STS), and middle temporal gyrus (MTG). The middle panel depicts the equivalent current dipole (ECD) method, which is traditionally used for MEG source estimation. On the left is a sagittal MR image with the ECD source estimate location, with the black bar indicating the dipole direction. On the right, the dipole representing the peak in the magnetic mismatch response is registered onto an inflated cortex. Notice that, in this case, the dipole is near the anterior portion of HG. Integration of the fMRI, the MEG data, and the inflated cortex data are combined ( lowest panel ) to create a “movie” or spatiotemporal map. The fMR imaging data are superimposed on source echoplanar imaging blood oxygenation level–dependent MR images, which is then transformed onto the inflated cortex, illustrating how the activity from MEG is transformed, using the minimum norm estimate (MNE), onto an inflated cortical source. Next, the fMR imaging data and the MNE results are combined into a spatiotemporal movie (dSPM). Some activity of the mismatch response is seen at 93 milliseconds, which peaks at approximately 103 milliseconds, and is diminished by 113 milliseconds.


Most , but not all , studies have reported a reduced mismatch response in neural diseases that have a reduction in inhibition, such as schizophrenia. Electrophysiologic investigations suggest that the reduced mismatch response is localized in the posterior STG , the site of reduced volume measurements in the STG as measured by MR imaging. The mismatch response reliability is excellent , which makes it a good tool for clinical use.


For use in the imaging of normal and impaired cognition, the mismatch response may be used to determine the integrity of particular neurotransmitter systems. For example, some studies indicate that the excitatory glutamate receptors interacting with the inhibitor gamma-aminobutyric acid (GABA) system are closely related to the abnormalities of the mismatch response in many disorders of cognition, and typified by schizophrenia . The monkey equivalent of the human mismatch response was abolished by microinjections of competitive and noncompetitive glutamate N-methyl d-aspartate (NMDA) receptor antagonists, without noticeable effects on the temporally preceding activity in lower laminae or the deviant alone stimuli . A similar inherent modulation was observed on surface potentials in schizophrenia . Further, Javitt has shown that exogenously administered NMDA receptor antagonists also result in a reduction of the mismatch response similar to that observed in schizophrenia. These findings are supported by animal models that demonstrate recurrent inhibition of NMDA-receptor neurons by way of GABAergic interneurons . These results suggest that noninvasive imaging of the mismatch negativity may have a direct impact on the clinical care of patients, such as in schizophrenia, especially when combined with other measurements and imaging techniques.


Spectral analysis of cognitive processes using visual stimuli


Bertrand and colleagues have investigated visual cognition using various paradigms and showed that during a visual object recognition, a burst of neural activity near 40 Hz (known as the gamma-band frequency) correlates strikingly with many cognitive acts. The gamma-band activity typically occurs within two time windows after presentation of a visual stimulus: one phase-locked (evoked) to the stimulus at about 180 milliseconds and another, which is not phase-locked to the stimulus (induced), which peaks at approximately 400 milliseconds after stimulus presentation .


The timing of neural events is crucial for normal brain function, and is probably most important for the coordination of cognitive acts, especially for object recognition. Disruption of the precise timing of cortical events may explain the diverse cognitive deficits in dyslexia, depression, and schizophrenia . Neural activity in the gamma frequency band (30 – 80 Hz or broader, centered on 40 Hz) has been hypothesized to reflect the synchronization of neural assemblies involved in the “binding” of various features of an object within a single sensory modality and across modalities, and also across time . At the cellular level, Traub and colleagues have presented evidence that interneurons with GABA A -mediated synaptic transmission interacting with glutamate-mediated cells provide the substrate for gamma oscillations in the cortex. A rich literature suggests that gamma-band activity may bind information across distant sites in the brain, known as the temporal binding hypothesis . Neural modeling, using sophisticated, biologically based simulations based on cellular physiology, suggest a complex interplay in the excitatory and inhibitory cells in generating intrinsic rhythms in the brain, and that different frequency bands have complementary roles in brain function , which is supported by recent cellular research .


