Research Articles, Systems/Circuits

Electrical Stimulation of Temporal and Limbic Circuitry Produces Distinct Responses in Human Ventral Temporal Cortex

Harvey Huang, Nicholas M. Gregg, Gabriela Ojeda Valencia, Benjamin H. Brinkmann, Brian N. Lundstrom, Gregory A. Worrell, Kai J. Miller and Dora Hermes
Journal of Neuroscience 14 June 2023, 43 (24) 4434-4447; https://doi.org/10.1523/JNEUROSCI.1325-22.2023

Abstract

The human ventral temporal cortex (VTC) is highly connected to integrate visual perceptual inputs with feedback from cognitive and emotional networks. In this study, we used electrical brain stimulation to understand how different inputs from multiple brain regions drive unique electrophysiological responses in the VTC. We recorded intracranial EEG data in 5 patients (3 female) implanted with intracranial electrodes for epilepsy surgery evaluation. Pairs of electrodes were stimulated with single-pulse electrical stimulation, and corticocortical evoked potential responses were measured at electrodes in the collateral sulcus and lateral occipitotemporal sulcus of the VTC. Using a novel unsupervised machine learning method, we uncovered 2-4 distinct response shapes, termed basis profile curves (BPCs), at each measurement electrode in the 11-500 ms after stimulation interval. Corticocortical evoked potentials of unique shape and high amplitude were elicited following stimulation of several regions and classified into a set of four consensus BPCs across subjects. One of the consensus BPCs was primarily elicited by stimulation of the hippocampus; another by stimulation of the amygdala; a third by stimulation of lateral cortical sites, such as the middle temporal gyrus; and the final one by stimulation of multiple distributed sites. Stimulation also produced sustained high-frequency power decreases and low-frequency power increases that spanned multiple BPC categories. Characterizing distinct shapes in stimulation responses provides a novel description of connectivity to the VTC and reveals significant differences in input from cortical and limbic structures.

SIGNIFICANCE STATEMENT Disentangling the numerous input influences on highly connected areas in the brain is a critical step toward understanding how brain networks work together to coordinate human behavior. Single-pulse electrical stimulation is an effective tool to accomplish this goal because the shapes and amplitudes of signals recorded from electrodes are informative of the synaptic physiology of the stimulation-driven inputs. We focused on targets in the ventral temporal cortex, an area strongly implicated in visual object perception. By using a data-driven clustering algorithm, we identified anatomic regions with distinct input connectivity profiles to the ventral temporal cortex. Examining high-frequency power changes revealed possible modulation of excitability at the recording site induced by electrical stimulation of connected regions.

Introduction

Distributed networks in the human brain, interacting with each other at multiple scales, are the basis for our interactions with the sensory environment (Edelman and Mountcastle, 1982; Goldman-Rakic, 1988; Mesulam, 1990; Bressler, 1995; Sporns et al., 2004). Throughout the visual stream, feedforward processes from the stimulus-driven primary visual cortex converge with feedback influences from higher-level cognitive networks involved in attention, memory, and expectation (Moran and Desimone, 1985; Felleman and Van Essen, 1991; Motter, 1993; Schlack and Albright, 2007; McManus et al., 2011). Convergence of these processes in the ventral temporal cortex (VTC) permits object recognition and the encoding of perceptual content, such as faces and scenery (Kanwisher et al., 1997; Epstein and Kanwisher, 1998). Functional connectivity within these networks has been established primarily through correlated signals during behavioral tasks and at rest, through a variety of electrophysiologic and imaging modalities (Friston, 1994; Kay and Yeatman, 2017). Measurements and analyses that can causally and independently probe different connections provide a novel angle to understanding the numerous inputs to the VTC.

An increasingly common method to characterize brain connectivity has been to measure evoked potentials in awake humans in response to single-pulse electrical stimulation through intracranial electrodes (Matsumoto et al., 2004; Lacruz et al., 2007; Kundu et al., 2020). Termed corticocortical evoked potentials (CCEPs), this technique enables directional quantification of connectivity between any pair of implanted sites. In other words, it is a perturbational approach to determining effective connectivity. By measuring responses at one recording site of interest, we can calculate, with high temporal resolution, the influence of electrical stimulation from multiple stimulated electrodes located across different brain networks (Miller et al., 2021). Earlier responses in CCEPs represent signals propagated through direct pathways, while later responses represent signals propagated through indirect or recurrent cortico-subcortico-cortical pathways (Matsumoto et al., 2004; Keller et al., 2014b; Araki et al., 2015). This approach leverages CCEPs to generate a global map of converging brain inputs to a recording site of interest.

Previous studies have primarily quantified CCEP strength by measuring the amplitude or latency to negative or positive peaks within predetermined time intervals (e.g., the early “N1” peak) (Matsumoto et al., 2004, 2007; Krieg et al., 2017; Kundu et al., 2020; Silverstein et al., 2020), but this approach often considers only the fastest, direct pathways and neglects other response shapes that are often present. For instance, prior studies have highlighted that stimulation at nodes within different networks produces different CCEP waveforms (Shine et al., 2017; Veit et al., 2021). A recent technique has captured this waveform diversity by performing unsupervised clustering on CCEPs based on temporal shape (Miller et al., 2021). A unique canonical shape, termed basis profile curve (BPC), is calculated for CCEPs from each cluster of stimulation sites to the same measurement site of interest. Temporal deflections in intracranial EEG (iEEG) represent synchronous synaptic activity at the measurement site (Mitzdorf and Singer, 1978). Since the measurement site is held constant, different temporal motifs across BPCs may indicate activation of different cortical layers in the same ensemble of neurons, or different ensembles of neurons altogether around the measurement electrode. Uncovering these input differences has significant functional and anatomic implications. For instance, the laminar patterns of neuronal inputs, as assessed by injected tracers, correlate with feedforward and feedback connectivity within the visual system (Felleman and Van Essen, 1991; Markov et al., 2014). We use the BPC method here as a means of uncovering the signatures of such afferent inputs in awake human participants.

As the human VTC simultaneously receives inputs from many networks to accomplish its goal of visual perception, it serves as an optimal candidate for disambiguating network inputs by single-pulse electrical stimulation. We have measured CCEPs at electrodes in the VTC to parse input connectivity from independently stimulated sites across multiple brain networks. By applying the BPC method, we separate those inputs into distinct physiologically relevant categories informed by temporal shape.

Materials and Methods

Subjects

This research was performed in accordance with the Mayo Clinic Institutional Review Board (IRB #15-006530), which also authorizes sharing of the data. Each patient/representative voluntarily provided independent written informed consent/assent to participate in this study.

iEEG voltage data were measured in 5 human subjects (3 female, see Table 1) who had been implanted with stereo EEG (sEEG) electrodes for epilepsy monitoring and had electrodes placed in the VTC. Recorded data were filtered between 0.01 and 878 Hz and then digitized at 2048 Hz on a Natus Quantum amplifier.

View this table:
Table 1.

Subject demographicsa

Electrode localization

Subject preoperative T1 MRIs were first transformed into AC-PC space through affine transformations and trilinear voxel interpolation (Huang et al., 2021). sEEG electrodes were then localized from the postoperative CT scans and coregistered to the T1 MRIs (Hermes et al., 2010), so that electrode positions were in AC-PC space.

Subject T1 MRIs were segmented using the autosegmentation algorithm in Freesurfer 7 (Dale et al., 1999), which also produced a 3D mesh rendering of the pial surface. Gyral and sulcal labels were generated for each subject's pial surface, during autosegmentation, by aligning surface topology to the Destrieux cortical atlas (Destrieux et al., 2010). Electrodes were matched to cortical or subcortical labels corresponding to the most frequent voxel label within a 3 mm radius, and these labels were visually reviewed for correct assignment. Electrodes were visualized in AC-PC space on individual subject T1 MRI slices and on subject pial and inflated pial surfaces.

For visualization purposes, electrode positions were also transformed to the standard MNI 152 space using nonlinear segmentation-based normalization of the T1 scan in SPM12 (Penny et al., 2011) (https://www.fil.ion.ucl.ac.uk/spm/), so that recording and stimulation sites across subjects can be collected on the same MNI 152 pial surface rendering and MNI 152 T1 MRI slices. Positions in the right hemisphere were reflected across the mid-sagittal (yz) plane when rendered to the MNI 152 pial surface, so that all positions would be visible from one angle on a single (left) hemispheric rendering.

Electrode of interest selection

After preprocessing, we studied stimulation inputs to measurement electrodes in the collateral sulcus (Destrieux label S_oc-temp_med_and_Lingual) and the lateral occipitotemporal sulcus (Destrieux label S_oc-temp_lat). Electrodes that were in seizure onset zones, as indicated by physician records during intracranial monitoring, were excluded from analysis. The collateral sulcus and occipitotemporal sulcus were implanted in multiple subjects and allowed us to test the hypothesis that inputs cluster systematically across subjects.

Experimental design and preprocessing

Electrode pairs were stimulated up to 36 times each (average 10) with a single biphasic pulse of 200 μs pulse width and either 4 or 6 mA amplitude every 3-7 s, using a Nicolet Cortical Stimulator. Stimulation sites that overlapped seizure onset zones, per physician records, were excluded from analysis. Electrodes and individual trials were then visually inspected, and those that contained artifacts or epileptiform activity were also excluded from analysis. Finally, stimulation sites with <8 remaining stimulation trials were excluded from analysis. Table 1 shows the final number of stimulation sites analyzed in each subject. This number includes stimulation sites that were excluded in the analysis of individual measurement electrodes, when the stimulation pair contained the measurement electrode. Data were structured to conform to the iEEG-BIDS format (Holdgraf et al., 2019).

All preprocessing was performed using custom MATLAB code (see Code availability, below). Data in all subjects were high-pass filtered above 0.5 Hz by forward-reverse filtering with a second-order Butterworth filter and then rereferenced on a trial-by-trial basis to a modified common average that avoids introducing large early stimulation-evoked potentials in other electrodes. Specifically, the reference was the mean of all channels that were not stimulated, excluding noisy channels whose variance between 500 and 1000 ms after stimulation was in the top 5 percentile, as well as channels with a large initial response, identified as having variance between 15 and 100 ms after stimulation that is both (1) in the top quartile and (2) >1.5-fold the variance between 500 and 1000 ms after stimulation. The common average reference was calculated separately for each 64-channel headbox to account for differences in headbox-specific noise. Line noise at 60 Hz and the first two harmonics were then removed using the spectrum interpolation method implemented by Mewett et al. (2001). Finally, the median baseline voltage value from 500 to 50 ms before stimulation was subtracted from each individual trial and channel.

