Our interactions with our natural environment rely on a myriad of cognitive functions. These include auditory, visual & other sensory perception, motor behaviour, attention, memory & learning, executive functions, speech & language and consciousness & awareness. Cognitive neuroscience investigates how the brain accomplishes these cognitive functions, i.e. their underlying neural processes. Given their complexity, it is not surprising that cognitive processes are mediated by interactions amongst several regions within widespread neural systems.
To establish relationships between cognitive functions and neural structures, research in cognitive neuroscience is therefore governed by two fundamental, yet complementary principles. First, from the perspective of functional specialization, we aim to identify neural structures that are specialized for processing specific types of information or functional components (e.g. motion area V5). Second, from the perspective of functional integration, we aim to characterize how complex cognitive functions emerge from interactions amongst these specialized subsystems.
Structure-function mapping is complicated by the fact that there is no isomorphism between cognitive function and neural structure. Instead, a region may be involved in multiple cognitive functions – also coined pluripotentiality. Conversely, multiple neural systems may be able to perform the same cognitive task. Hence, structure-function mapping in cognitive neuroscience requires the combination of functional imaging techniques with permanent or transient (transcranial magnetic stimulation, TMS) lesion approaches. Furthermore, since the underlying neural processes evolve at multiple timescales we need to integrate findings from electrophysiological techniques with high temporal resolution (e.g. MEG, EEG) and imaging techniques with high spatial resolution (e.g. MRI, fMRI, PET).
To gain a more informed understanding of the underlying neural processes, functional imaging is combined with computational modelling. This allows one to relate parameters from the computational model with signatures of neural activity as measured with non-invasive imaging techniques. In the following, we will briefly review the various imaging methods employed in cognitive neuroscience
Hemodynamic metabolic methods such as PET and fMRI are based on increases in blood flow accompanying neuronal activity. Local hemodynamic effects occur when changes in the activity of a neural population lead to changes in metabolic demands. Thus, a change in timing (e.g. synchronous vs. asynchronous) may have little or no hemodynamic correlate despite its likely functional significance. Two methods, fMRI and PET are commonly employed to measure hemodynamic correlates of neural activity.
Positron emission tomography (PET) utilizes a variety of radioactively labelled biological probes (e.g. H215O, 18F-FDG, 18F-Dopa) to provide insight into physiological (e.g. blood flow), metabolic (e.g. glucose metabolism) or neurotransmitter processes (e.g. dopamine receptors). Most widely used are H215O PET studies that measure regional cerebral blood flow (rCBF) in order to make inferences about neuronal activations. As the H215O tracer has a relatively short half-life of 2.07 minutes, the isotope is produced in a cyclotron close to the PET scanner. It is introduced into the human body by intravenous injection. The distribution of these molecules in brain tissue can then be measured by making regional measurements of PET counts in the brain.
Basically, the isotopic nuclei are unstable because of their excessive positive charge and will decay by emitting a positron and a neutrino. The positron will travel a short distance, lose energy by colliding with atomic electrons and finally annihilate with an electron by producing two 511-keV gamma rays (i.e. photons) that are emitted with an angular separation of 180°. Due to their high energy, most of these photons will not be attenuated by the brain tissue or the skull and can thus be detected by the array of scintillation detectors around the head, which constitute the recording apparatus of the PET scanner. The basic logic is to acquire PET images for two experimental conditions that differ only in the cognitive operation of interest (e.g. listening to words vs. listening to reversed words). The two sets of images are then contrasted on a pixel by pixel basis to identify neural structures in which the regional cerebral blood flow (rCBF) differs across the two conditions.
Functional magnetic resonance imaging (fMRI) is primarily based on the Blood Oxygenation Level Dependent (BOLD) contrast, which makes use of the fact that the magnetic susceptibility is greater for deoxyhemoglobin than oxyhemoglobin. Since the blood supply triggered by increases in neural activity exceeds the amount of oxygen that is utilized by the neural activity, neural activity usually leads paradoxically to an increase in oxyhemoglobin concentration. This regional change in the ratio of oxy- and deoxyhemoglobin can then be measured non-invasively using fMRI. In contrast to PET, fMRI enables experimental designs that go beyond simple block designs and present stimuli in a randomized fashion. This is very important for many cognitive neuroscience experiments that are intended to dissociate sustained state (e.g. attention) from transient stimulus-driven activity or where the conditions are defined based on subjects responses.