Auditory steady-state response


A robust early neural response is the auditory steady-state response (ASSR). It has a maximal peak when the stimulation frequency, typically presented as brief auditory clicks, is presented at 40 Hz. Several studies using the ASSR in schizophrenia show a deficit in the 40-Hz response ( Fig. 2 ) in the ability to follow the stimulus, in frequency and phase . In a separate study, the ASSR was increased in Alzheimer’s disease patients, compared with normal controls, which was suggested to represent decreased inhibition and adaptation in auditory processing .




Fig. 2


ASSR in schizophrenia. ( A ) Magnetic field strength (fT/cm) versus time (stimulus presented at time 0, lasting ∼ 500 milliseconds) in a single MEG channel over the temporal lobe. Note the appearance of middle latency evoked response, N100 m (N1M), and the overriding sustained field. ( B ) Power spectral density in a temporal lobe gradiometer (maximal response for auditory stimuli) over the time interval 0 to 500 milliseconds. Note that in the control subjects, 40 Hz is the largest response (greater than both the 20-Hz and 30-Hz power peaks, often twice the magnitude). In this case, however, the 40-Hz power is less than one half the power in either the 20-Hz or the 30-Hz power, although the 20-Hz response is maximal at a frequency somewhat less than 20 Hz. ( C ) Time frequency representation (TFR) showing the spectral power as a function of time. Sustained neuromagnetic field from MEG gamma-band response to periodic auditory clicks at 40 Hz in a control subject. The color scale shows increased spectral power of the Morlet wavelet (blue to red for increasing power). ( D ) Source estimation localization using the same response to same auditory clicks from A , with the fMR imaging spatial information (not shown) using a Morlet wavelet–based minimum norm estimate centered at 40 Hz at time 230 milliseconds after beginning of click train. It is displayed on an inflated cortical surface of the same subject. The activity appears to involve middle and medial Heschl’s gyrus, extending into the planum temporale. Note that the MEG power spectrum calculated with a Fast Fourier Transform in B does not include any temporal information, unlike the wavelet-transformed measures such as the TFR in C and the spectral spatiotemporal map in D. ( Courtesy of Thomas Witzel, Charlestown, MA.)




Using sensory stimulation to probe for disorders of higher cognitive function


Although cognitive studies often use tasks that involve higher cognitive function, using sensory stimuli that have a cognitive correlate, such as the mismatch negativity and steady-state responses, has a long history .


Mismatch negativity


The auditory mismatch response is a component of the MEG/EEG response elicited by any change (“deviant”) in auditory stimulation (“standard”), usually peaking at 80 to 200 milliseconds from change onset . The mismatch response is elicited, for example, when a sound changes in frequency, duration, or intensity. The mismatch response is elicited in the absence of attention (ie, when no task performance is required), although clear modulation by attention of the mismatch responses has been described . MEG/EEG source analysis , and depth recordings in humans and animals , suggest that the mismatch response is generated in the posterior superior temporal gyrus (STG), in or near Heschl’s gyrus. Prior work has suggested that the mismatch response is a composite of several different neural generators, including parietal and frontal generators, in addition to STG sources. Others suggest that the STG source of the mismatch response results from the anterior and posterior components of the N1 response, known as N1a and N1p, respectively . The N1p may be generated in the associative auditory cortex and the planum temporale (earlier latency than the anterior component) and the N1a may have a genesis just anterior and lateral to the primary auditory cortex ( Fig. 1 ). The mismatch response may originate in or near the N1a generator or slightly anterior to it .




Fig. 1


Auditory mismatch response and combining MEG with fMR imaging (fMRI) data. Upper panel shows a three-dimensional (3-D) rendering of a brain computed from the subject’s MR image (MRI), with the corresponding inflated cortex patch around the auditory areas. Labeled areas include Heschl’s gyrus (HG), planum temporale (PT), STG, superior temporal sulcus (STS), and middle temporal gyrus (MTG). The middle panel depicts the equivalent current dipole (ECD) method, which is traditionally used for MEG source estimation. On the left is a sagittal MR image with the ECD source estimate location, with the black bar indicating the dipole direction. On the right, the dipole representing the peak in the magnetic mismatch response is registered onto an inflated cortex. Notice that, in this case, the dipole is near the anterior portion of HG. Integration of the fMRI, the MEG data, and the inflated cortex data are combined ( lowest panel ) to create a “movie” or spatiotemporal map. The fMR imaging data are superimposed on source echoplanar imaging blood oxygenation level–dependent MR images, which is then transformed onto the inflated cortex, illustrating how the activity from MEG is transformed, using the minimum norm estimate (MNE), onto an inflated cortical source. Next, the fMR imaging data and the MNE results are combined into a spatiotemporal movie (dSPM). Some activity of the mismatch response is seen at 93 milliseconds, which peaks at approximately 103 milliseconds, and is diminished by 113 milliseconds.