Subject-level BPCs

For each measurement electrode in each subject, BPCs were calculated from the single-pulse CCEP data based on the method described by Miller et al. (2021), with two slight modifications that are noted in the outline below. Each BPC is a single unit-normalized curve that characterizes the temporal shape common to CCEPs from a set of stimulation sites to the measurement electrode. It is also possible for some stimulation sites (usually those with weak CCEP responses) to be not assigned to any BPC. Figure 1 provides a brief illustration of the modified process, described here. We began with a matrix of preprocessed voltage data corresponding to all stimulation trials recorded at one measurement electrode, in this case in the collateral sulcus or lateral occipitotemporal sulcus (Fig. 1A). Each trial was then weighted by an exponential decay function with a time constant (τ) of 100 ms (Fig. 1B). The weighting by exponential decay places more value on deflections that are the result of direct, rapid, responses, without eliminating indirect responses that are later in the signal. This weighting step was not present in the original BPC algorithm, but it allowed us to use a relatively large window of CCEP data for analysis, without concern that later stochastic noise might drown out meaningful signal in the first 100 ms. The interval between 11 and 500 ms after stimulation was extracted for all weighted trials as a matrix, V, to use as input for the main BPC algorithm (Fig. 1B, in red). V has dimensions K × T, with K total stimulation trials and T total time points between 11 and 500 ms. The 11 ms start time was chosen as it was the earliest time point after the sharp transient stimulation artifact, which was consistent across all CCEP trials and identified visually by overlaying all CCEP trials in one plot (Extended Data Fig. 1-1). According to the BPC algorithm, a scalar projection matrix was calculated between all pairs of trials, P=VV̂T, where V̂ denotes V with all trials l2-normalized (Fig. 1C). All pairwise projections in P were then grouped into sets by stimulation site pairs, with self-projections omitted, and the t value of each set (mean divided by SE) was calculated as an entry in the significance matrix, Ξ (Fig. 1D). This significance matrix has dimensions N × N, with N unique stimulation sites, and quantifies the temporal similarity between inputs from all pairs of stimulation sites (e.g., CCEPs from stimulation sites 1 and 2 in Fig. 1A are highly similar in shape to each other and thus had a high t value in Ξ). Ξ was subjected to a non-negativity constraint, with negative t values set to 0, and non-negative matrix factorization (NNMF) was performed to decompose Ξ into component matrices W and H, such that Ξ ∼ WH (Lee and Seung, 1999) (Fig. 1E). W has dimensions N × Q, and H has dimensions Q × N. The inner dimension Q, which determines the final number of BPCs, was always initialized at 6 and then reduced iteratively. The goal of NNMF is to minimize the error, η = |Ξ −WH|2, with the non-negativity constraint in both W and H. NNMF is necessary to ensure that temporal dynamics from each cluster do not contribute both positively and negatively to CCEPs recorded at the same electrode (i.e., laminar anatomy is not invertible). To minimize redundancy in the dimensionality of the NNMF output, we iteratively reduced the inner dimension, Q, until the sum of the upper-half off-diagonal elements in HHT(ζ) was ≤1. In the original BPC algorithm, ζ was defined as the maximum of the upper-half off-diagonal elements in HHT, and the threshold on this maximum value was set at 0.5. Using the sum rather than the maximum sets a global constraint on overall shared structure across all BPCs. For stability, NNMF was rerun 50 times with randomly generated initial W and H matrices at each value of Q, and the decomposition with minimum error η was chosen in each case. This ultimately resulted in a low-rank decomposition of the CCEP signals. The Q rows in the final H corresponded to Q BPC groups, and the N columns in H corresponded to the ordered list of N stimulation sites. A “winner-take-all” operation assigned each stimulation site column to the row in H with the largest coefficient if that value exceeded 1/2N, and to no BPC group otherwise (Fig. 1E). BPC shapes were then found as the first principal component of CCEPs from each NNMF group of stimulation sites, calculated using Linear Kernel PCA (Fig. 1F), and they were visualized after multiplication by the reciprocal of the exponential weighting function (Fig. 1G). For example, stimulation sites 1 and 2 in Figure 1A grouped into the second BPC (cyan) shown in Figure 1G. If a BPC was assigned for a stimulation site, a numerical signal-to-noise ratio (SNR), quantifying the fit of the BPC relative to residual noise, was also calculated for each trial at that stimulation site as follows: SNR=αk‖εk(t)‖2 where the kth CCEP trial Vk(t)=αkB(t)+εk(t), and B(t) was the BPC assigned for that stimulation site. The mean SNR across all trials of a stimulation site is referred to subsequently as the SNR of that stimulation site.

Figure 1.
Figure 1.

Modified procedure for identifying BPCs from CCEP data. A, Voltage data (positive is up) were recorded in a convergent paradigm from a fixed electrode while multiple electrode pairs were stimulated (top) to yield input CCEP data (bottom). Multiple stimulation trials (average = 10) were performed at each stimulation site. Top, Adapted from Miller et al. (2021). B, Each CCEP trial was multiplied by an exponential decay weighting function with time constant, τ, equal to 100 ms. All CCEPs from 11 to 500 ms after stimulation were used as input matrix, V, to calculate BPCs. C, Scalar projection matrix, P, calculated between all pairs of trials. D, Significance matrix, Ξ, was calculated from P and quantified the similarity in measured CCEP shape between pairs of stimulation sites. E, NNMF on Ξ with iterative reduction in component dimensionality resulted in a low-rank, non-negative decomposition of all CCEP signals. The winner-take-all operation assigned stimulation sites to BPC groups. F, Each BPC corresponded to the first principal component of all CCEPs assigned to its respective group. G, BPCs were multiplied by the reciprocal of the exponential weighting function in B for visualization.

Figure 1-1

The transient stimulation artifact was identified visually by overlaying all CCEP trials in one plot, shown here for Subject 1 electrode 1 (376 trials, each a single curve). The artifact ended just before 11 ms for all subjects. Download Figure 1-1, TIF file.

Group-level consensus BPCs

We calculated a single set of “consensus” BPCs across all collateral sulcus electrodes in Subjects 1-4, as follows. All BPCs across electrodes and subjects were concatenated to form a matrix, A, where columns correspond to BPCs and rows correspond to time points. Singular value decomposition was applied to A, to yield orthogonal eigenvectors of AAT in descending order of variance explained. This equated to principal component analysis without centering on features (time points). The first three eigenvectors of AAT formed a low-dimensional basis on which subject BPCs were clustered using K-means clustering. K = 4 clusters were chosen based on visual inspection of the subject BPCs projected to the first three eigenvectors. The consensus BPCs were obtained by projecting the centroid of each cluster back into signal space and labeled numerically by order of increasing projection coefficient to the first principal component. Each subject BPC was assigned to a consensus BPC based on its K-means cluster. This ultimately yielded consensus BPC labels for all stimulation sites that possessed subject-level BPC labels, across all input electrodes and subjects. Leave-one-subject-out cross-validation was used to evaluate the robustness of this model in the absence of individual subjects. BPCs from each of the 4 subjects were withheld independently as a test set, while singular value decomposition and K-means clustering steps were performed on BPCs from the remaining 3 subjects to yield consensus BPCs for that cross-validation fold. To maintain naming consistency to the full model, these consensus BPCs were labeled by order of increasing similarity (scalar projection) to the first principal component of the full model. BPCs in the test set were projected onto the principal component space created from the other subjects, and each BPC was assigned the consensus BPC label of its closest K-means centroid. Accuracy was evaluated as the proportion of test set BPCs predicted to have the same consensus BPC label as in the full consensus BPC model.

For each consensus BPC, we tallied the number of stimulation sites that were located in each of six large cortical or subcortical regions. Stimulation sites whose SNR were <1 were excluded from this analysis. First, each stimulation site was matched to an anatomic label in a similar way as for electrodes, with the position of each stimulation site defined as the midpoint position between the pair of stimulated electrodes. Next, gyral and sulcal labels (based on the Destrieux atlas) were binned together into six regions as follows. The ventral temporal region included the fusiform gyrus (G_oc-temp_lat-fusifor), lingual gyrus (G_oc-temp_med-Lingual), parahippocampal gyrus (G_oc-temp_med-parahip), lateral occipitotemporal sulcus (S_oc-temp_lat), collateral sulcus (S_oc-temp_med_and_Lingual), anterior collateral sulcus (S_collat_transv_ant), and posterior collateral sulcus (S_collat_transv_post). The lateral temporal region included the inferior temporal gyrus (G_temporal_inf), middle temporal gyrus (G_temporal_middle), superior temporal gyrus (G_temp_sup-Lateral, G_temp_sup-Plan_polar, G_temp_sup-Plan_tempo), temporal pole (Pole_temporal), Heschl's gyrus (G_temp_sup-G_T_transv), inferior temporal sulcus (S_temporal_inf), superior temporal sulcus (S_temporal_sup), and transverse temporal sulcus (S_temporal_transverse). The insula included the long insular gyrus and central sulcus of the insula (G_Ins_Ig_and_S_cent_ins), short insular gyri (G_insular_short), and circular sulcus of the insula (S_circular_insula_ant, S_circular_insula_inf, S_circular_insula_sup). The hippocampus and amygdala were each their own region, and the “other” category included all other anatomic labels. χ2 test of independence was used to test for independence between region and consensus BPC across all stimulation sites (with robust SNR ≥ 1). One-way ANOVA and post hoc Tukey's HSD tests for multiple comparisons were used to determine whether significant differences in mean Euclidean distance between stimulation site and measurement electrode existed across consensus BPCs (and stimulation sites not assigned to BPCs). Pearson's correlation was used to test for a relationship between Euclidean distance from measurement electrode and BPC SNR for all stimulation sites assigned to BPCs, for stimulation sites within each consensus BPC separately, and also across all consensus BPCs together.

We investigated whether similar consensus BPC shapes and anatomic distributions would result if the BPC algorithm were applied to CCEP data concatenated across all collateral sulcus measurement electrodes in Subjects 1-4 in a single-step analysis. For this, CCEP data at each measurement electrode were normalized by the SD across all trials, to account for subject differences in iEEG amplitude (Mercier et al., 2022), before being concatenated across subjects. CCEP trials recorded from different measurement electrodes in the same subject but evoked by the same stimulation site were pooled under one stimulation site condition for the BPC algorithm, resulting in fewer total stimulation sites than in the main consensus BPC analysis. We relaxed the threshold on ζ to allow for the same number of clusters as the main consensus BPC analysis, so as to be directly comparable. χ2 test of independence was used again to test for independence between region and consensus BPC across all stimulation sites with SNR ≥ 1.

While subject-level BPCs were calculated for each of the four lateral occipitotemporal sulcus electrodes in Subjects 4 and 5, consensus BPC analysis was not performed on these BPCs as we deemed it underpowered to generalize from the two subjects.

Spectral and broadband analysis by consensus BPC category

Before calculation of wavelet spectrograms and broadband estimates, the stimulation artifact was removed from each CCEP trial so as to mitigate distribution of the artifact in time bins around stimulation during wavelet transformation and IIR filtering. Similar to the method used by Crowther et al. (2019), we removed the artifactual data between −11 and 11 ms around stimulation, and replaced it with a sum of reversed tapered copies of the preceding (−22 to −11 ms) and following (11-22 ms) signals. These replaced data have the same amplitude and spectral distribution as that of the background signal.

Induced (i.e., single trial-level) wavelet spectrograms were calculated for each CCEP trial using the MATLAB cwt package on the preprocessed time series data. Power was calculated for each time-frequency bin as the square of the amplitude of the wavelet transform. Power was then normalized to baseline separately for each frequency by dividing by the mean power between 500 and 100 ms before stimulation. An average spectrogram was determined for each stimulation site by taking the geometric mean across CCEP trials. A single average t statistic spectrogram was then calculated for each consensus BPC by collecting the average log10-transformed spectrograms from all stimulation sites assigned to that consensus BPC and calculating the one-sample t statistic at each frequency-time bin against the null hypothesis of 0 (no log-fold change over baseline). As in the anatomic analysis, stimulation sites whose SNR were <1 were excluded from this analysis. An average t statistic spectrogram was also calculated in the same way from all stimulation sites that were algorithmically excluded from BPC assignment, as a negative control.

A time-varying broadband estimate was calculated for each CCEP trial, as follows. For each recording site in each subject, the preprocessed and artifact-removed signals were bandpass filtered by forward-reverse filtering (MATLAB filtfilt) using a third-order Butterworth filter between 70 and 170 Hz. Subsequently, the broadband estimate was calculated as the log10 squared absolute value (log power) of the Hilbert transform of each bandpass filtered signal. Each trial was normalized by baseline by subtracting the mean broadband estimate between 500 and 100 ms before stimulation. The result was then averaged across trials for each stimulation site, and then averaged again across all stimulation sites assigned to a given consensus BPC to yield a single time-varying broadband estimate for each consensus BPC. An average broadband estimate was also calculated the same way from all stimulation sites that were algorithmically excluded from BPC assignment, as a negative control.

