A hemispheric decoding principle for vestibular heading perception in the posterior sylvian area

Significance

The vestibular system plays a crucial role in important fundamental as well as high cognitive functions. Supporting this, vestibular signals have been found widely distributed in the brain including the cerebral neocortex, yet they are mostly studied under stimulus-neural activity correlation, leaving it unverified about their exact functional implications. Here by combining behavioral task, neural recording, and perturbation with microstimulation in the visual posterior sylvian (VPS), a unique area in the brain for being dominated by visual-vestibular conflict neurons, we verify causal contributions of vestibular signals to heading perception. Strikingly, the vestibular, but not visual signals in VPS are decoded primarily with respect to hemisphere (ipsiversive), rather than the labeled-line code typically seen in other sensory systems.

Abstract

Identifying functional links between neural activity and behavioral perception is a fundamental neuroscience issue for inferring causality and means of information-decoding. Using perturbation methods, numerous studies have provided direct evidence showing that sensory information in many systems (e.g., vision) is causally read out for subjects’ perceptual choice with respect to the manipulated neurons’ preferred feature (labeled-line code). Here we report an alternative decoding principle in the visual posterior sylvian area (VPS). Specifically, VPS neurons in each hemisphere encode heading information symmetrically in terms of leftward vs. rightward preference in the earth-horizontal plane based on inertial or visual motion cues. In a heading discrimination task, correlation between neural activity and perceptual choice on a trial-by-trial basis (choice probability) was also significantly dependent on the neurons’ tuning functions. However, electrical microstimulation in VPS significantly biased macaques’ heading judgments toward the ipsilateral hemisphere, irrespective of the artificially-stimulated neurons’ encoded heading preference. This effect was observed exclusively in the vestibular but not visual condition, and in VPS but not neighboring areas. The microstimulation induced perceptual bias could be complemented by chemical inactivation manipulations. A noninvasive galvanic vestibular stimulation applied at the peripheral inner ear organs produced similar bias in the subjects’ behavioral performance, as well as activation in the central VPS. Our findings reveal a hemispheric-readout algorithm that predominates over the labeled-line code in the central vestibular system, which may constrain the way of information integration across sensory modalities.

Precise self-motion perception is essential for animals to survive in natural environment, a process highly dependent on multiple sensory inputs. Among them, the vestibular system is crucial for mediating perception of self-motion and action, for example, when computing and planning path for vector-based navigation (1). Supporting this, robust vestibular signals have been found to exist broadly in many cortical areas including parieto-insular vestibular cortex (PIVC), 2v, 3a, visual posterior sylvian area (VPS), dorsal portion of medial superior temporal area (MSTd), ventral intraparietal area (VIP), smooth eye movement area of the frontal eye field (FEFsem), posterior cingulate cortex (PCC), retrosplenial cortex, area 7a, superior temporal polysensory area (2, 3). Interestingly, many of these areas also contain visual signals related to self-motion (e.g., optic flow), indicating possible cross-modality interactions (1). Indeed, many neurons in these areas encode congruent visual and vestibular heading directions such that their tuning functions are even strengthened during self-motion when both cues are available, which may mediate the improved heading discriminability when compared to that under unimodal conditions (1).

Although vestibular modulations to self-motion are prevalent in the brain including neocortex, most of the studies only focus on neural encoding, upon which potential ecological functions are implied. Instead, whether and how these sensory signals are exactly read out by downstream neurons for behavioral perception is elusive. So far there is only one study that has performed electrical microstimulation in MSTd, but it fails to evoke significant influence on the animals’ heading judgment based on vestibular cues (4). The negative result observed in Gu’s study may be due to two factors. First, unlike the well-clustered visual signals, MSTd lacks clustered vestibular tuning which may have limited efficiency of microstimulation technique. Second, vestibular signals in MSTd may not be deployed for the function of heading judgment. In any case, compared to successful application of manipulation methods including microstimulation (5, 6) and optogenetic (7, 8) tools in other sensory systems, it is urgently needed to reveal causal readout mechanisms in the vestibular system.

We here therefore examined causal readout effects of vestibular signals in the VPS. We focused on this area mainly due to its unique feature of visual-vestibular interactions. That is, unlike the other areas that contain mixed visual-vestibular congruent and conflict cells [MSTd: (9, 10); VIP: (11); FEFsem: (12)], VPS is predominated by conflict neurons (13). Based on this fact, the previous study inferred that this area is unlikely to be involved in heading perception because the two types of self-motion signals would give conflicting information regarding ongoing heading direction (13). We thus have two goals here. First, we want to identify whether vestibular and optic flow signals in VPS are causally read out for heading perception in each individual modality condition. Second, if they do, are they decoded with respect to the preferred heading in each stimulus condition, or in a modality independent way (e.g., by both referenced to vestibular, or visual, or their intermediate)? To address these questions, in the current study we recorded neural activity in VPS. We found well clustered vestibular signals, providing a basis for applying microstimulation perturbation. Indeed, delivering small electrical currents significantly biased the animals’ perceptual judgments. However surprisingly, the bias direction was not predicted from the neurons’ tuning functions, but rather from the hemisphere. Subsequent control experiments further showed that this hemispheric-dependent, causal readout effect was specific to the vestibular signals in VPS, but not visual modality or other brain regions. Possible mechanisms mediating this distinct readout effect of the vestibular signals from other sensory systems (e.g., visual) were discussed.