Most , but not all , studies have reported a reduced mismatch response in neural diseases that have a reduction in inhibition, such as schizophrenia. Electrophysiologic investigations suggest that the reduced mismatch response is localized in the posterior STG , the site of reduced volume measurements in the STG as measured by MR imaging. The mismatch response reliability is excellent , which makes it a good tool for clinical use.


For use in the imaging of normal and impaired cognition, the mismatch response may be used to determine the integrity of particular neurotransmitter systems. For example, some studies indicate that the excitatory glutamate receptors interacting with the inhibitor gamma-aminobutyric acid (GABA) system are closely related to the abnormalities of the mismatch response in many disorders of cognition, and typified by schizophrenia . The monkey equivalent of the human mismatch response was abolished by microinjections of competitive and noncompetitive glutamate N-methyl d-aspartate (NMDA) receptor antagonists, without noticeable effects on the temporally preceding activity in lower laminae or the deviant alone stimuli . A similar inherent modulation was observed on surface potentials in schizophrenia . Further, Javitt has shown that exogenously administered NMDA receptor antagonists also result in a reduction of the mismatch response similar to that observed in schizophrenia. These findings are supported by animal models that demonstrate recurrent inhibition of NMDA-receptor neurons by way of GABAergic interneurons . These results suggest that noninvasive imaging of the mismatch negativity may have a direct impact on the clinical care of patients, such as in schizophrenia, especially when combined with other measurements and imaging techniques.


Spectral analysis of cognitive processes using visual stimuli


Bertrand and colleagues have investigated visual cognition using various paradigms and showed that during a visual object recognition, a burst of neural activity near 40 Hz (known as the gamma-band frequency) correlates strikingly with many cognitive acts. The gamma-band activity typically occurs within two time windows after presentation of a visual stimulus: one phase-locked (evoked) to the stimulus at about 180 milliseconds and another, which is not phase-locked to the stimulus (induced), which peaks at approximately 400 milliseconds after stimulus presentation .


The timing of neural events is crucial for normal brain function, and is probably most important for the coordination of cognitive acts, especially for object recognition. Disruption of the precise timing of cortical events may explain the diverse cognitive deficits in dyslexia, depression, and schizophrenia . Neural activity in the gamma frequency band (30 – 80 Hz or broader, centered on 40 Hz) has been hypothesized to reflect the synchronization of neural assemblies involved in the “binding” of various features of an object within a single sensory modality and across modalities, and also across time . At the cellular level, Traub and colleagues have presented evidence that interneurons with GABA A -mediated synaptic transmission interacting with glutamate-mediated cells provide the substrate for gamma oscillations in the cortex. A rich literature suggests that gamma-band activity may bind information across distant sites in the brain, known as the temporal binding hypothesis . Neural modeling, using sophisticated, biologically based simulations based on cellular physiology, suggest a complex interplay in the excitatory and inhibitory cells in generating intrinsic rhythms in the brain, and that different frequency bands have complementary roles in brain function , which is supported by recent cellular research .


Auditory steady-state response


A robust early neural response is the auditory steady-state response (ASSR). It has a maximal peak when the stimulation frequency, typically presented as brief auditory clicks, is presented at 40 Hz. Several studies using the ASSR in schizophrenia show a deficit in the 40-Hz response ( Fig. 2 ) in the ability to follow the stimulus, in frequency and phase . In a separate study, the ASSR was increased in Alzheimer’s disease patients, compared with normal controls, which was suggested to represent decreased inhibition and adaptation in auditory processing .