Basis profile spectrograms

Unique spectral patterns and anatomic distributions might emerge if time-frequency data, instead of voltage data, were used as the basis for clustering by the BPC algorithm. This is motivated by differences in the physiology underlying evoked potentials and power changes in frequency bands (Winawer et al., 2013; van Kerkoerle et al., 2014). To accomplish this, the BPC algorithm was applied to the wavelet spectrogram data to yield “basis profile spectrograms” (BPSs). For each collateral sulcus measurement electrode, the induced wavelet spectrograms from all CCEP trials were individually calculated, normalized to their respective baselines, and log10-transformed, as in the above subsection. Each trial spectrogram was downsampled in time by a factor of 16 to reduce dimensionality. All time-frequency bins between 12.5 and 200 Hz and between 100 and 500 ms after stimulation were flattened to yield a length-2091 vector for each trial. The flattened trials were vertically concatenated and input as V into the BPC algorithm in the same way as was voltage data (see Subject-level BPCs) to yield BPSs, after unflattening, as well as SNR values quantifying the fit of each trial and stimulation site to its associated BPS. Unlike the voltage data, the spectrograms were not exponentially weighted before the application of the BPC algorithm. No analogous “consensus” BPSs were calculated across subjects.

For each measurement electrode, we quantified the similarity in stimulation site distributions between the BPC assignments and the BPS assignments using the Adjusted Rand Index (ARI) (Hubert and Arabie, 1985). ARI is bounded between −1 and 1. It is equal to 1 when two methods yield identical clusters, and it is close to 0 when the two methods yield random clusters (Steinley, 2004; Yang, 2017). ARI does not require that the two methods yield the same number of clusters or the same cluster labels. In our analysis, we considered all stimulation sites not assigned a BPC to be in one collective “excluded” cluster, and likewise for the BPS stimulation sites. To compare only the top 50% most robust BPC and BPS assignments, ARI calculations were repeated after disqualifying, separately for BPC and BPS assignments, those with SNR in the bottom 50 percentile of values calculated across all measurement electrodes. Stimulation sites with disqualified BPC or BPS assignments were effectively reassigned to the “excluded” BPC or BPS cluster in this repeat ARI calculation.

Statistical analysis

Time-frequency bins significantly different from rest in the consensus BPC t statistic heatmaps and samples significantly different from rest in the consensus BPC broadband estimates, by one-sample t tests, were controlled for multiple comparisons by using a 5% false discovery rate (FDR) under dependency, as implemented by Benjamini and Yekutieli (2001). This method was chosen as nearby time-frequency bins often exhibit positive or negative regression dependency with each other. All other statistical analyses are described above in their respective Materials and Methods subsections.

Data and code accessibility

The data that support the findings of this study are available, on manuscript publication, in BIDS format on OpenNeuro (https://openneuro.org/datasets/ds004457). In addition, the OpenNeuro dataset also contains quality check figures supporting our line noise removal algorithm and the validity of the BPS analysis. The code used to generate all results and figures is available on GitHub (https://github.com/hharveygit/VTCBPC_JNS_Manu).

Results

To characterize distinct types of input connectivity to the VTC, we identified canonical waveforms using the BPC method on voltage CCEP data recorded from electrodes in the collateral sulcus and lateral occipitotemporal sulcus. We calculated a single set of “consensus BPCs” across subjects and identified the anatomic distribution of stimulation sites that were assigned to each BPC. We also investigated the induced spectral power of CCEPs in each consensus BPC to understand the relationship between local neuronal activity and waveform shape. Finally, we applied the BPC algorithm to the induced spectrograms themselves and compared the distribution of “BPSs” to the distribution of voltage BPCs.

BPCs from collateral sulcus input CCEPs in each subject

Collateral sulcus inputs were recorded from 4 subjects (1-5 electrodes each) while stimulating between 30 and 51 electrode pairs (Table 1). Application of the BPC algorithm uncovered sets of 2-4 subject-level BPCs at each measurement electrode, representing distinct canonical CCEP shapes between 11 and 500 ms after stimulation.

BPCs from electrode 1 in Subject 1 are presented in Figure 2. Here, we observed three distinct BPC shapes (Fig. 2B). The first BPC was characterized by a prominent negative deflection at 27 ms after stimulation, followed by a wide positive peak centered at 255 ms; the second BPC was characterized by an initial negative deflection at 31 ms, followed by two positive peaks centered at 81 ms and 330 ms; the third BPC was characterized by a first positive peak at 25 ms and a gradual tapering to baseline over the entire duration.

Figure 2.
Figure 2.

BPCs calculated from CCEP inputs to Subject 1, electrode 1 (collateral sulcus). A, Coronal slice of subject T1-weighted MRI, depicting location of measurement electrode. B, Three distinct BPCs were determined using CCEP data from 36 stimulation sites (11 excluded by algorithm), in the 11-500 ms interval after stimulation. C, Spatial representation of BPCs by stimulation site (squares) on the brain rendering of Subject 1. Colors match BPCs in B; size and color intensity indicate mean SNR across trials at each stimulation site. Stimulation sites are placed at the midpoint of electrode pairs (white circles). Gray represents stimulation sites discarded by BPC algorithm thresholding. D, Mean CCEP across trials from each stimulation site to measurement electrode, separated by BPC category and labeled by anatomic location of stimulation site. Amyg, Amygdala; Collat S, collateral sulcus; Hipp, hippocampus; Ins, insula; MFG, middle frontal gyrus; MTG, middle temporal gyrus; WM, white matter. E, Mid-hippocampal axial and sagittal T1 MRI slices, with electrodes and stimulation sites within 8 mm distance of each slice plotted. Stimulation sites (squares) were colored and scaled by BPC as in C. F, Cortical and white matter stimulation sites within 6 mm of the pial surface viewed on the subject's inflated pial rendering. Lighter areas represent gyri. Darker areas represent sulci. Stimulation sites (squares) were colored and scaled by BPC as in C and E. Results in the absence of common average rereferencing are shown in Extended Data Figure 2-1.

Figure 2-1

BPCs calculated from monopolar CCEP inputs to Subject 1, electrode 1 (collateral sulcus), in the absence of common average re-referencing. All figure panels are arranged in the same format as in Figure 2. Anatomical legends for D: ACC = Anterior Cingulate Cortex, Amyg = Amygdala, Call S = Callosal Sulcus, Collat S = Collateral Sulcus, Hipp = Hippocampus, Ins = Insula, MTG = Middle Temporal Gyrus, SFS = Superior Frontal Sulcus, WM = White Matter. Download Figure 2-1, TIF file.

Across all subjects, between 3 and 25 stimulation sites did not evoke robust responses at the measurement electrodes and did not exceed the threshold set in the algorithm in the calculation of BPC shapes. These stimulation sites were not assigned to any BPCs. Each stimulation site that was assigned to a BPC fit with a degree of robustness quantified by SNR. For example, Figure 2C, E shows robust contributions from hippocampal stimulation for the first BPC and from amygdala stimulation for the second BPC in Subject 1. Major contributions to the third BPC from stimulation in the lateral temporal cortex and insula can be easily seen in Figure 2D,F.

Similar patterns were observed at other electrodes and subjects. The BPCs tended to be more consistent in shape across measurement electrodes within the same subject, although the number of BPCs varied sometimes across measurement electrodes within subject because of the absence of a predetermined cluster number. Across all subjects, SNR was negatively correlated with distance between stimulation site and measurement electrode (Pearson's r(322) = −0.37, p = 1.01 × 10−11).

The overall results were not substantially affected by common average rereferencing. Extended Data Figure 2-1 presents the monopolar BPCs and their anatomic distributions in Subject 1, electrode 1, in the absence of rereferencing.

Four consensus BPCs across subjects

In individual subjects, the collateral sulcus revealed various contrasting BPC shapes from cortical and subcortical gray stimulation. We tested whether a single set of “consensus” BPCs could map these CCEP shapes across subjects. For this, we first applied singular value decomposition (effectively, principal component analysis without centering) to the set of all 38 BPCs merged across collateral sulcus measurement electrodes in Subjects 1-4 (Fig. 3A). The first three principal components collectively explained 89% of total variance across BPCs. Next, K-means clustering was performed on the three principal components to identify four clusters, their centroids projected back to signal space to yield a set of four consensus BPCs. K = 4 clusters were chosen based on visual identification of four salient clusters of BPCs when projected to the first three principal components. However, we note that the dimensionality reduction was unnecessary in this case, as the same four clusters were reliably obtained by K-means clustering when any other number of total principal components were kept.

Figure 3.
Figure 3.

Consensus BPCs and anatomic locations of stimulation sites across 4 subjects. A, Consensus BPCs were calculated by applying K-means clustering on all 38 subject-level BPCs from the 14 collateral sulcus electrodes in Subjects 1-4, after reduction to three principal components by singular value decomposition. B, Temporal shapes of the four consensus BPCs. C, Spatial representation of consensus BPCs by stimulation site from Subjects 1-4, transformed to MNI 152 space, on a standard left hemisphere brain rendering. Stimulation sites with SNR ≥ 1 are colored by consensus BPC category, with size and color intensity scaled by subject-level SNR. Gray represents stimulation sites with SNR < 1, which were discarded by subject-level BPC algorithm thresholding. D, Distribution of stimulation sites across anatomic regions differed significantly between the four consensus BPCs (χ2 test of independence, χ2(15, N = 211) = 235.6, p = 1.15 × 10−41). Amyg, Amygdala; Hipp, hippocampus; LT, lateral temporal; VT, ventral temporal. Cross-validation accuracy of the consensus BPC model is provided in Extended Data Figure 3-1. The effect of distance between stimulation site and measurement electrode on consensus BPC assignments is provided in Extended Data Figure 3-2. For comparison, consensus BPCs alternatively calculated in a single step by applying the BPC algorithm to CCEP trials concatenated across all subjects are shown in Extended Data Figure 3-3.

Figure 3-1

Leave-one-subject-out cross-validation of consensus BPC model. A, Example of cross-validation fold, where all 12 BPCs from Subject 1 were withheld as a test set, while the remaining 26 BPCs from subjects 2 through 4 were used as training data. B, Consensus BPC clusters in principal component space (left) and temporal shapes (right), calculated from the training data in A. Each circle is a single BPC from the training set. C, BPCs from the withheld Subject 1 were projected into the principal component space in B, and assigned the consensus BPC label of the closest cluster centroid calculated from the training data. Here, test set BPCs are plotted as opaque squares and training set BPCs are plotted as semi-transparent circles. Accuracy was calculated as the number of test set BPCs predicted to have the same consensus BPC label as in the full consensus BPC model. D, Consensus BPC shapes and prediction accuracy for the other three cross-validation folds. Download Figure 3-1, TIF file.

Figure 3-2

Relationship between distance from stimulation site to measurement electrode and consensus BPC statistics. A, All stimulation sites plotted by consensus BPC color, with size scaled by SNR, on an MNI 152 pial surface rendering. B, Distributions of MNI 152 Euclidean distance between stimulation site and measurement electrode, grouped and colored by consensus BPC assignment. Each white dot is a single stimulation site. “Exc” = stimulation sites excluded by BPC algorithm. * indicates pairwise significant difference in means, Tukey's Honest Significant Difference Test, p < 0.05. C, MNI 152 Euclidean distance between stimulation site and measurement electrode, plotted against the BPC SNR of the stimulation site. Each dot is a stimulation site, colored by its consensus BPC assignment. Download Figure 3-2, TIF file.