Results

Behavioral Tasks.

Three male rhesus monkeys were trained to perform two versions of behavioral tasks. In the fixation-only task, the animals were only required to maintain fixation at a central target for 1,500 ms while passively experiencing two types of heading stimuli: inertial force applied through translation of a motion base (vestibular heading), and optic flow simulating self-moved direction (visual heading) (Fig. 1A). In the heading-discrimination task, however, the animals were required to actively report their experienced heading direction, whether leftward or rightward relative to straight ahead at the end of each trial, by making a saccadic eye movement to one of the two choice targets at each side of the screen. Correct choice leads to reward of water or juice. After training, all animals showed stable performance with heading discriminability of a few degrees (Fig. 2B, solid lines).

Fig. 1.

Encoded heading tuning properties in VPS neurons. (A) (Left) Schematic of the virtual reality system. (Right) Timeline of heading discrimination task based on vestibular or visual heading stimuli. (B) (Top) Anatomical location of VPS. (Bottom) MRI reconstruction of microstimulation sites in VPS of three monkeys. Blue: vestibular tuning only; Red: visual tuning only; Green: multisensory tuning; White: no significant spatial tuning. White dashed contour: VPS boundary, predominantly localized at the posterior lateral sulcus. (C) (Top Left) An example of a vestibular (blue) and visual (red) tuning functions to heading direction in the horizontal plane for a visuo-vestibular conflict cell. (Bottom Left) Distribution of the difference in preferred direction between visual and vestibular tuning of VPS neurons. (Right) Proportion of different types of VPS neurons. (D) Distributions of vestibular (Upper) and visual (Lower) heading preference of VPS neurons in each hemisphere. In each plot, color represents data from each of the three animals. (E) Distribution of Choice probability (CP) of VPS neurons. (Upper) signs are assigned according to the neuron’s preferred heading direction. (Lower) signs are assigned according to hemisphere. Filled bars are CP with statistical significance (P < 0.05, permutation test) from the chance level (0.5). Open bars are nonsignificant cases (P > 0.05, permutation test). Blue: vestibular; Red: visual. Arrows represent the means. ***: P < 0.001, **: P < 0.01, *: P < 0.05, NaN: P > 0.05. Same format in the following figures.

Fig. 2.

Microstimulation in VPS induces ipsiversive bias in vestibular heading judgment. (A) Clustering of vestibular (Left) and visual (Right) heading signals in VPS assessed by Pearson correlation coefficient of tuning curves between neighboring sites apart by 100 μm (Top) and 200 μm (Bottom). (B) Examples of a microstimulation experiment, conducted first by a three consecutive sites mapping of tuning functions under fixation-only condition (Left), and then by stimulation during the heading discrimination task (Right). Dashed curves are stimulated trials. Solid curves are nonstimulated trials. Blue: vestibular; Red: visual. (C) Comparison of PSE between microstimulated (vertical axis) and control trials (horizontal axis) for all three monkeys (Left), and individuals (Right three panels). Filled dots are cases with significant difference (P < 0.05, logistic test). Open dots are cases with unchanged PSE (P > 0.05, logistic test). (D) Distribution of PSE shift induced by microstimulation for all three animals (Left) and individual animals (Right three). Positive value means the biased direction is consistent with the preferred heading of the stimulated site, and vice versa for the negative values. Filled and open bars are cases with significant (P < 0.05, logistic test) and insignificant difference (P > 0.05, logistic test), respectively. (E) (Top) similar in D, but positive value means the biased direction consistent with the hemisphere of the stimulated site (ipsilateral), and vice versa for the negative value (contralateral). Horizontal and vertical patch indicates data with significant PSE shift from Left and Right hemisphere, respectively. (Bottom) Data in the Top Left panel were separated into three groups according to the relationship between the stimulated neurons’ heading preference and hemisphere: i) ipsiversive heading preference (Left); ii) orthogonal (fore/aft) heading preference (Middle); iii) contraversive heading preference (Right). The gray semicircles represent hemispheres and the superimposed arrows represent heading preference of the stimulated neurons. Specifically, neurons with heading preference within ±60° straight to the left or to the right are defined as ← and →, respectively. Neurons with heading preference within ±30° around straight forward or backward are defined as ↑ and ↓, respectively.

Heading Tuning Properties in VPS.

We first recorded spiking activity from VPS neurons in six hemispheres of all three animals during the fixation-only task to assess their basic tuning properties in response to vestibular and visual heading stimuli (Fig. 1B). Recordings were made in the posterior portion of the lateral sulcus according to locations defined in previous studies (2, 13). Tuning curves were measured in the horizontal plane with 30° apart, covering the whole 360°. Both single-unit (SU) and multi-unit (MU) activities were recorded. The multi-units were collected for the purpose of clustering mapping to guide site selection for subsequent electrical microstimulation (see later section). However, we found a fairly consistent tuning between SU and MU on the same recording site (SI Appendix, Fig. S1), so we did not distinguish the two types of units in the following analysis.

Many neurons showed significant modulations to vestibular (75.3%) or optic flow (43.7%) stimuli. Among them, about a third (35.6%) neurons showed significant modulations to both heading stimuli, and they were defined as bimodal neurons (Fig. 1C, right pie graph). Overall the strength of vestibular tuning is higher than the visual tuning, indicating that VPS is a multisensory area but with more vestibular-dominant signals (2). Importantly, for those bimodal neurons, majority of the neurons show conflicting vestibular and visual heading preference (Fig. 1C, Left), which is consistent with a previous report (13).

