Page 40 - 2018Fall

Imaging the Listening Brain
Magnetoencephalography and Electroencephalography
Magnetoencephalography (MEG) and electroencephalogra- phy (EEG or M/EEG when combined) record the magnetic fields outside the head and potentials on the scalp, respec- tively. These signals reflect the synchronous activation of thousands or millions of neurons (Hämäläinen et al., 1993) compared with fMRI, M/EEG directly measures the elec- tromagnetic activity of the brain. Both technologies can de- tect activity on the millisecond timescale; this fine temporal resolution is of particular interest when studying auditory processing given the importance of temporal information in audition. However, their spatial resolution is not as fine as fMRI. MEG and EEG have different sensitivity profiles, and when they are used simultaneously, they can provide complementary information about the underlying cortical activities (Sharon et al., 2007).
Magnetoencephalography and Electroencephalography Neural Source Analysis Interpreting M/EEG data is a challenge. One can start by analyzing the M/EEG topographical patterns directly (ex- trapolated from the measurements at each sensor location). However, we face yet another ill-posed problem similar to the cocktail party problem. It is physically impossible to completely and uniquely determine which brain areas were active given the measured M/EEG data. Still, there are two general approaches to relate M/EEG data to the neural sources: (1) source localization and (2) inverse imaging.
The source localization approach assumes that only a small handful of brain regions are active and uses equivalent cur- rent dipoles to represent that activity when modeling the M/EEG data. This approach is often favored when the ex- perimenter wants to make inferences about how activity at specific regions of interest (e.g., the bilateral auditory cor- tex) differs across experimental conditions (e.g., actively at- tending to a sound vs. passively listening while watching a silent movie). However, a disadvantage of this approach is that other neural activities originating outside the modeled dipoles can change the estimated activity at the modeled lo- cations, leading to erroneous interpretations of results (e.g., failing to account for premotor activities in an auditory at- tention task can influence our interpretation of the atten- tional modulated activities modeled in the auditory cortex; Bharadwaj et al., 2014).
Inverse imaging is an alternative approach that also mod- els M/EEG data, but unlike source localization, it estimates activity across the entire brain. Using other information to constrain the ill-posed problem (e.g., by constraining the lo- cation of neural activity to the cortex by using the individ- ual’s MRI structural scans) allows the experimenter to map M/EEG sensor data to neural sources directly interpretable on the cortex. However, a disadvantage of this approach is the idiosyncrasies associated with the specific inverse model chosen; for example, the often-used minimum-norm model produces estimates of brain activity that are potentially over- smoothed in space.
Electrocorticography
Electrocorticography (ECoG), sometimes also called intra- cranial EEG (iEEG), measures brain signals using electrodes that are implanted either above or below the dura, the tough outermost membrane enveloping the brain. Before surgi- cal intervention, some epilepsy patients are implanted with ECoG electrodes to better localize the part of the brain re- sponsible for the seizure onsets. These patients sometimes volunteer for different sensory and cognitive tasks while they are in an extended hospital stay, which provides an op- portunity for experimenters to record brain activity from a relatively large patch of cortex with unmatched temporal and spatial resolution. For ethical reasons, the sites of re- cording are dictated by clinical needs, and thus, cortical cov- erage varies across patients.
Different Imaging Techniques to Answer
Different Questions
As outlined above, each recording technique has its own strengths and weaknesses (Figure 3), so experimenters choose different imaging techniques to answer different questions. For example, to carefully examine the subdivi- sions of the human auditory cortical areas, fMRI is often preferred due to its superior spatial resolution. Conversely, to track how the cortical dynamics change with ongoing sound, MEG and EEG are preferred due to their high tem- poral resolution. ECoG has both fine spatial and temporal resolution, but the cortical coverage is limited and out of the hands of the experimenter. This limits the opportunity to address an array of system neuroscience questions such as how one area of the brain is functionally correlated with another region to coordinate active listening. Overall, the experimental question must be matched with an appropri- ate recording modality.
38 | Acoustics Today | Fall 2017