Figure 3-3

Single-step consensus BPCs calculated by applying the BPC algorithm to all CCEP trials concatenated across all collateral sulcus measurement electrodes of Subjects 1-4. A, Temporal shapes of the 4 single-step consensus BPCs. B, Spatial representation of the single-step consensus BPCs by stimulation site, transformed to MNI 152 space, on a standard left hemisphere brain rendering. Stimulation sites are colored by consensus BPC category, with size and color intensity scaled by subject-level SNR. Gray depicts stimulation sites that were not assigned by the BPC algorithm. C, Distribution of stimulation sites across anatomical regions differed significantly between the 4 single-step consensus BPCs (Chi-Square test of independence, Χ2(15, N = 52) = 81.1, p = 4.41*10−11). Amyg = Amygdala, Hipp = Hippocampus, LT = Lateral Temporal, VT = Ventral Temporal). Download Figure 3-3, TIF file.

Figure 3B shows the consensus BPCs that emerged across the subjects, and their waveform shapes are also described in Table 2. Multiple subject-level BPCs with slight differences (e.g., in timing) at one measurement electrode could cluster to the same consensus BPC. This allowed for high level consolidation of possibly redundant subject-level BPCs.

View this table:
Table 2.

Origin and shape of collateral sulcus consensus BPCsa

It is important to note from Table 2 that each consensus BPC was constructed from multiple electrodes across at least 2 subjects; thus, no individual subject drove a particular consensus BPC. Indeed, leave-one-subject-out cross-validation revealed high consensus BPC prediction accuracy for each test subject when the remaining 3 subjects were used to calculate the consensus BPC model (Extended Data Fig. 3-1). If any individual subject were responsible for driving a single consensus BPC, we would have expected to find near-chance level (25%) prediction accuracy when that subject was withheld as the test set. Furthermore, the consensus BPC shapes calculated from each cross-validation training set were highly similar to each other and to the consensus BPC shapes predicted by the full model.

Anatomical distribution of consensus BPCs

Figure 2C demonstrates the anatomic segregation of stimulation sites by BPC cluster in Subject 1. For example, stimulation at hippocampal sites often elicited CCEPs assigned to a subject-level BPC exhibiting a single early negative deflection. We tested whether a consistent anatomic pattern would emerge across multiple subjects and electrodes, when grouping stimulation sites by consensus BPC. To answer this question, we tallied the cortical Destrieux and subcortical atlas labels for all stimulation sites that clustered to each consensus BPC and binned them into six major anatomic regions (hippocampus, amygdala, ventral temporal, lateral temporal, insula, and other) (Fig. 3D). Stimulation sites that were not assigned to any BPCs (123 of 447) or were assigned with SNR <1 (113 of 447, lower BPC fit than residual noise) were excluded from this analysis. The anatomic distributions differed significantly across the four consensus BPCs (χ2 test of independence, χ2(15, N = 211) = 235.6, p = 1.15 × 10−41). As we had observed in single subjects, the majority of stimulation sites that elicited the first consensus BPC were in the hippocampus (27 of 29, 93%), while the majority of stimulation sites that elicited the second consensus BPC were in the amygdala (10 of 14, 71%). The third consensus BPC was elicited mostly by stimulation in lateral temporal cortex (52 of 78, 67%), and the fourth consensus BPC was evoked by stimulation at a broad distribution of sites comprising hippocampus, amygdala, VTC, and insula, but with insula leading in prevalence (39 of 90, 43%). Extended Data Fig. 3-2 examines the effect of distance between stimulation site and measurement electrode on the consensus BPC assignments. Mean distance between stimulation site and measurement electrode did not differ significantly across the four consensus BPCs (Tukey's HSD test between all pairs of consensus BPCs, p > 0.05), but mean distance was greatest for stimulation sites that were excluded by the BPC algorithm (one-way ANOVA, F(4,442) = 49.9, p = 1.21 × 10−34; Tukey's HSD test between each consensus BPC and excluded stimulation sites, p < 0.05). Within each consensus BPC, the SNR was negatively correlated with distance between stimulation site and measurement electrode (Pearson's r ranged between −0.33 and −0.43 for each consensus BPC, p < 0.05).

To visualize the spatial distribution of stimulation sites belonging to each consensus BPC measured in the collateral sulcus, electrodes and stimulation sites from all subjects were transformed to MNI 152 coordinates and plotted to a standard left hemisphere MNI 152 pial rendering (Fig. 3C) and to axial slices of a standard MNI 152 T1-weighted MRI (Fig. 4). On the MNI 152 pial rendering, right hemispheric electrodes were reflected across the midline. It was apparent that, although hippocampal stimulation contributed in part to all four consensus BPCs, the finer spatial distribution differed along the longitudinal axis: notably, the second and fourth consensus BPCs tended to be elicited by stimulation more anteriorly in the amygdala/hippocampus, compared with the first consensus BPC. This pattern occurred in both hemispheres and can be seen in greater detail by visualizing stimulation sites on axial and sagittal slices, at the level of the hippocampus, of each subject's T1 MRI (Extended Data Fig. 4-1). Cortical and white matter stimulation sites within 6 mm of the pial surface were also rendered onto each subject's inflated pial renderings (Extended Data Fig. 4-2).

Figure 4.
Figure 4.

Stimulation sites from all electrodes in Subjects 1-5 visualized on axial slices of MNI 152 T1 MRI. Stimulation sites (squares) were plotted to the nearest slice. Stimulation sites with SNR ≥ 1 are colored by consensus BPC category, with size and color intensity scaled by subject-level SNR. Gray represent stimulation sites with SNR < 1, which were discarded by subject-level BPC algorithm thresholding. Bottom right, z positions of axial slices depicted on sagittal slice.

Figure 4-1

Hippocampal and amygdala stimulation sites for collateral sulcus measurement electrodes. Mid-hippocampal axial and sagittal T1 MRI slices in each subject, with electrodes and stimulation sites within 8 mm distance of each slice visible. Stimulation sites (squares) and subject BPCs were colored according to their assigned consensus BPC. Size and color intensity are scaled by subject-level SNR. Download Figure 4-1, TIF file.

Figure 4-2

Stimulation sites viewed on inflated pial renderings for collateral sulcus measurement electrodes. Cortical and white matter stimulation sites within 6 mm of the pial surface in each subjects were projected to inflated pial renderings. Stimulation sites (squares) and subject BPCs were colored according to their assigned consensus BPC. Size and color intensity are scaled by subject-level SNR. Download Figure 4-2, TIF file.

Direct application of the BPC algorithm to all CCEP trials concatenated across all collateral sulcus measurement electrodes of Subjects 1-4 provided an alternative, single-step approach to calculating consensus BPCs (Extended Data Fig. 3-3). ζ was allowed to increase to 1.17 in the BPC algorithm to yield four single-step consensus BPCs for direct comparison to the main consensus BPC results above. The single-step consensus BPCs possessed similar shapes and anatomic distribution as the main consensus BPCs; the anatomic distributions of stimulation sites with SNR ≥ 1 differed significantly across the four single-step consensus BPCs (χ2 test of independence, χ2(15, N = 52) = 81.1, p = 4.41 × 10−11), and showed the same spatial patterns described for the main consensus BPCs. Fifty-six stimulation sites were not assigned to any single-step consensus BPC, and 51 stimulation sites were assigned weakly with SNR < 1. There were fewer total stimulation sites (N = 159) than present for the main consensus BPCs because trials recorded at different measurement electrodes within the same subject were pooled for each stimulation site in the single-step approach.

Spectral and broadband changes

We observed that stimulation of different anatomic clusters evoked differently shaped waveforms in the collateral sulcus. To better understand whether local neuronal populations increase or decrease in activity in response to these various inputs, we calculated time-frequency heatmaps and time-varying broadband power estimates (Fig. 5). Broadband power, in particular, has been associated with local neuronal activity (Manning et al., 2009; Miller et al., 2009; Ray and Maunsell, 2011).

Figure 5.
Figure 5.

Spectral and broadband power changes of CCEPs across consensus BPC categories. A, Example of spectrogram calculation. CCEPs (middle) were measured at Subject 2, electrode 1 following stimulation at a hippocampus site assigned to consensus BPC 1 (top). Gray represents individual CCEP trials (N = 9). Black represents mean across trials. An average spectrogram for the stimulation site was calculated by geometrically averaging across all wavelet-transformed CCEP trials (bottom, top layer). Average spectrograms were calculated similarly for all other stimulation sites in Subjects 1-4 assigned to consensus BPC 1 (bottom, stacked layers). B, Spectrograms from stimulation sites were combined for each consensus BPC, as well as for all sites excluded by the BPC algorithm, by calculating the one-sample t statistic at each time-frequency bin across stimulation sites. Time-frequency bins with significant power change over baseline (corrected for multiple comparisons at FDR < 0.05) are outlined in green. C, Time-varying broadband power for the same consensus BPCs in B were calculated by averaging across broadband-filtered CCEPs from stimulation sites. Samples with significant power increase and decrease, relative to baseline, (corrected for multiple comparisons at FDR < 0.05) are marked above by red and blue circles, respectively. Inset, Zoom in on the 0-500 ms poststimulation interval. Shaded intervals indicate 95% confidence interval of the mean across stimulation sites.

We determined the average induced spectrograms for each stimulation site by calculating the geometric mean of spectrograms across trials, after normalizing by baseline. The spectrograms and time-varying broadband power estimates from stimulation sites were combined for each consensus BPC, as well as for all stimulation sites algorithmically excluded from BPC assignment (Fig. 5B,C). Immediately after stimulation, we observed a rapid broadband power increase for all consensus BPCs up to ∼100 ms after stimulation. Interestingly, the first and second consensus BPCs demonstrated significant sustained decreases in power >30 Hz, relative to baseline, between 100 and 400 ms after stimulation (FDR < 0.05, corrected for multiple comparisons). This interval of high-frequency power decrease was present to a lesser degree for the fourth consensus BPC, minimally present for the third consensus BPC, and not present in excluded stimulation sites. In addition, the first and second consensus BPCs induced prolonged intervals of significantly decreased high-frequency power from 500 to 1000 ms after stimulation. Significant sustained increases in power in the 12-30 Hz (β) range were also observed for the first and second consensus BPCs between 300 and 800 ms after stimulation. We note that, despite the large differences in evoked potential responses, relatively similar spectral changes emerged across multiple consensus BPCs.

Lateral occipitotemporal sulcus inputs

We briefly explored whether BPCs identical to those in the collateral sulcus would be observed from recording sites in a different ventral temporal area. In our analysis thus far, we have carefully identified electrodes in the collateral sulcus. Similar BPCs in a different VTC area may suggest physiological invariance in the VTC to electrical stimulation inputs. Therefore, we repeated our BPC analysis on inputs to the lateral occipitotemporal sulcus electrodes (Table 1) in Subjects 4 (2 electrodes) and 5 (2 electrodes). Application of the BPC algorithm uncovered sets of 2 or 3 BPCs at each electrode (Fig. 6A,C). As with collateral sulcus inputs, we observed consistent BPC shapes and stimulation site clusters between measurement electrodes within each of the two subjects. Subject 4, who possessed both collateral sulcus electrodes (1 and 2) and lateral occipitotemporal sulcus electrodes (4 and 5), allowed for a direct comparison of input connectivity to the two VTC regions in the same subject. We observed major differences between these two recording regions. Stimulation of amygdala evoked a robust (SNR ≥ 1) biphasic response in the collateral sulcus (Fig. 6B, cyan), but a weak response in the lateral occipitotemporal sulcus (Fig. 6A). Stimulation of hippocampus also evoked a robust biphasic response in the collateral sulcus (Fig. 6B, cyan), but a robust negative-only response in the lateral occipitotemporal sulcus (Fig. 6A, blue). These differences underscore the heterogeneity of waveforms recorded in different regions of the VTC.