We then examined how the encoded tuning properties, in particular, the two neuron types preferring leftward and rightward, might be related to hemisphere because this would be highly relevant to our later analysis of their decoding effect. We found that the proportion of neurons preferring leftward and rightward heading directions was roughly equal for both vestibular and visual modalities in each hemisphere (Fig. 1D). In addition, tuning strength and maximum firing rate are also similar between neurons preferring ipsiversive or contraversive heading (SI Appendix, Fig. S2 C and D). These results indicate that similar to the other central areas [MSTd: (9, 14); VIP: (11); FEFsem: (12); PIVC: (15); VN: (16); Nodulus: (17)], vestibular and visual heading information is encoded in all possible directions in VPS without clear bias toward either hemisphere.

Although here we have focused on the translational heading signals, previous studies have shown robust rotational signals prevalently existing in the extrastriate visual cortex including the VPS area (18). Therefore we also reanalyzed rotational tuning data from our previous study to examine whether rotational heading preference might instead show a hemispheric organization. Across the three hemispheres we collected data previously (left and right hemispheres of monkey Q, and right hemisphere of monkey N), there was also no significant bias in rotational tuning preference in the vestibular modality. The same conclusion was nearly held in the visual modality, except that only the left hemisphere of monkey Q showed a statistically significant preference for clockwise rotation (P < 0.01, t test), which might be attributed to a sampling artifact due to relatively small sampling size (SI Appendix, Fig. S3). In summary, similar to the translation signals, rotational signals were also roughly equally distributed in terms of ipsiversive or contraversive preference (i.e., hemisphere-independent).

Finally, in addition to sensory information, we also examined whether VPS neurons may carry nonsensory signals (e.g., memory, saccadic motor) by recording their activities under a memory-guided saccade task. We found VPS neurons lacked clear saccade-related modulations (SI Appendix, Fig. S2E), which was very different from those sensory-motor association areas including the intraparietal area (LIP) and FEF, in which neurons are activated especially when making saccade to contralateral visual field (19). This result suggests that VPS should be considered mainly as a sensory area.

VPS Neural Activity Correlates with Perceptual Choice with Respect to Tuning Preference.

We then examined whether heading signals encoded in VPS are functionally linked to the animals’ perception. We first assessed this by computing a correlation metric, which is the covariation between individual neuron’s activity and the animal’s perceptual choice on a trial-by-trial basis (Choice probability, CP, see Materials and Methods) (20). CP has been used to infer a potential bottom–up sensory-driven readout processing, yet it is also debated because alternative sources exist such as top–down feedback, or correlated noise among sensory neurons (Discussion). Nevertheless, we were able to collect 81 single-units from one macaque (Monkey F) while the animal simultaneously performed the fine heading discrimination task with small heading angles varied around straight ahead (−8°, −4°, −2°, −1°, 0°, 1°, 2°, 4°, 8°). We found on average, the mean CP across all neurons was significantly greater than the chance level (0.5, or 50%) in both vestibular (0.536, P < 0.001, t test) and visual conditions (0.537, P < 0.001, t test), indicating that overall, neuronal activity in VPS correctly predicted the animal’s perceptual choice with respect to the neurons’ heading preference (Fig. 1 E, Top). Instead, predicting the animals’ choice based on hemisphere (ipsi- or contra-lateral) only led to a chance level in the vestibular condition (0.4995, P > 0.05, Fig. 1 E, Bottom Left), or even wrong direction in the visual condition (0.477, P < 0.01, Fig. 1 E, Bottom Right).

Thus, the CP results suggest that heading signals encoded in VPS neurons are functionally coupled with perceptual choice with respect to neuronal tuning, instead of hemispheres. However, CP is a correlation metric that does not certainly indicate causality. In particular, positive CP values (>0.5) may arise from top–down feedback signals related to decision formation, rather than a bottom–up sensory driven process (21). Thus the observed positive CP values in VPS neurons may not necessarily imply that the heading signals in VPS are causally read out for forming heading perception with respect to their heading preference. Consistent with this rationale, we found in our data that although the mean CP across the population was statistically significant (>0.5), only a small number of individual neurons showed significant CP values (Fig. 1E filled bars). Compared to previous studies, this proportion is low, suggesting that the heading signals in VPS may not fully capture the functional properties of this area. Next, we turned to the microstimulation technique by precisely manipulating neural activity within a relatively restricted local area, and measured potential change in the animals’ behavioral performance.

Microstimulation in VPS Induces Ipsiversive Perceptual-Bias in Vestibular Condition.

To assess causality of VPS activity in heading perception, we applied electrical microstimulation with low currents (20 to 30 µA). A precondition for this technique to work is that neurons around the electrode tip need to carry similar heading tuning functions in order to generate a consistent signal evoked by the delivered currents. Therefore, we first measured heading tuning curves from multi-unit activity during the fixation-only task along the electrode penetration for three consecutive recording sites with 100 μm apart. We then evaluated clustering of heading signals in VPS using Pearson correlation coefficients between neighboring sites (Fig. 2A). For vestibular tuning, the median correlation coefficient was significantly greater than zero across both 100 μm and 200 μm intervals (P < 0.001, two-tailed sign test), indicating strong vestibular clustering in VPS. The visual tuning also showed a similar trend albeit a bit weaker (Fig. 2 A, Right). Thus, heading signals in VPS are well clustered in local spatial domains, providing a chance to produce positive microstimulation effects. To compare clustering of heading signals in VPS with the other areas, we conducted a cross-area and cross-modality comparison with data from previous studies (SI Appendix, Fig. S4). Briefly, for the vestibular modality, the clustering in VPS is much better than that in the multisensory MSTd (4), or the vestibular-dominant PIVC (see later section). For the visual modality, visual heading signals were less clustered than in MST, MT, or VIP (4, 22–24).