Fig. 2


ASSR in schizophrenia. ( A ) Magnetic field strength (fT/cm) versus time (stimulus presented at time 0, lasting ∼ 500 milliseconds) in a single MEG channel over the temporal lobe. Note the appearance of middle latency evoked response, N100 m (N1M), and the overriding sustained field. ( B ) Power spectral density in a temporal lobe gradiometer (maximal response for auditory stimuli) over the time interval 0 to 500 milliseconds. Note that in the control subjects, 40 Hz is the largest response (greater than both the 20-Hz and 30-Hz power peaks, often twice the magnitude). In this case, however, the 40-Hz power is less than one half the power in either the 20-Hz or the 30-Hz power, although the 20-Hz response is maximal at a frequency somewhat less than 20 Hz. ( C ) Time frequency representation (TFR) showing the spectral power as a function of time. Sustained neuromagnetic field from MEG gamma-band response to periodic auditory clicks at 40 Hz in a control subject. The color scale shows increased spectral power of the Morlet wavelet (blue to red for increasing power). ( D ) Source estimation localization using the same response to same auditory clicks from A , with the fMR imaging spatial information (not shown) using a Morlet wavelet–based minimum norm estimate centered at 40 Hz at time 230 milliseconds after beginning of click train. It is displayed on an inflated cortical surface of the same subject. The activity appears to involve middle and medial Heschl’s gyrus, extending into the planum temporale. Note that the MEG power spectrum calculated with a Fast Fourier Transform in B does not include any temporal information, unlike the wavelet-transformed measures such as the TFR in C and the spectral spatiotemporal map in D. ( Courtesy of Thomas Witzel, Charlestown, MA.)




Connectivity


Although early sensory processes may be used to measure processes that are involved in disorders of cognition, cognitive functions require precise coordination of neural activity across long distances in the brain. These connections are typically made through large white matter tracts in the brain, which ensure precise timing across distant brain areas. Imaging connectivity in the brain involves structural and functional components, both of which are accessible through functional imaging. The nonstructural connectivity is often further classified as functional connectivity and effective connectivity. Functional connectivity quantifies the correlation of activity between two areas, and effective connectivity measures the influence of one area on another. The covariance approach is a type of effective connectivity of a large-scale interconnected network, in contrast to traditional subtraction methods (stimulus-type contrasts) that assume that there is a change in activation in discrete brain areas .


Structural connectivity


Diffusion tensor imaging


Structural connectivity measures, such as DT imaging, have recently demonstrated the capability to detect subtle white matter microstructural variations across the brain and among individuals, and to find alterations in a number of brain disorders. Microscopic white matter alterations have been observed with DT imaging in a number of conditions, including Alzheimer’s disease, HIV infection, premature birth, chronic alcoholism, autism, multiple sclerosis, schizophrenia, and numerous other conditions . Fractional anisotropy (FA) reflects diffusion anisotropy, presumably from microstructural obstacles that cause water molecules to diffuse in a particular direction. The obstacles that the water may encounter may be caused by the amount and integrity of myelin sheaths around axons, the diameter of the neurons, and the degree of facilitated diffusion that may reflect active transport within neurons.


Diffusion tensor imaging combined with other neuroimaging technologies


To understand the relationship of structure-function, diffusion measures such as FA and function (PET or fMR imaging) may be correlated within subjects or across subjects. The function might be the age of the subject, the timing of a behavioral response, or the latency of an evoked potential. For example, the normal aging prefrontal white matter is selectively susceptible to FA decline, consistent with prefrontal cognitive impairment in normal aging . Antisaccade latency in schizophrenia is strongly correlated with FA in established white matter pathways in the antisaccade processing network.


Another study found a relationship between the timing of behavior, in this case the choice reaction time (CRT), and the physical properties of white matter . Using DT imaging, the FA was found to be correlated regionally with the CRT on a visual task. Specifically, the timing of CRT across a group of young, healthy subjects was highly correlated with the FA in visual projection and association pathways supporting visuospatial attention ( Fig. 3 ). The investigators suggest that the localization of the CRT-FA correlations to predominately visual and parietal fiber tracts, but not to motor pathways or the corpus callosum, indicates that individual differences in visual CRT performance are associated with variations in the white matter underlying the visuospatial attention network, as opposed to pathways supporting motor movement or interhemispheric transmission.