Figure 6.
Figure 6.

BPCs calculated from CCEP inputs to lateral occipitotemporal sulcus electrodes in Subjects 4 and 5. A, BPCs calculated from CCEP inputs to Subject 4, electrodes 3 and 4 in the lateral occipitotemporal (OT) sulcus. The following are Destrieux atlas labels for stimulation sites that elicited each BPC with SNR ≥ 1. BPC 1 (blue): hippocampus, parahippocampal gyrus, and white matter; BPC 2 (cyan): lateral ventricle, superior temporal sulcus, and white matter. B, For comparison, BPCs calculated from CCEP inputs to Subject 4, electrode 1 in the collateral sulcus. BPC 1 (blue): parahippocampal gyrus; BPC 2 (cyan): amygdala and hippocampus; BPC 3 (green): insula, putamen, superior temporal sulcus, and white matter. C, BPCs calculated from CCEP inputs to Subject 5, electrodes 1 and 2 in the lateral occipitotemporal sulcus. BPC 1 (blue): hippocampus; BPC 2 (cyan, electrode 1 only): hippocampus; BPC 3 (green): hippocampus, insula, and white matter. Subject 5 did not have any electrodes in the collateral sulcus.

Basis profile spectrograms

Average spectral power changes appeared similar across multiple consensus BPCs, but interesting spectral patterns might have been obfuscated when averaged by consensus BPC category. As a confirmatory analysis, we directly clustered the spectral power changes using the wavelet-transformed CCEP trials (induced spectrograms) as input into the BPC algorithm. For each collateral sulcus measurement electrode, the induced wavelet spectrograms between 12.5 and 200 Hz and between 100 and 500 ms after stimulation were downsampled in time and flattened for all CCEP trials, and then used as input into the BPC algorithm to yield BPSs (Fig. 7). BPSs show the average spectrogram clusters, rather than the average of spectrograms assigned to each voltage-derived consensus BPC (Fig. 5). The anatomic distribution of BPSs was compared with that of BPCs at each measurement electrode (Fig. 7A). This consistently resulted in two BPSs at all measurement electrodes, where either one or both BPSs exhibited a strong decrease in broadband power similar to the spectral pattern that was prominent when grouped by consensus BPCs. The stimulation sites exhibiting the decrease in broadband power tended to localize to the hippocampus and therefore overlapped considerably with the BPC exhibiting an early negative deflection (first consensus BPC).

Figure 7.
Figure 7.

Comparison of subject-level BPCs and BPSs. A, Subject-level BPCs and BPSs shown at one collateral sulcus measurement electrode in each subject. Stimulation sites corresponding to each BPS are plotted on the pial rendering in red or orange, matching the color that outlines the BPSs below. BPS time-frequency bins are plotted in consistent l2-normalized units across subjects. ARI values above pial renderings quantify similarity between BPC and BPS stimulation site assignments. ARI values in parentheses were calculated when keeping only the top 50 percentile (in terms of SNR) of BPC and BPS assignments across all measurement electrodes. ARI ranges between −1 and 1, where positive values indicate above-chance overlap and negative values indicate below-chance overlap. B, ARI calculated at all collateral sulcus measurement electrodes in Subjects 1-4.

At each measurement electrode, we quantified the similarity between the BPC anatomic distribution and the BPS anatomic distribution using the ARI (Fig. 7B). Overall, we found weak but above-chance level (positive) agreement between the two anatomic distributions. Notably, stimulation sites excluded from BPC assignment also tended to be excluded from BPS assignment. To evaluate similarity in the absence of weakly fit BPCs or BPSs, we repeated the ARI calculations after disqualifying BPC and BPS assignments on stimulation sites with SNR values in the bottom 50 percentile (Fig. 7B, orange). A percentile threshold was used, instead of the fixed numerical threshold of 1 used earlier, because the SNR of BPSs was significantly lower than that of BPCs, with nearly all (278 of 286) assigned stimulation sites having BPS SNR values <1. The ARI increased at 9 of the 12 collateral sulcus electrodes after disqualifying the weakly fit stimulation sites.

Discussion

This study characterizes single-pulse electrical stimulation-driven inputs to the collateral sulcus and the lateral occipitotemporal sulcus within the VTC. CCEP shapes varied substantially by stimulation site, and those shapes broadly segregated by the anatomic regions stimulated. The unique CCEP input shapes were summarized at each measurement electrode by sets of 2-4 waveforms (BPCs), and again across electrodes and subjects by a single set of four consensus BPC waveforms. Stimulation produced broadband (>30 Hz) power decreases between 100 and 400 ms after stimulation. This broadband decrease was strongest with stimulation sites that evoked the first and second consensus BPCs, from predominantly hippocampal and amygdala stimulation.

Across subjects, stimulation in the hippocampus elicited a prominent early negative deflection, while stimulation in lateral temporal cortex sites and insula elicited early positive deflections. The advantage in fixing the recording site is that differences in CCEP shape reflect different physiologic connectivity to the same neuronal population. Studies using the current source density method in cat and macaque visual cortices have found that local field potential polarity recorded from microelectrodes largely depends on synaptic configuration (Eccles, 1951; Mitzdorf and Singer, 1978; Schroeder et al., 1998; Keller et al., 2014b). For example, pronounced negative deflections in surface that local field potentials often reflect dipolar current sinks produced by synchronized excitatory postsynaptic potentials at apical dendrites of superficial pyramidal cells (Mitzdorf, 1985). More recent efforts have investigated laminar physiology directly in human disease and behavioral states (Ulbert et al., 2001; Csercsa et al., 2010; Tóth et al., 2021; Ujma et al., 2021). For instance, surface negative macroelectrode potentials in human entorhinal cortex ∼200 ms after word recognition primarily reflected current sinks in layers 2 and 3 (Halgren et al., 2006).

BPC waveforms may thus help to reveal the type of pathway that connects the stimulated and measured neuronal populations. In our analysis, lateral temporal cortex inputs to the collateral sulcus contained a long positive deflection. Previous studies showed that the superior temporal sulcus is involved in processing dynamic facial features, such as gaze (Gobbini and Haxby, 2007; Pitcher and Ungerleider, 2021; Babo-Rebelo et al., 2022), despite no direct connections to the fusiform face area, which neighbors the collateral sulcus (Ethofer et al., 2011; Gschwind et al., 2012; Pyles et al., 2013). Therefore, the ventral visual stream may be indirectly connected to the lateral temporal cortex through the parietal cortex or through the inferior temporal cortex (Babo-Rebelo et al., 2022). These models of indirect connectivity are consistent with the long time course of the lateral temporal input to the collateral sulcus, uncovered by our BPC analysis.

Other useful information potentially contained in CCEP waveforms includes the coherence of arriving inputs (Mitzdorf, 1985) and the degree to which action potentials and neuronal spikes are because of orthodromic and antidromic propagation (Fuortes et al., 1957; Phillips, 1959; Stuart et al., 1997; Kaiser et al., 2001). Large deflections in CCEPs might reflect low-frequency rhythms that have been phase reset and synchronized by stimulation (Nakae et al., 2018). Further analysis of temporal dynamics using BPCs may help clarify the contribution from these various factors.

The results presented here indicate that CCEP responses may be more complex than previously assumed. Many ECoG-CCEP studies have focused on the earliest negative potential in the initial 50 ms, termed the N1, which is often followed by a later negative deflection termed the N2 (Matsumoto et al., 2004, 2007; Keller et al., 2014a; Araki et al., 2015; Krieg et al., 2017; van Blooijs et al., 2018; Kundu et al., 2020; Silverstein et al., 2020). The BPCs we derived from sEEG recordings contained reliable waveforms that differed from this pattern. The absence of N1s and N2s is likely attributable, in part, to electrical field differences measured at depth using sEEG electrodes, compared with measurements at the cortical surface with ECoG. Our sEEG electrodes were positioned close to the gray-white matter boundary. At this depth, the N1/N2 peaks would reverse: that is, surface N1s are the result of depolarization in laminar layers 3 and 4, and signals measured deeper to these layers would result in inverted signals (Mitzdorf and Singer, 1978; Schroeder et al., 1998; Keller et al., 2014b). Indeed, the two positive peaks in the third consensus BPC (Fig. 3B), measured at deeper cortical layers (e.g., see location of electrode in Fig. 2A), might represent inverted N1 and N2 peaks if recorded at the cortical surface instead.

Our data in Subject 4 showed differences in BPC waveforms measured in the collateral sulcus compared with the lateral occipitotemporal sulcus, the two regions separated by the mid-fusiform sulcus (MFS). The VTC is sharply delineated by the MFS into medial and lateral components with differing functional and anatomic features (Grill-Spector and Weiner, 2014). Functionally, cortex medial to the MFS is preferentially activated by visual stimuli that are peripheral, inanimate, large, and which contain places, whereas cortex lateral to the MFS is preferentially activated by stimuli that are foveal, animate, small, and which contain faces (Nasr et al., 2011; Konkle and Oliva, 2012; Weiner et al., 2014). Structurally, the MFS marks a medial-lateral separation in cytoarchitectonics and white matter connectivity (Saygin et al., 2011; Weiner et al., 2014; Gomez et al., 2015). The place-selective region in the collateral sulcus further differs from regions lateral to the MFS in terms of endpoint connectivity profiles (Kubota et al., 2023). Such anatomic boundaries are important to consider when comparing CCEP waveforms across different recording sites.

CCEPs permit directed measurements of connectivity between stimulated and recording sites and are regarded as a form of effective connectivity (Keller et al., 2014b). CCEP-derived connectivity may be complementary to other measures of brain connectivity, and CCEP amplitude networks have been shown to overlap with and extend beyond functional ECoG connectivity networks (Hebbink et al., 2019). Furthermore, local CCEP amplitude connectivity has been shown to better correlate with functional connectivity, while long range CCEP amplitude connectivity better correlates with structural (DTI) connectivity (Crocker et al., 2021). Our BPC method demonstrates that it may be possible to tease apart sets of more complex inputs to the VTC.

The spectral analysis revealed that certain inputs to the collateral sulcus decreased high-frequency broadband power between 100 and 400 ms after stimulation. The inputs producing the largest decreases in power clustered primarily in the hippocampus and amygdala. Broadband power has been correlated with underlying neuronal (input) spiking activity (Manning et al., 2009; Miller et al., 2009; Ray and Maunsell, 2011). As such, our results suggest inhibited neuronal activity following single-pulse electrical stimulation of connected sites. Previous studies have similarly found that single-pulse or low-frequency electrical stimulation in gray matter tended to produce long-lasting neuronal inhibition (Alarcón et al., 2012; Westin et al., 2018; Mohan et al., 2020). Plausible physiologic explanations for this include hyperpolarization mediated by GABAergic interneurons or slow after-hyperpolarization (Toprani and Durand, 2013); indeed, inhibition in monkey visual cortex induced by electrical stimulation in lateral geniculate nucleus could be disrupted with GABA antagonists (Logothetis et al., 2010). Recurrent feedback inhibition by interneurons is key to cortical circuits as a means of gain control (Markram et al., 2004; Douglas and Martin, 2007; Ozeki et al., 2009).