The three-site mapping process was performed in each experiment before applying microstimulation (Fig. 2 B, Left). After identifying consecutive sites with similar tuning functions, we typically chose the middle site for conducting microstimulation experiment while the animals performed the heading discrimination task. The animals’ performance was quantified by fitting psychometric functions with cumulative Gaussian functions. Microstimulation was applied in half of the trials and were randomly interleaved with the other half control trials without microstimulation. Take one experiment for example (Fig. 2 B, Right), microstimulation (dashed curves) significantly biased the animal’s choice (P < 0.05, probit regression) toward leftward in the vestibular condition, and toward rightward in the visual condition compared to the control trials (solid curves). To quantify the bias in the animals’ perceptual choice, we computed the difference in the point of subjective equality (PSE) of the psychometric curves between microstimulated and control trials (PSE_stim − PSE_control). Across population, microstimulation frequently altered PSE (P < 0.05, probit regression) in about half cases in all three animals (monkey M: 59.3%, monkey D: 41.8%, monkey F: 53.6%, Fig. 2C, solid symbols), indicating strong microstimulation perturbation effects in VPS under the vestibular condition.

We then examined whether the microstimulation induced PSE was shifted to a direction that was consistent, or expected from the stimulated neurons’ encoded heading preference. Specifically, a positive value was assigned if the biased direction was consistent with the preferred heading of the stimulated site, and vice versa for a negative sign with biased direction opposite to the preferred heading of the stimulated site. For the exemplary case in Fig. 2B, PSE shift in both vestibular and visual condition was in the opposite direction with respect to the stimulated site’s tuning preference, and thus a negative sign was assigned accordingly for each modality condition. The exact heading preference of each stimulation site was first determined by a vector sum algorithm, and then a coarse categorization of leftward or rightward was assigned according to whether the residual vector was within the 2nd and 3rd quadrants, or 1st and 4th quadrants in the Cartesian coordinate, respectively. Across population, we found the biased direction of PSE was not predictable from the heading preference of the stimulated sites, leading to a symmetric distribution of the PSE shift when the sign was assigned with respect to tuning (mean = −0.157, P > 0.05, two-tailed t test, Fig. 2D). Moreover, the microstimulation effects were largely independent on the tuning strength of the stimulated site (SI Appendix, Fig. S2B). These results suggest that tuning may not be a key factor to the PSE shift in the vestibular condition.

Rather than the tuning, we found that the PSE shift direction was significantly correlated with the hemisphere in which microstimulation was applied (Fig. 2 E, Top row). Specifically, microstimulation consistently induced ipsiversive biases, as evident by microstimulation effects in the six hemispheres of all three animals (mean = 1.757°, P ≪ 0.001, two-tailed t test). One possibility, however, was that the microstimulation effect might only be valid for those ipsilateral-tuned neurons (i.e., left heading preference in the left hemisphere, and right heading preference in the right hemisphere). To examine this possibility, we separated the microstimulation data into three groups: ipsilateral-tuned neurons (Fig. 2 E, Bottom Left), fore/aft-tuned neurons (Fig. 2 E, Bottom Middle), contralateral-tuned neurons (Fig. 2 E, Bottom Right). We found that the microstimulation effect was similar across all these three types of neurons, excluding the possibility of a tuning-hemisphere interaction. Particularly for those contralateral-tuned sites, microstimulation frequently induced ipsiversive PSE shift (Fig. 2 E, Bottom Right), contrary to the prediction based on their preferred heading. Such a result provides strong evidence against a labeled-line decoding and instead supports a hemispheric readout mechanism.

To further quantify relative contributions of each factor (e.g., neural tuning, hemisphere) for microstimulation-induced PSE shift, we applied a multiple linear regression analysis by including a number of regressors (SI Appendix, Fig. S5). We found that only the hemisphere factor had a significant contribution (coefficient = 1.77, P ≪ 0.001), while contribution from the other factors, for example, tuning, was negligible (coefficient = −0.041, P = 0.92). The interaction term was also not significant, excluding the possibility that the tuning may contribute via interaction with the hemisphere factor.

Hemispheric Readout Is Specific to Vestibular Signals in VPS.

To further examine whether the hemispheric readout effect is specific to the vestibular signals in VPS, we applied the following analyses and additional experiments: first, we tested the visual modality condition using the same microstimulation method in VPS; second, we applied microstimulation in a different, neighboring brain region (PIVC); and third, we employed a different, complementary perturbation method (chemical inactivation).

First, microstimulation in the visual condition also frequently evoked PSE shift (27.8%), yet the shifted direction was neither dependent on visual tuning (P > 0.05, two-tailed t test, SI Appendix, Fig. S6A), nor on hemispheres (P > 0.05, two-tailed t test, SI Appendix, Fig. S6B). Since all stimulus conditions including vestibular vs. visual modality, and microstimulation vs. non-microstimulation trials are randomly interleaved trial by trial in one experimental session, this distinct microstimulation effect indicates the hemispheric readout effect is specific to the vestibular signals instead of other modalities like visual.