Fig. 3


Cognition and white matter from diffusion MR imaging. The correlation coefficient ( r s ) between CRT and FA is displayed as a map overlaid on the T1 template from a standard atlas. The small frames show axial slices through the optic radiation. The superior-inferior level of the axial slices is indicated by the yellow lines in the sagittal images. The sagittal images at top right are the Montreal Neurological Institute (MNI) individual T1 template ( left ) and the group average FA map ( right ). The correlation map shows the trajectory of the right visual pathway from lateral geniculate nucleus (LGN), through Meyer’s loop (ML), to the optic radiation (OR), terminating at the junction between the optic radiation and the posterior forceps of the corpus callosum (OR-PF). ( From Tuch DS, Salat DH, Wisco JJ, et al. Choice reaction time performance correlates with diffusion anisotropy in white matter pathways supporting visuospatial attention. P Natl Acad Sci USA 2005;102(34):12212–7; with permission.)


Clinically, it has been shown that evoked potentials and evoked magnetic fields may also be correlated with measures from DT imaging in disorders that affect cognition. Evoked potentials and neuromagnetic fields are an effective clinical diagnostic tool in a wide number of neurologic and psychiatric disorders . Despite the clinical usefulness of evoked fields and potentials, the exact underlying mechanisms causing evoked potential abnormalities, such as prolonged latency or reduced magnitude, now are better understood with the use of structural and functional neuroimaging. Recent reports have demonstrated correlations between neuromagnetic and neuroelectric evoked response abnormalities and atrophy of the supporting white matter in a wide range of disorders (eg, multiple sclerosis , schizophrenia , and cognitive impairment in normal aging) . For example, white matter lesion load in multiple sclerosis correlates with prolonged latencies . The concordance between evoked response abnormalities and white matter pathology has promoted a model for neurologic cognitive impairment, based on alterations in nerve conduction velocity. One controversial model maintains that white matter degeneration (eg, demyelination or axon loss) leads to altered nerve conduction velocity, which disrupts the proper temporal synchronization of nerve impulses that are required for normal cognitive function .


Klingberg and colleagues found certain function-structure relationships in the human brain by investigating patients who had reading dyslexia. They reported that the FA values in the bilateral temporal junction were decreased in subjects with poor reading ability, compared with age-matched control subjects. Specifically, the FA was correlated with the reading ability score. This finding suggests that myelination is decreased in or near the classic Wernicke’s area in patients who have reading disorders.


Functional connectivity


Resting state connectivity


A particular case of functional connectivity is the so-called “default mode” or resting state, first characterized using PET and extended with fMR imaging studies. It quantified the spatial correlation of brain activity in the absence of a cognitive task, typically performed by having the subjects fixate on a visual cue, such as a cross-hair, without any other cognitive task . The resting state map is typically computed by calculating the resting state connectivity, using a seed voxel to calculate voxels that are correlated with the seed voxel ( Fig. 4 ). This baseline activity in the resting state has spatial correlations that involve the same prefrontal medial temporal lobe and parietal lobe systems involved in some memory tasks. Although speculative, it may be that the memory system, especially the declarative memory system, might be activated in the so-called “resting state” .




Fig. 4


Resting state networks. Correlation maps are illustrated for the four seed regions. Left middle temporal (L MT+) and right middle temporal (L MT+) are left and right regions defined around the MT+ complex; L HC and R HC are left and right regions defined within the hippocampal formation. Each image shows the mean voxel-wise correlation (a measure of functional connectivity) to the specific region displayed on the left. Four independent participant groups were analyzed. Only positive correlations are shown. Correlation maps are overlaid onto the average anatomic image for each group. Note the consistency of the difference between correlation with MT+ versus HC seed regions and also the reproducibility of the topography across independent data sets. ( From Vincent JL, Snyder AZ, Fox MD, et al. Coherent spontaneous activity identifies a hippocampal-parietal memory network. J Neurophysiol 2006;96(6):3517–31; with permission.)

Only gold members can continue reading. Log In or Register to continue

Mar 28, 2017 | Posted by in NEUROLOGICAL IMAGING | Comments Off on Mapping Cognitive Function
Premium Wordpress Themes by UFO Themes