The recordings in our analysis were conducted in patients affected by intractable epilepsy, with variations in age, sex, and epilepsy etiology, and electrodes were placed in regions indicated by the clinical team. The variable placement of electrodes may have contributed to between-subject variability in the BPC shapes calculated. Both the collateral sulcus and lateral occipitotemporal sulcus span several centimeters longitudinally, so there may be different types of inputs to measurement electrodes placed along different parts of each sulcus (Weiner et al., 2018; Kubota et al., 2023). This variability was mitigated by excluding electrodes in the anterior and posterior collateral transverse sulci, which were segmented as their own areas. Disparity in electrode placement also meant that CCEPs may have been affected to varying degrees by volume conduction of stimulation (Prime et al., 2020), which might produce longer-lasting influences beyond the sharp transient artifacts removed. This may have partially contributed to initial deflections observed in some BPCs. More detailed anatomic segmentations may further delineate differences in BPC categorization. For example, waveform differences along the longitudinal axis of the hippocampus may have arisen from different hippocampal subfields (Strange et al., 2014). Despite these potential sources of variability, consensus BPC waveforms correlated relatively well to broad anatomic regions across multiple measurement electrodes and subjects, were robust to cross-validation, and remained similar when alternatively derived in a single step from CCEPs.

In conclusion, single-pulse electrical stimulation inputs to two areas in the VTC could be described by characteristic time-varying waveforms. Some of these shapes were distinct from the archetypal “N1” and “N2” peaks. The shapes segregated by stimulation location in cortical and limbic areas with robustness across subjects and may contain valuable information for mapping synaptic physiology at the measurement electrode. Broadband power decreases poststimulation spanned across stimulation categories and may reflect stimulation-induced inhibition of neuronal activity.

Footnotes

  • This work was supported by National Institute of Mental Health Award R01MH122258, National Institute of General Medical Sciences Award T32GM065841, and American Epilepsy Society Award 937450. The content is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health. We thank the patients for participation in this study; and Cindy Nelson, Karla Crockett, and other staff at Saint Mary Hospital, Mayo Clinic, Rochester, MN for assistance.

  • N.M.G. is an investigator for the Medtronic Deep Brain Stimulation Therapy for Epilepsy Post-Approval Study, and industry consultant (NeuroOne, Inc.; money to Mayo Clinic). B.H.B. declares licensed intellectual property to Cadence Neuroscience Inc. B.N.L. declares intellectual property licensed to Cadence Neuroscience Inc (contractual rights waived), site investigator (Medtronic EPAS, NeuroPace RESPONSE, Neuroelectrics tDCS for Epilepsy), and industry consultant (Epiminder, Medtronic, Neuropace, Philips Neuro; money to Mayo Clinic). G.A.W. declares intellectual property licensed to Cadence Neuroscience Inc. and NeuroOne, Inc. The remaining authors declare no competing financial interests.