Second, we applied the same microstimulation experiment in a different region of the PIVC. Anatomically, PIVC is anterior to VPS along the lateral sulcus (SI Appendix, Fig. S6C). Functionally, this area has been thought to be a core vestibular region among the cortical vestibular network since similar to VPS, it also receives many inputs from the other vestibular nodes (2). However, microstimulation in PIVC evoked only a few cases with significant PSE shift in either the vestibular condition (4/55, 7%) or the visual condition (6/55, 11%) (SI Appendix, Fig. S6D, filled bars). Importantly, microstimulation induced PSE shift was neither dependent on tuning preference (SI Appendix, Fig. S6 D, Top), nor on hemispheres (SI Appendix, Fig. S6 D, Bottom). Thus, the hemispheric readout effect is specific to VPS instead of other regions like PIVC.

Third, it has long been argued that microstimulation technique suffers pitfall of possible activation of passing fibers. Thus we applied an alternative method of chemical-inactivation (SI Appendix, Fig. S6E). Specifically, a low volume muscimol (4.5 µL, 10 mg/mL), GABA-A receptor agonist was applied unilaterally in VPS in two of the monkeys (monkey D and F) to suppress its neural activity. We found muscimol injection induced significant contraversive bias (SI Appendix, Fig. S6F) in the vestibular condition (P < 0.01, paired t tests), but not in the visual condition (P = 0.2, paired t tests). Since chemical inactivation is more likely to directly impact the targeted areas, the observed consistent, complementary effect compared to that from microstimulation may exclude the possibility that the latter effect is due to activation of passing fibers in VPS.

Peripheral Galvanic Vestibular Stimulation (GVS) Induces Ipsiversive Bias.

The ipsiversive bias effect revealed by manipulation of the vestibular signals in the cortical VPS is reminiscent of those experiments in humans in which noninvasive caloric, or CVS and GVS, respectively are used to activate peripheral vestibular end organs in inner ears (Fig. 3A). These manipulations typically induce strong perceptual balance bias toward the activated side (25). Does the peripheral manipulation share similar mechanisms as our current results from cortical microstimulation? To explore this, we further conducted GVS experiment on our animals for two goals: in behavior, whether the animal would show bias in the heading discrimination task; and at the neural level, whether VPS would be activated particularly in the hemisphere ipsiversive to the behavioral bias.

Fig. 3.

Noninvasive GVS applied at peripheral vestibular end organs. (A) Schematic of the GVS system for humans (Left) and monkeys (Right). (B) Two examples of GVS experiment with cathode electrode on the Left (Left) and on the Right (Right). Format is same as in Fig. 2B. (C) Population results of GVS induced PSE shift in the vestibular (Left) and visual (Right) condition. Positive values indicate a bias toward the cathode side. (D) (Top) Schematic of fUS setup and the direct current (DC) and alternative current (AC) used in fUS task. (Bottom) Examples of fUS imaging, showing VPS activation during DC (Left) and AC (Right) GVS stimulation. (E) Electrophysiological recordings of VPS neurons during GVS stimulation in the vestibular (Left) and visual (Right) condition.

Specifically, one of the animals (monkey F) performed the heading discrimination task while receiving GVS with 1 mA direct current delivered noninvasively through the mastoid bone behind the two ears (Fig. 3B). Similar to the microstimulation experiments, GVS trials were applied for half of the total trials, which were randomly interleaved with nonstimulated trials. Cathodal electrode, typically activating the vestibular end organs and afferents, was placed on either the left or right ear in different experimental blocks (Fig. 3 B, Top). We found GVS significantly biased the animal’s heading perception to the cathodal side in the vestibular condition (Fig. 3 B and C, blue curves and bars), but the effect was negligible in the visual condition (Fig. 3 B and C, red curves and bars). Thus, GVS applied at peripheral vestibular system produced analogous effect on the subjects’ heading judgments as that produced when microstimulation was applied in the cortical sylvian area.

We then examined whether and how VPS may be activated by GVS. We first recorded VPS activity under GVS through functional ultrasound (fUS) imaging (3), which is a technique that conveniently measures hemodynamics related to neural activity with a relatively high spatial (100 μm) and temporal (400 ms) resolution across a relatively extensive spatial domain (~2 cm). We found GVS evoked significant hemodynamic activity in VPS, however, only when using alternating (2 Hz) rather than direct current (Fig. 3D). Thus, fUS imaging successfully confirms activation of VPS by the peripheral GVS with alternating current. However, under direct current GVS, whether VPS will be more activated in the hemisphere ipsilateral to the cathodal side is elusive.

Therefore, we next sought electrophysiological recordings. We found GVS with direct current evoked significant neural activity in VPS, particularly in the hemisphere ipsilateral to the cathode side than the contralateral side (P < 0.01, sign test, Fig. 3 E, Left), as well as the non- stimulation trials (P < 0.05, sign test, Fig. 3 E, Left). By contrast, this effect was absent in the visual condition (Fig. 3 E, Right). Thus, at the neural level, GVS with direct current enhances VPS activity particularly in the hemisphere ipsilateral to the cathodal electrode side, which may mediate the behavioral bias.