Unlike hemodynamic metabolic methods, electrophysiological methods measure neural activity directly. EEG and MEG take advantage of the fact that a pool of neurons that are oriented along the same direction and are depolarized synchronously create an electromagnetic field that can be measured from outside the head using the electroencephalogram (EEG) or the magnetoencephalogram (MEG). Hence, MEG and EEG can define the underlying cortical neuronal events in real-time (10-100 msec).
However, both MEG and EEG suffer from a number of draw-backs. Most notably, since they can infer electric/magnetic sources within the brain only indirectly from measurements on the scalp, they provide relatively poor spatial resolution. In fact, localization of sources of neural activity from the time-varying magnetic or electric fields is even an ill-posed problem (also referred to as inverse problem) and can only be approximately solved by imposing anatomical (or other e.g. smoothness) constraints based on additional information from fMRI/PET or lesion studies.
Furthermore, EEG and MEG can only measure synchronous neural activity of neuronal populations that are oriented such that the individual neuronal dipole moments summate. This precludes detection of neural activity in neural structures with random cell orientation such as the basal ganglia. Finally, since electric and particularly magnetic fields rapidly decline with the distance from the source, the neural activity in deep brain structures is less detectable for M/EEG. In addition to these common drawbacks, unlike EEG, MEG is sensitive only to sources oriented tangentially to the scalp, so that the event-related magnetic field primarily reflects neuronal activity in the cortical sulci.
Transcranial Magnetic Stimulation (TMS) is a technique that enables stimulation of the human brain non-invasively with limited discomfort for the subject. It is based on the principles of electromagnetic induction. By applying a brief pulse of current through a stimulation coil held on the subject’s scalp, a time-varying magnetic field perpendicular to the current is produced. In turn, this magnetic field induces an electrical current in the brain. Depending on the stimulation parameters, the current induced will transiently interfere with cortical activity, which can result in altered behavior.
The TMS equipment consists of a pulse generation unit, the stimulator, and an electromagnetic stimulation coil. The most widely used types of coil are the circular coil and the figure-of-eight coil. Figure-of-eight coils can be approximately viewed as two round coils mounted side to side, with the current rotating in opposite directions in the two coils. While with circular coils the maximum electric field induced in the brain lies in an annulus under the coil, the electric field with figure-of-eight coils is strongest under the centre of the coil. Figure-of-eight coils thus maximize the focality of TMS.
Another important feature of TMS is that the induced magnetic field is inversely proportional to the square of the distance between the coil and the cortex, meaning that only superficial neural structures are sensitive to direct stimulation, whereas deeper areas might only be excited indirectly by propagated activity from the region beneath the coil. TMS can be applied as single pulses or as train of pulses (rTMS). The effects of single TMS pulses are short-lived and several seconds are needed before the next pulse can be delivered. Conversely, rTMS can modulate the excitability of the stimulated area even beyond the duration of the TMS application with the specific effect depending on the prior history of activation.
Despite the extensive use of TMS, the exact mechanisms still need to be defined. Two main mechanisms have been proposed. The virtual lesion approach assumes that the application of a TMS pulse temporarily lesions or disturbs the function of a population of neurons. Alternatively, TMS may induce neural activity by adding ‘extra activity’ to ongoing processes. Both mechanisms will alter neural processing and may manifest themselves in behavioral changes such as longer/shorter responses, errors, lowered detection threshold. Yet, TMS may not only change local neuronal activity directly, but also affect processing in remote brain regions either via neural interregional connections or due to compensatory mechanisms.
Such remote effects are not revealed by standard TMS studies, where inference is usually limited to the targeted site of stimulation. Instead, the combination of TMS and fMRI (e.g. interleaved TMS & fMRI) allows one to monitor the effects of TMS on neural activity directly not only at the site of stimulation but also within the entire brain, allowing also for inferences about the connectivity between the stimulated and co-activated areas.
Cognitive neuroscience relies on the combination of functional imaging methods with high temporal (EEG, MEG) and spatial (fMRI, PET) resolution. Furthermore, structure-function relationships need be identified in an iterative procedure involving functional imaging and transient (TMS) or permanent lesion approaches. Most importantly, research in cognitive neuroscience is guided by clearly defined hypotheses that are based on our current theories and understanding of cognitive functions and their underlying neural processes. Thoughtful, creative and innovative experimental designs are thus required that allow us to dissociate and selectively manipulate individual cognitive processes that contribute to the complex tasks we are facing in our natural environment.