  • Correspondence should be addressed to Dora Hermes at hermes.dora{at}mayo.edu

References

  1. ↵
    2. Alarcón G,
    3. Martinez J,
    4. Kerai SV,
    5. Lacruz ME,
    6. Quiroga RQ,
    7. Selway RP,
    8. Richardson MP,
    9. García Seoane JJ,
    10. Valentín A
    (2012) In vivo neuronal firing patterns during human epileptiform discharges replicated by electrical stimulation. Clin Neurophysiol 123:1736–1744. https://doi.org/10.1016/j.clinph.2012.02.062 pmid:22410162
    OpenUrlCrossRefPubMed
  2. ↵
    2. Araki K,
    3. Terada K,
    4. Usui K,
    5. Usui N,
    6. Araki Y,
    7. Baba K,
    8. Matsuda K,
    9. Tottori T,
    10. Inoue Y
    (2015) Bidirectional neural connectivity between basal temporal and posterior language areas in humans. Clin Neurophysiol 126:682–688. https://doi.org/10.1016/j.clinph.2014.07.020 pmid:25190148
    OpenUrlPubMed
  3. ↵
    2. Babo-Rebelo M,
    3. Puce A,
    4. Bullock D,
    5. Hugueville L,
    6. Pestilli F,
    7. Adam C,
    8. Lehongre K,
    9. Lambrecq V,
    10. Dinkelacker V,
    11. George N
    (2022) Visual information routes in the posterior dorsal and ventral face network studied with intracranial neurophysiology and white matter tract endpoints. Cereb Cortex 32:342–366. https://doi.org/10.1093/cercor/bhab212 pmid:34339495
    OpenUrlPubMed
  4. ↵
    2. Benjamini Y,
    3. Yekutieli D
    (2001) The control of the false discovery rate in multiple testing under dependency. Ann Statist 29:1165–1188. https://doi.org/10.1214/aos/1013699998
    OpenUrl
  5. ↵
    2. Bressler SL
    (1995) Large-scale cortical networks and cognition. Brain Res Brain Res Rev 20:288–304. https://doi.org/10.1016/0165-0173(94)00016-i pmid:7550362
    OpenUrlCrossRefPubMed
  6. ↵
    2. Crocker B,
    3. Ostrowski L,
    4. Williams ZM,
    5. Dougherty DD,
    6. Eskandar EN,
    7. Widge AS,
    8. Chu CJ,
    9. Cash SS,
    10. Paulk AC
    (2021) Local and distant responses to single pulse electrical stimulation reflect different forms of connectivity. Neuroimage 237:118094. https://doi.org/10.1016/j.neuroimage.2021.118094 pmid:33940142
    OpenUrlCrossRefPubMed
  7. ↵
    2. Crowther LJ,
    3. Brunner P,
    4. Kapeller C,
    5. Guger C,
    6. Kamada K,
    7. Bunch ME,
    8. Frawley BK,
    9. Lynch TM,
    10. Ritaccio AL,
    11. Schalk G
    (2019) A quantitative method for evaluating cortical responses to electrical stimulation. J Neurosci Methods 311:67–75. https://doi.org/10.1016/j.jneumeth.2018.09.034 pmid:30292823
    OpenUrlCrossRefPubMed
  8. ↵
    2. Csercsa R, et al
    . (2010) Laminar analysis of slow wave activity in humans. Brain 133:2814–2829. https://doi.org/10.1093/brain/awq169 pmid:20656697
    OpenUrlCrossRefPubMed
  9. ↵
    2. Dale AM,
    3. Fischl B,
    4. Sereno MI
    (1999) Cortical surface-based analysis: I. Segmentation and surface reconstruction. Neuroimage 9:179–194. https://doi.org/10.1006/nimg.1998.0395 pmid:9931268
    OpenUrlCrossRefPubMed
  10. ↵
    2. Destrieux C,
    3. Fischl B,
    4. Dale A,
    5. Halgren E
    (2010) Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. Neuroimage 53:1–15. https://doi.org/10.1016/j.neuroimage.2010.06.010 pmid:20547229
    OpenUrlCrossRefPubMed
  11. ↵
    2. Douglas RJ,
    3. Martin KA
    (2007) Mapping the matrix: the ways of neocortex. Neuron 56:226–238. https://doi.org/10.1016/j.neuron.2007.10.017 pmid:17964242
    OpenUrlCrossRefPubMed
  12. ↵
    2. Eccles JC
    (1951) Interpretation of action potentials evoked in the cerebral cortex. Electroencephalogr Clin Neurophysiol 3:449–464. https://doi.org/10.1016/0013-4694(51)90033-8 pmid:14887631
    OpenUrlCrossRefPubMed
  13. ↵
    2. Edelman GM,
    3. Mountcastle VB
    (1982) The mindful brain: Cortical organization and the group-selective theory of higher brain function. MIT press.
  14. ↵
    2. Epstein R,
    3. Kanwisher N
    (1998) A cortical representation of the local visual environment. Nature 392:598–601. https://doi.org/10.1038/33402
    OpenUrlCrossRefPubMed
  15. ↵
    2. Ethofer T,
    3. Gschwind M,
    4. Vuilleumier P
    (2011) Processing social aspects of human gaze: a combined fMRI-DTI study. Neuroimage 55:411–419. https://doi.org/10.1016/j.neuroimage.2010.11.033 pmid:21095230
    OpenUrlCrossRefPubMed
  16. ↵
    2. Felleman DJ,
    3. Van Essen DC
    (1991) Distributed hierarchical processing in the primate cerebral cortex. Cereb Cortex 1:1–47. https://doi.org/10.1093/cercor/1.1.1
    OpenUrlCrossRefPubMed
  17. ↵
    2. Friston KJ
    (1994) Functional and effective connectivity in neuroimaging: a synthesis. Hum Brain Mapp 2:56–78. https://doi.org/10.1002/hbm.460020107
    OpenUrlCrossRef
  18. ↵
    2. Fuortes MG,
    3. Frank K,
    4. Becker MC
    (1957) Steps in the production of motoneuron spikes. J Gen Physiol 40:735–752. https://doi.org/10.1085/jgp.40.5.735 pmid:13428986
    OpenUrlAbstract/FREE Full Text
  19. ↵
    2. Gobbini MI,
    3. Haxby JV
    (2007) Neural systems for recognition of familiar faces. Neuropsychologia 45:32–41. https://doi.org/10.1016/j.neuropsychologia.2006.04.015 pmid:16797608
    OpenUrlCrossRefPubMed
  20. ↵
    2. Goldman-Rakic PS
    (1988) Topography of cognition: parallel distributed networks in primate association cortex. Annu Rev Neurosci 11:137–156. https://doi.org/10.1146/annurev.ne.11.030188.001033 pmid:3284439
    OpenUrlCrossRefPubMed
  21. ↵
    2. Gomez J,
    3. Pestilli F,
    4. Witthoft N,
    5. Golarai G,
    6. Liberman A,
    7. Poltoratski S,
    8. Yoon J,
    9. Grill-Spector K
    (2015) Functionally defined white matter reveals segregated pathways in human ventral temporal cortex associated with category-specific processing. Neuron 85:216–227. https://doi.org/10.1016/j.neuron.2014.12.027 pmid:25569351
    OpenUrlCrossRefPubMed
  22. ↵
    2. Grill-Spector K,
    3. Weiner KS
    (2014) The functional architecture of the ventral temporal cortex and its role in categorization. Nat Rev Neurosci 15:536–548. https://doi.org/10.1038/nrn3747 pmid:24962370
    OpenUrlCrossRefPubMed
  23. ↵
    2. Gschwind M,
    3. Pourtois G,
    4. Schwartz S,
    5. Van De Ville D,
    6. Vuilleumier P
    (2012) White-matter connectivity between face-responsive regions in the human brain. Cereb Cortex 22:1564–1576. https://doi.org/10.1093/cercor/bhr226 pmid:21893680
    OpenUrlCrossRefPubMed
  24. ↵
    2. Halgren E,
    3. Wang C,
    4. Schomer DL,
    5. Knake S,
    6. Marinkovic K,
    7. Wu J,
    8. Ulbert I
    (2006) Processing stages underlying word recognition in the anteroventral temporal lobe. Neuroimage 30:1401–1413. https://doi.org/10.1016/j.neuroimage.2005.10.053 pmid:16488158
    OpenUrlCrossRefPubMed
  25. ↵
    2. Hebbink J,
    3. van Blooijs D,
    4. Huiskamp G,
    5. Leijten FS,
    6. van Gils SA,
    7. Meijer HG
    (2019) A comparison of evoked and non-evoked functional networks. Brain Topogr 32:405–417. https://doi.org/10.1007/s10548-018-0692-1 pmid:30523480
    OpenUrlPubMed
  26. ↵
    2. Hermes D,
    3. Miller KJ,
    4. Noordmans HJ,
    5. Vansteensel MJ,
    6. Ramsey NF
    (2010) Automated electrocorticographic electrode localization on individually rendered brain surfaces. J Neurosci Methods 185:293–298. https://doi.org/10.1016/j.jneumeth.2009.10.005 pmid:19836416
    OpenUrlCrossRefPubMed
  27. ↵
    2. Holdgraf C, et al
    . (2019) iEEG-BIDS, extending the Brain Imaging Data Structure specification to human intracranial electrophysiology. Sci Data 6:1. https://doi.org/10.1038/s41597-019-0105-7
    OpenUrl
  28. ↵
    2. Huang H,
    3. Valencia GO,
    4. Hermes D,
    5. Miller KJ
    (2021) A canonical visualization tool for sEEG electrodes. 2021 43rd Annual International Conference of the IEEE Engineering in Medicine Biology Society, pp 6175–6178. https://doi.org/10.1109/EMBC46164.2021.9630724 pmid:34892526
  29. ↵
    2. Hubert LJ,
    3. Arabie P
    (1985) Comparing partitions. J Classification 2:193–218. https://doi.org/10.1007/BF01908075
    OpenUrlCrossRef
  30. ↵
    2. Kaiser KM,
    3. Zilberter Y,
    4. Sakmann B
    (2001) Back-propagating action potentials mediate calcium signalling in dendrites of bitufted interneurons in layer 2/3 of rat somatosensory cortex. J Physiol 535:17–31. https://doi.org/10.1111/j.1469-7793.2001.t01-1-00017.x pmid:11507155
    OpenUrlCrossRefPubMed
  31. ↵
    2. Kanwisher N,
    3. McDermott J,
    4. Chun MM
    (1997) The fusiform face area: a module in human extrastriate cortex specialized for face perception. J Neurosci 17:4302–4311. https://doi.org/10.1523/JNEUROSCI.17-11-04302.1997
    OpenUrlAbstract/FREE Full Text
  32. ↵
    2. Kay KN,
    3. Yeatman JD
    (2017) Bottom-up and top-down computations in word- and face-selective cortex. Elife 6:e22341. https://doi.org/10.7554/eLife.22341
    OpenUrlCrossRef
  33. ↵
    2. Keller CJ,
    3. Honey CJ,
    4. Entz L,
    5. Bickel S,
    6. Groppe DM,
    7. Toth E,
    8. Ulbert I,
    9. Lado FA,
    10. Mehta AD
    (2014a) Corticocortical evoked potentials reveal projectors and integrators in human brain networks. J Neurosci 34:9152–9163. https://doi.org/10.1523/JNEUROSCI.4289-13.2014 pmid:24990935
    OpenUrlAbstract/FREE Full Text
  34. ↵
    2. Keller CJ,
    3. Honey CJ,
    4. Mégevand P,
    5. Entz L,
    6. Ulbert I,
    7. Mehta AD
    (2014b) Mapping human brain networks with cortico-cortical evoked potentials. Philos Trans R Soc Lond B Biol Sci 369:20130528. https://doi.org/10.1098/rstb.2013.0528
    OpenUrlCrossRefPubMed
  35. ↵
    2. Konkle T,
    3. Oliva A
    (2012) A real-world size organization of object responses in occipitotemporal cortex. Neuron 74:1114–1124. https://doi.org/10.1016/j.neuron.2012.04.036 pmid:22726840
    OpenUrlCrossRefPubMed
  36. ↵
    2. Krieg J,
    3. Koessler L,
    4. Jonas J,
    5. Colnat-Coulbois S,
    6. Vignal J P,
    7. Bénar CG,
    8. Maillard LG
    (2017) Discrimination of a medial functional module within the temporal lobe using an effective connectivity model: A CCEP study. NeuroImage 16:219–231.
    OpenUrl
  37. ↵
    2. Kubota E,
    3. Grotheer M,
    4. Finzi D,
    5. Natu VS,
    6. Gomez J,
    7. Grill-Spector K
    (2023) White matter connections of high-level visual areas predict cytoarchitecture better than category-selectivity in childhood, but not adulthood. Cereb Cortex 33:2485–2506. https://doi.org/10.1093/cercor/bhac221
    OpenUrl
  38. ↵
    2. Kundu B,
    3. Davis TS,
    4. Philip B,
    5. Smith EH,
    6. Arain A,
    7. Peters A,
    8. Newman B,
    9. Butson CR,
    10. Rolston JD
    (2020) A systematic exploration of parameters affecting evoked intracranial potentials in patients with epilepsy. Brain Stimul 13:1232–1244. https://doi.org/10.1016/j.brs.2020.06.002 pmid:32504827
    OpenUrlPubMed
  39. ↵
    2. Lacruz ME,
    3. García Seoane JJ,
    4. Valentin A,
    5. Selway R,
    6. Alarcón G
    (2007) Frontal and temporal functional connections of the living human brain. Eur J Neurosci 26:1357–1370. https://doi.org/10.1111/j.1460-9568.2007.05730.x pmid:17767512
    OpenUrlCrossRefPubMed
  40. ↵
    2. Lee DD,
    3. Seung HS
    (1999) Learning the parts of objects by non-negative matrix factorization. Nature 401:788–791. https://doi.org/10.1038/44565 pmid:10548103
    OpenUrlCrossRefPubMed
  41. ↵
    2. Logothetis NK,
    3. Augath M,
    4. Murayama Y,
    5. Rauch A,
    6. Sultan F,
    7. Goense J,
    8. Oeltermann A,
    9. Merkle H
    (2010) The effects of electrical microstimulation on cortical signal propagation. Nat Neurosci 13:1283–1291. https://doi.org/10.1038/nn.2631 pmid:20818384
    OpenUrlCrossRefPubMed
  42. ↵
    2. Manning JR,
    3. Jacobs J,
    4. Fried I,
    5. Kahana MJ
    (2009) Broadband shifts in local field potential power spectra are correlated with single-neuron spiking in humans. J Neurosci 29:13613–13620. https://doi.org/10.1523/JNEUROSCI.2041-09.2009 pmid:19864573
    OpenUrlAbstract/FREE Full Text
  43. ↵
    2. Markov NT,
    3. Vezoli J,
    4. Chameau P,
    5. Falchier A,
    6. Quilodran R,
    7. Huissoud C,
    8. Lamy C,
    9. Misery P,
    10. Giroud P,
    11. Ullman S,
    12. Barone P,
    13. Dehay C,
    14. Knoblauch K,
    15. Kennedy H
    (2014) Anatomy of hierarchy: feedforward and feedback pathways in macaque visual cortex. J Comp Neurol 522:225–259. https://doi.org/10.1002/cne.23458 pmid:23983048
    OpenUrlCrossRefPubMed
  44. ↵
    2. Markram H,
    3. Toledo-Rodriguez M,
    4. Wang Y,
    5. Gupta A,
    6. Silberberg G,
    7. Wu C
    (2004) Interneurons of the neocortical inhibitory system. Nat Rev Neurosci 5:793–807. https://doi.org/10.1038/nrn1519
    OpenUrlCrossRefPubMed
  45. ↵
    2. Matsumoto R,
    3. Nair DR,
    4. LaPresto E,
    5. Najm I,
    6. Bingaman W,
    7. Shibasaki H,
    8. Lüders HO
    (2004) Functional connectivity in the human language system: a cortico-cortical evoked potential study. Brain 127:2316–2330. https://doi.org/10.1093/brain/awh246 pmid:15269116
    OpenUrlCrossRefPubMed
  46. ↵
    2. Matsumoto R,
    3. Nair DR,
    4. LaPresto E,
    5. Bingaman W,
    6. Shibasaki H,
    7. Lüders HO
    (2007) Functional connectivity in human cortical motor system: a cortico-cortical evoked potential study. Brain 130:181–197. https://doi.org/10.1093/brain/awl257 pmid:17046857