Large Currents Impair Behavioral Sensitivity.

Electrical stimulation will inevitably introduce noise in addition to signal, which may consequently impair behavioral sensitivity (SI Appendix, Fig. S7A). This is the main reason that in the previous sections, we have conducted microstimulation experiments with low currents (20 to 30 μA), trying to avoid impact on psychophysical threshold. However, if an impaired behavioral sensitivity, i.e., increased psychophysical threshold is observed, it can potentially serve to indicate a necessary causal role of the perturbed area in certain cognitive functions (26, 27). To test this possibility, in some of the experiments, we additionally increased the current to 100 μA (SI Appendix, Fig. S7A). We found that large current stimulation indeed induced a significant increase in the animals’ psychophysical threshold in the vestibular condition (P ≪ 0.001, two-tailed paired t tests), but not much change in the visual condition (SI Appendix, Fig. S7C). In addition, the ipsilateral PSE bias effect observed in the vestibular condition under low current stimulation was less consistent under the large currents, potentially due to corruption from the increased noise (SI Appendix, Fig. S7D).

In summary, our microstimulation experiments with small and large currents support that vestibular signals in VPS are both sufficient and necessary, respectively, for heading discrimination.

Discussion

In the current study, we investigated causal roles of vestibular and visual motion signals in the posterior Sylvian area in heading perception. We found that both signals, especially the vestibular, were well clustered in local spatial domains, providing a basis for applying microstimulation to artificially manipulate neural activity with high spatial and temporal precision. Indeed, microstimulation with low currents reliably biased macaques’ heading judgments particularly when based on vestibular cues. Interestingly, the bias direction was largely ipsilateral, rather than being consistent with prediction from the stimulated neurons’ tuning preference. This microstimulation effect is in sharp contrast to the tuning-dependent effect as seen in other sensory systems (e.g., visual) in previous studies.

Probe Functional Links between Vestibular Coding in Neocortex and Perception.

There are a number of vestibular pathways in the brain mediating several fundamental as well as high cognitive functions. Specifically, the two subcortical pathways (mainly via the brainstem and cerebellum) of vestibulo-ocular-reflex (VOR) and the vestibulospinal reflex are responsible for basic functions of visual stability and balance, postural control (28). The thalamus instead mediates two cortical pathways: hippocampal-entorhinal pathway for spatial navigation, and cerebral neocortex (1). Although robust vestibular signals have been identified to exist in many sensory cortices, their exact functions are largely elusive, due to lack of studies involving neural activity recording, manipulation, and behavioral task at the same time. In fact, such an issue also applies to the subcortical vestibular-related areas.

Choice probability (CP) is a window to look into potential functional links between neural activity and perceptual judgments (20), which has been largely used in the visual (29–32), somatosensory (33), and auditory (34, 35) system. In the vestibular system, positive CP has also been found in the central brain including MSTd (36), VIP (37), PIVC (38), vestibular nucleus (39), but not the peripheral afferents (40). However, the debate about CP is that it is a correlation metric per se that does not necessarily indicate causality because of possible sources of top–down feedback (21), noise correlation among neurons (41), and motor choice (42) rather than a bottom–up sensory-driven process. Indeed, in a recent study by combining neural tuning measurement, CP computation, and microstimulation perturbation on the same site in extrastriate visual cortex, it reveals a divergence of CP and microstimulation effects (22). Similarly in our current study, we also found a divergence between the two measurements: individual neurons in VPS showed fairly weak coupling with the animals’ perceptual choice, and the population mean is shifted in the direction consistent with the encoded heading preference, rather than the hemisphere as in the microstimulation effect.

Inactivation is a frequently used method to perturb neural activity in the targeted areas to explore their causal contributions of necessity to behaviors. For example, chemical inactivation with GABA-A agonist muscimol has previously been applied in PIVC and causes deficits in the animals’ heading judgments based on vestibular cue, demonstrating that PIVC is necessary in self-motion perception (43). This result makes it surprising in terms of the absence of microstimulation effect in PIVC in the current study. There are a couple of possibilities. First, necessity and sufficiency are two different, complementary causal roles that may not have to be yoked. Second, we found that although PIVC neurons frequently exhibit robust vestibular tuning, their clustering structures in local spatial domains are much weaker compared to that in VPS. Thus, the poor clustered vestibular signals in PIVC may have limited microstimulation efficiency in this area, in a way similar to the lack of microstimulation effect on the vestibular heading perception in the MSTd (4). Third, VPS is a relatively small area in which microstimulation or chemical inactivation may easily produce significant effects. By contrast, PIVC extends along the anterior–posterior axis, covering a much larger anatomical domain. With this respect, PIVC is likely to contain subdivisions with different functions (44, 45). Therefore, extensive perturbation is probably needed to be able to affect PIVC in a certain behavioral task, such as in Chen et al.’s study. Our microstimulation in PIVC instead, may have missed these key subregions.

In addition to inactivation, microstimulation is another frequently used method to perturb neural activity. Compared to chemical inactivation, the electrical currents can be more precisely confined to local spatial domains. Importantly, microstimulation is typically thought to mainly activate neural activity and thus tests a sufficient causal role (5, 6). Microstimulation has previously been applied in MSTd, producing a positive result in the visual condition (4, 22, 23), but the effect is lacking in the vestibular condition (4). Similar aforementioned reasons for the lack of microstimulation effect in PIVC may also apply to MSTd. However, unlike PIVC, chemical inactivation in MSTd with exactly the same paradigms (e.g., drug dose, 4-sites, bilateral injection) produces fairly small effective size (by ~10% change in psychophysical threshold) under vestibular condition, compared to the visual condition (by several folds change in psychophysical threshold) (4). Therefore, it is likely that the vestibular signals in MSTd may not be critical for instantaneous heading judgments.

Hemispheric Readout of Vestibular Signals for Lateralized Heading Perception.

Previous microstimulation experiments conducted in sensory systems other than vestibular typically reveal readout algorithms based on a labeled-line code, that is, perception is biased toward the preferred feature of the artificially-stimulated neurons (4, 5, 22, 46–50). Here we show, however, microstimulation in the posterior sylvian area mainly induces ipsiversive heading perception based on inertial motion cue, without much relationship with the neural-encoded heading preference. We infer such a distinct readout mechanism observed in the neocortex may be related to the strong lateralized information processing originating from the peripheral vestibular system. Specifically, the mechanical sensors in inner ears—cilia on hair cells for the horizontal canals are arranged in the same direction on each side of the head, so that they are activated by ipsiversive rotation (yaw) and suppressed by contralateral rotations, a “push–pull” mechanism that smartly enhances signal-to-noise ratio of neural response to self-motion (51). For the otolith, although cilia are arranged in both directions in each side, the ones toward ipsiversive direction are predominant (51). Thus the peripheral vestibular afferent inputs are mainly activated by ipsiversive rotation or translation stimulation. Indeed, artificial stimulation with caloric, Galvanic, or sound stimulation also mainly activate the afferents on one side but suppress the other side (52, 53). However, how the peripheral “push–pull” organization is preserved or transformed as signals ascend through the central vestibular pathway remains poorly understood. For example, at the level as early as the brainstem, inputs from both sides already converge and single neurons in the vestibular nucleus could be modulated by either ipsiversive or contraversive self-rotation, or both. This central spatial-temporal convergence of the vestibular signals from both hemispheres (16), is quite unlike the other sensory systems in which peripheral information is mainly transmitted to the contralateral hemisphere. Indeed, bilateral vestibular activations are observed in many central areas with unilateral stimulation of the peripheral end organs (54, 55). Meanwhile, several studies clearly have shown stronger activity in the ipsi-lateral than the contra-lateral hemisphere (56–58).In our current study, we also applied peripheral GVS with direct currents, and showed significantly stronger activation in hemisphere ipsilateral to the cathodal side, which was accompanied by ipsiversive bias in the animal’s heading judgment.

Therefore, the microstimulation effects we observed in the posterior lateral sulcus may reflect an intrinsic characteristic of the vestibular system: signals are read out with respect to hemispheres. This effect may have largely masked contributions from the labeled-line code of individual neurons in VPS. However, this does not mean that a hemispheric code would be true for all other areas in the vestibular network. For example, our microstimulation in PIVC failed to produce such an effect. Thus it is likely that PIVC neurons may contribute to vestibular perception through a labeled-line code. Unfortunately, vestibular tuning functions in PIVC are not well clustered, which may have prevented seeing a labeled-line code-dependent microstimulation effect. Future studies are required to further examine in other areas of the vestibular network, for example, 2 V, 3a, PCC, VN, VPL, etc., whether and how a hemisphere and labeled-line code may respectively contribute to self-motion perception.

In contrast to the vestibular system, a hemispheric decoding effect is much less clear in the visual condition. There might be some hints, however. For example, a previous study (42) discovered that visual neurons in the middle temporal area (MT) tended to show higher activity when choice was made toward their receptive field (i.e., contralateral). Interestingly, we also observed a negative CP pattern of visual signals (Fig. 1 E, Bottom Right), meaning a contraversive effect. In addition, under large current (100 μA) stimulation condition, one animal showed contralateral bias effect (SI Appendix, Fig. S7 E, Bottom Right). Therefore, these data may hint a hemispheric effect, albeit contraversive in the visual pathway. On the other hand, however, more solid evidence needs to be collected to further validate this conclusion in the future.

Uniqueness of VPS in the Multisensory Self-Motion Network.

VPS, located at the posterior part of the lateral sulcus in nonhuman primates, originally defined as the temporal-parietal association area (59), is possibly homologous to the posterior insular cortex (PIC) in humans (54, 55). Albeit close, VPS/PIC is clearly anatomically separated from the more anterior vestibular cortex of PIVC. Both VPS/PIC and PIVC have extensive connections with the other areas, thus they have been proposed to be core in the vestibular network (2). However, there is evidence suggesting that the two areas may execute distinctive functions. First, our microstimulation experiments reveal a positive causal, sufficient effect in VPS but negative in PIVC. This effect is unlikely due to lack of tuning clustering in the latter area, since our analysis shows that the ipsilateral effect in VPS is not related much to the neural tuning. Thus, the same effect in VPS would have been expected in PIVC regardless of how different types of neurons are spatially clustered. Second, PIVC neurons lack clear visual response to complex optic flow (60) while VPS neurons are clearly modulated (13). Although some other studies report that PIVC neurons are modulated by visual motion stimuli, these stimuli are mainly gratings (44) instead of optic flow. Thus, the visual modulation in PIVC may be more linked to optokinetic nystagmus which typically happens in the laminar plane, rather than optic flow-based heading perception which typically happens during forward locomotion (61). Third, previous studies indicate that during visually induced self-motion illusion (i.e., vection), VPS/PIC is activated together with other extrastriate visual cortices including MT+, VIP, V6, V3A, whereas PIVC is instead suppressed (62). Thus the two areas may be involved in different circuitry mediating the bi-stable perception under ambiguous stimuli.