    OpenUrlCrossRefPubMed
  47. ↵
    2. McManus JN,
    3. Li W,
    4. Gilbert CD
    (2011) Adaptive shape processing in primary visual cortex. Proc Natl Acad Sci USA 108:9739–9746. https://doi.org/10.1073/pnas.1105855108 pmid:21571645
    OpenUrlAbstract/FREE Full Text
  48. ↵
    2. Mercier MR, et al
    . (2022) Advances in human intracranial electroencephalography research, guidelines and good practices. Neuroimage 260:119438. https://doi.org/10.1016/j.neuroimage.2022.119438 pmid:35792291
    OpenUrlPubMed
  49. ↵
    2. Mesulam MM
    (1990) Large-scale neurocognitive networks and distributed processing for attention, language, and memory. Ann Neurol 28:597–613. https://doi.org/10.1002/ana.410280502 pmid:2260847
    OpenUrlCrossRefPubMed
  50. ↵
    2. Mewett DT,
    3. Nazeran H,
    4. Reynolds KJ
    (2001). Removing power line noise from recorded EMG. 2001 Conference Proceedings of the 23rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp 2190–2193. https://doi.org/10.1109/IEMBS.2001.1017205
  51. ↵
    2. Miller KJ,
    3. Sorensen LB,
    4. Ojemann JG,
    5. den Nijs M
    (2009) Power-law scaling in the brain surface electric potential. PLoS Comput Biol 5:e1000609. https://doi.org/10.1371/journal.pcbi.1000609 pmid:20019800
    OpenUrlCrossRefPubMed
  52. ↵
    2. Miller KJ,
    3. Müller KR,
    4. Hermes D
    (2021) Basis profile curve identification to understand electrical stimulation effects in human brain networks. PLoS Comput Biol 17:e1008710. https://doi.org/10.1371/journal.pcbi.1008710 pmid:34473701
    OpenUrlCrossRefPubMed
  53. ↵
    2. Mitzdorf U
    (1985) Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena. Physiol Rev 65:37–100. https://doi.org/10.1152/physrev.1985.65.1.37 pmid:3880898
    OpenUrlCrossRefPubMed
  54. ↵
    2. Mitzdorf U,
    3. Singer W
    (1978) Prominent excitatory pathways in the cat visual cortex (A 17 and A 18): a current source density analysis of electrically evoked potentials. Exp Brain Res 33:371–394. https://doi.org/10.1007/BF00235560 pmid:215431
    OpenUrlCrossRefPubMed
  55. ↵
    2. Mohan UR,
    3. Watrous AJ,
    4. Miller JF,
    5. Lega BC,
    6. Sperling MR,
    7. Worrell GA,
    8. Gross RE,
    9. Zaghloul KA,
    10. Jobst BC,
    11. Davis KA,
    12. Sheth SA,
    13. Stein JM,
    14. Das SR,
    15. Gorniak R,
    16. Wanda PA,
    17. Rizzuto DS,
    18. Kahana MJ,
    19. Jacobs J
    (2020) The effects of direct brain stimulation in humans depend on frequency, amplitude, and white-matter proximity. Brain Stimul 13:1183–1195. https://doi.org/10.1016/j.brs.2020.05.009 pmid:32446925
    OpenUrlPubMed
  56. ↵
    2. Moran J,
    3. Desimone R
    (1985) Selective attention gates visual processing in the extrastriate cortex. Science 229:782–784. https://doi.org/10.1126/science.4023713 pmid:4023713
    OpenUrlAbstract/FREE Full Text
  57. ↵
    2. Motter BC
    (1993) Focal attention produces spatially selective processing in visual cortical areas V1, V2, and V4 in the presence of competing stimuli. J Neurophysiol 70:909–919. https://doi.org/10.1152/jn.1993.70.3.909 pmid:8229178
    OpenUrlCrossRefPubMed
  58. ↵
    2. Nakae T,
    3. Matsumoto R,
    4. Togo M,
    5. Takeyama H,
    6. Kobayashi K,
    7. Shimotake A,
    8. Matsuhashi M,
    9. Yamao Y,
    10. Kikuchi T,
    11. Yoshida K,
    12. Kunieda T,
    13. Ikeda A,
    14. Miyamoto S
    (2018) Oscillatory responses evoked by single-pulse electrical stimulation in human cerebral cortex: a Cortico-Cortical Evoked Potential (CCEP) study. Clin Neurophysiol 129:e189–e190. https://doi.org/10.1016/j.clinph.2018.04.488
    OpenUrl
  59. ↵
    2. Nasr S,
    3. Liu N,
    4. Devaney KJ,
    5. Yue X,
    6. Rajimehr R,
    7. Ungerleider LG,
    8. Tootell RB
    (2011) Scene-selective cortical regions in human and nonhuman primates. J Neurosci 31:13771–13785. https://doi.org/10.1523/JNEUROSCI.2792-11.2011 pmid:21957240
    OpenUrlAbstract/FREE Full Text
  60. ↵
    2. Ozeki H,
    3. Finn IM,
    4. Schaffer ES,
    5. Miller KD,
    6. Ferster D
    (2009) Inhibitory stabilization of the cortical network underlies visual surround suppression. Neuron 62:578–592. https://doi.org/10.1016/j.neuron.2009.03.028 pmid:19477158
    OpenUrlCrossRefPubMed
  61. ↵
    2. Penny WD,
    3. Friston KJ,
    4. Ashburner JT,
    5. Kiebel SJ,
    6. Nichols TE
    (2011) Statistical parametric mapping: the analysis of functional brain images. Amsterdam: Elsevier.
  62. ↵
    2. Phillips CG
    (1959) Actions of antidromic pyramidal volleys on single Betz cells in the cat. Q J Exp Physiol Cogn Med Sci 44:1–25. https://doi.org/10.1113/expphysiol.1959.sp001364 pmid:13624009
    OpenUrlPubMed
  63. ↵
    2. Pitcher D,
    3. Ungerleider LG
    (2021) Evidence for a third visual pathway specialized for social perception. Trends Cogn Sci 25:100–110. https://doi.org/10.1016/j.tics.2020.11.006 pmid:33334693
    OpenUrlCrossRefPubMed
  64. ↵
    2. Prime D,
    3. Woolfe M,
    4. O'Keefe S,
    5. Rowlands D,
    6. Dionisio S
    (2020) Quantifying volume conducted potential using stimulation artefact in cortico-cortical evoked potentials. J Neurosci Methods 337:108639. https://doi.org/10.1016/j.jneumeth.2020.108639 pmid:32156547
    OpenUrlCrossRefPubMed
  65. ↵
    2. Pyles JA,
    3. Verstynen TD,
    4. Schneider W,
    5. Tarr MJ
    (2013) Explicating the face perception network with white matter connectivity. PLoS One 8:e61611. https://doi.org/10.1371/journal.pone.0061611 pmid:23630602
    OpenUrlCrossRefPubMed
  66. ↵
    2. Ray S,
    3. Maunsell JH
    (2011) Different origins of gamma rhythm and high-gamma activity in macaque visual cortex. PLoS Biol 9:e1000610. https://doi.org/10.1371/journal.pbio.1000610 pmid:21532743
    OpenUrlCrossRefPubMed
  67. ↵
    2. Saygin ZM,
    3. Osher DE,
    4. Koldewyn K,
    5. Reynolds G,
    6. Gabrieli JD,
    7. Saxe RR
    (2011) Anatomical connectivity patterns predict face selectivity in the fusiform gyrus. Nat Neurosci 15:321–327. https://doi.org/10.1038/nn.3001 pmid:22197830
    OpenUrlCrossRefPubMed
  68. ↵
    2. Schlack A,
    3. Albright TD
    (2007) Remembering visual motion: neural correlates of associative plasticity and motion recall in cortical area MT. Neuron 53:881–890. https://doi.org/10.1016/j.neuron.2007.02.028 pmid:17359922
    OpenUrlCrossRefPubMed
  69. ↵
    2. Schroeder C,
    3. Mehta AD,
    4. Givre SJ
    (1998) A spatiotemporal profile of visual system activation revealed by current source density analysis in the awake macaque. Cereb Cortex 8:575–592. https://doi.org/10.1093/cercor/8.7.575 pmid:9823479
    OpenUrlCrossRefPubMed
  70. ↵
    2. Shine JM,
    3. Kucyi A,
    4. Foster BL,
    5. Bickel S,
    6. Wang D,
    7. Liu H,
    8. Poldrack RA,
    9. Hsieh LT,
    10. Hsiang JC,
    11. Parvizi J
    (2017) Distinct patterns of temporal and directional connectivity among intrinsic networks in the human brain. J Neurosci 37:9667–9674. https://doi.org/10.1523/JNEUROSCI.1574-17.2017 pmid:28893929
    OpenUrlAbstract/FREE Full Text
  71. ↵
    2. Silverstein BH,
    3. Asano E,
    4. Sugiura A,
    5. Sonoda M,
    6. Lee MH,
    7. Jeong JW
    (2020) Dynamic tractography: integrating cortico-cortical evoked potentials and diffusion imaging. Neuroimage 215:116763. https://doi.org/10.1016/j.neuroimage.2020.116763 pmid:32294537
    OpenUrlPubMed
  72. ↵
    2. Sporns O,
    3. Chialvo DR,
    4. Kaiser M,
    5. Hilgetag CC
    (2004) Organization, development and function of complex brain networks. Trends Cogn Sci 8:418–425. https://doi.org/10.1016/j.tics.2004.07.008 pmid:15350243
    OpenUrlCrossRefPubMed
  73. ↵
    2. Steinley D
    (2004) Properties of the Hubert-Arable Adjusted Rand Index. Psychol Methods 9:386–396. https://doi.org/10.1037/1082-989X.9.3.386 pmid:15355155
    OpenUrlCrossRefPubMed
  74. ↵
    2. Strange BA,
    3. Witter MP,
    4. Lein ES,
    5. Moser EI
    (2014) Functional organization of the hippocampal longitudinal axis. Nat Rev Neurosci 15:655–669. https://doi.org/10.1038/nrn3785
    OpenUrlCrossRefPubMed
  75. ↵
    2. Stuart G,
    3. Schiller J,
    4. Sakmann B
    (1997) Action potential initiation and propagation in rat neocortical pyramidal neurons. J Physiol 505:617–632. https://doi.org/10.1111/j.1469-7793.1997.617ba.x
    OpenUrlCrossRefPubMed
  76. ↵
    2. Toprani S,
    3. Durand DM
    (2013) Long-lasting hyperpolarization underlies seizure reduction by low frequency deep brain electrical stimulation. J Physiol 591:5765–5790. https://doi.org/10.1113/jphysiol.2013.253757 pmid:23981713
    OpenUrlCrossRefPubMed
  77. ↵
    2. Tóth E,
    3. Bokodi V,
    4. Somogyvári Z,
    5. Maglóczky Z,
    6. Wittner L,
    7. Ulbert I,
    8. Erőss L,
    9. Fabó D
    (2021) Laminar distribution of electrically evoked hippocampal short latency ripple activity highlights the importance of the subiculum in vivo in human epilepsy, an intraoperative study. Epilepsy Res 169:106509. https://doi.org/10.1016/j.eplepsyres.2020.106509 pmid:33310654
    OpenUrlPubMed
  78. ↵
    2. Ujma PP,
    3. Hajnal B,
    4. Bódizs R,
    5. Gombos F,
    6. Erőss L,
    7. Wittner L,
    8. Halgren E,
    9. Cash SS,
    10. Ulbert I,
    11. Fabó D
    (2021) The laminar profile of sleep spindles in humans. Neuroimage 226:117587. https://doi.org/10.1016/j.neuroimage.2020.117587 pmid:33249216
    OpenUrlPubMed
  79. ↵
    2. Ulbert I,
    3. Halgren E,
    4. Heit G,
    5. Karmos G
    (2001) Multiple microelectrode-recording system for human intracortical applications. J Neurosci Methods 106:69–79. https://doi.org/10.1016/s0165-0270(01)00330-2 pmid:11248342
    OpenUrlCrossRefPubMed
  80. ↵
    2. van Blooijs D,
    3. Leijten FS,
    4. van Rijen PC,
    5. Meijer HG,
    6. Huiskamp GJ
    (2018) Evoked directional network characteristics of epileptogenic tissue derived from single pulse electrical stimulation. Hum Brain Mapp 39:4611–4622. https://doi.org/10.1002/hbm.24309 pmid:30030947
    OpenUrlCrossRefPubMed
  81. ↵
    2. van Kerkoerle T,
    3. Self MW,
    4. Dagnino B,
    5. Gariel-Mathis MA,
    6. Poort J,
    7. van der Togt C,
    8. Roelfsema PR
    (2014) Alpha and gamma oscillations characterize feedback and feedforward processing in monkey visual cortex. Proc Natl Acad Sci USA 111:14332–14341. https://doi.org/10.1073/pnas.1402773111 pmid:25205811
    OpenUrlAbstract/FREE Full Text
  82. ↵
    2. Veit MJ,
    3. Kucyi A,
    4. Hu W,
    5. Zhang C,
    6. Zhao B,
    7. Guo Z,
    8. Yang B,
    9. Sava-Segal C,
    10. Perry C,
    11. Zhang J,
    12. Zhang K,
    13. Parvizi J
    (2021) Temporal order of signal propagation within and across intrinsic brain networks. Proc Natl Acad Sci USA 118:e2105031118. https://doi.org/10.1073/pnas.2105031118
    OpenUrlAbstract/FREE Full Text
  83. ↵
    2. Weiner KS,
    3. Golarai G,
    4. Caspers J,
    5. Chuapoco MR,
    6. Mohlberg H,
    7. Zilles K,
    8. Amunts K,
    9. Grill-Spector K
    (2014) The mid-fusiform sulcus: a landmark identifying both cytoarchitectonic and functional divisions of human ventral temporal cortex. Neuroimage 84:453–465. https://doi.org/10.1016/j.neuroimage.2013.08.068 pmid:24021838
    OpenUrlCrossRefPubMed
  84. ↵
    2. Weiner KS,
    3. Barnett MA,
    4. Witthoft N,
    5. Golarai G,
    6. Stigliani A,
    7. Kay KN,
    8. Gomez J,
    9. Natu VS,
    10. Amunts K,
    11. Zilles K,
    12. Grill-Spector K
    (2018) Defining the most probable location of the parahippocampal place area using cortex-based alignment and cross-validation. Neuroimage 170:373–384. https://doi.org/10.1016/j.neuroimage.2017.04.040 pmid:28435097
    OpenUrlPubMed
  85. ↵
    2. Westin K,
    3. Lundstrom B,
    4. Cooray G
    (2018) Neurophysiological effects of chronic subthreshold cortical stimulation for focal epilepsy: spike and spontaneous ECoG activity. Clin Neurophysiol 129:e67–e68. https://doi.org/10.1016/j.clinph.2018.04.168
    OpenUrl
  86. ↵
    2. Winawer J,
    3. Kay KN,
    4. Foster BL,
    5. Rauschecker AM,
    6. Parvizi J,
    7. Wandell BA
    (2013) Asynchronous broadband signals are the principal source of the BOLD response in human visual cortex. Curr Biol 23:1145–1153. https://doi.org/10.1016/j.cub.2013.05.001 pmid:23770184
    OpenUrlCrossRefPubMed
  87. ↵
    2. Yang Y
    (2017) Temporal data clustering. In: Temporal data mining via unsupervised ensemble learning ( Yang Y, ed), pp 19–34. Amsterdam: Elsevier. https://doi.org/10.1016/B978-0-12-811654-8.00003-8
Back to top

In this issue

Email
Print
View Full Page PDF
Citation Tools
Respond to this article
Electrical Stimulation of Temporal and Limbic Circuitry Produces Distinct Responses in Human Ventral Temporal Cortex
Harvey Huang, Nicholas M. Gregg, Gabriela Ojeda Valencia, Benjamin H. Brinkmann, Brian N. Lundstrom, Gregory A. Worrell, Kai J. Miller, Dora Hermes
Journal of Neuroscience 14 June 2023, 43 (24) 4434-4447; DOI: 10.1523/JNEUROSCI.1325-22.2023
Twitter logo Facebook logo Mendeley logo

Jump to section

Keywords

Responses to this article

Respond to this article

Jump to comment:

No eLetters have been published for this article.