VPS is also distinct from other multisensory areas (e.g., MSTd, VIP, FEFsem) by presenting predominant visual-vestibular conflict heading signals. Under conditions with congruent inertial and visual motion cues, tuning functions of VPS neurons are typically weakened compared to those in either single cue conditions. Based on this unique trait, researchers have speculated (13) that unlike other areas (e.g., MSTd, VIP, FEFsem) containing many visual-vestibular congruent neurons which can facilitate multisensory enhancement (36, 63), VPS is unlikely suitable for self-motion perception in natural environment (13). Our results however, challenge this speculation. First, we show that either heading signal is causally linked to heading perception. Specifically, vestibular signals in VPS exhibited a robust sufficiency and necessity role as evidenced by both microstimulation and inactivation experiments. Regarding the visual modality, microstimulation in VPS also significantly biased the animals’ heading judgments in many cases, although the overall effect was neither consistent with labeled-line, nor hemispheric code (SI Appendix, Fig. S6 A and B, filled bars), which requires further study. Second, we show that from the decoding perspective, the vestibular signals in VPS are read out with respect to hemisphere while the visual (optic flow) signals neither follow labeled-line nor a robust hemispheric code. Therefore, the two signals may be integrated in a way different from conventional thoughts based on their encoding properties.

Materials and Methods

Subjects and System Set-Up.

Three healthy adult rhesus monkeys (Macaca mulatta weighing 7 to 9 kg) were used in this study. Before the behavioral training, the animals were chronically implanted with a circular molded lightweight plastic ring (containing a bottom ring, a middle ring, and a lid). In brief, we used six titanium inverted T-bolts and dental acrylic to fix the bottom ring. The plastic rings were used as both head-fixed post and neural recording chamber. After recovery, monkeys were trained to seat in a customized primate chair with head restraint. The primate chair is fixed in the visuo-vestibular virtual reality system. All animal procedures were approved by the Animal Care Committee of the Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences.

Apparatus.

The visuo-vestibular virtual reality system was composed of a motion platform (MOOG MB-E-6DOF/12/1000 KG) and a vertically mounted LED display (Samsung ED55C), which provided vestibular and visual self-motion stimuli, respectively. The visual display provided approximately 90° × 90° of visual angle, positioned at a viewing distance of about 30 cm from the animals. The stimuli was controlled by custom C++ software and synchronized with electrophysiological recording (AlphaOmega SnR, Israel) and eye tracking (SR Research EyeLink 1,000 Plus, Canada) systems under the general control of the TEMPO (Reflective Computing, U.S.A.).

Behavioral Tasks.

Memory-guided saccade task.

Following established protocols, monkeys were trained to perform a memory-guided saccade task to examine whether VPS neurons may carry nonsensory signals. In brief, each trial was initiated when monkeys maintained fixation on a central point for 1,000 ms. A peripheral target was then briefly presented for 500 ms at one of eight possible locations, each positioned 10° from the display center. Monkeys were required to maintain central fixation. Following the target offset, monkeys were still required to maintain central fixation for an additional 1,000 ms, until the disappearance of the fixation point, which cued them to saccade toward the remembered target location.

Heading discrimination tasks.

Monkeys were comfortably seated in a custom-designed chair with head fixation, which was mounted within the visuo-vestibular virtual reality system. Upon acquiring and maintaining fixation on a central point, two choice targets appeared symmetrically at 10° eccentricity along the horizontal plane of the fixation point. The LED display was rigidly mounted on the motion platform. During vestibular stimulation, the fixation point on the display screen was moved together with the motion platform and the animals, thereby suppressing VOR-driven eye movements. Prior to the formal experiments, the location of the fixation point was carefully calibrated for each animal so that it was aligned with the animals’ straight-ahead gaze direction, which was usually around the center of the visual display. Then, a linear forward motion stimulus that was defined by either inertial motion or optic flow was presented with a small deviation from straight ahead, simulating heading in the horizontal plane. The heading stimuli lasted 1,500 ms. The animals were required to maintain central fixation within a 2° × 2° electronic window. After the stimulus offset, the fixation point disappeared, and indicated the animals to make eye movement toward one of the choice targets. Correct choice associated with the experienced heading led to a drop of juice as rewards.

Each experimental block consisted of two modality conditions: 1) Vestibular-only, in which self-motion cues were delivered exclusively through physical motion of the motion platform; 2) Visual-only, in which self-motion cues were delivered exclusively by optic flow presented on the display while the motion platform remained stationary. All stimuli were generated based on a Gaussian velocity profile combined with a biphasic acceleration pattern, so as to simulate natural transient self-motion within the environment. Task difficulty was determined by heading angles deviated from straight forward (monkey M: ±8°, ±4°, ±2°, ±1°, 0°; monkey D: ±8°, ±4°, ±2°, ±1°, 0°; monkey F: ±5°, ±2.5°, ±1.25°, ±0.625°, 0°). To roughly match the behavioral sensitivity between the two stimulus conditions, we lowered the visual coherence (monkey M: 15%, monkey D: 11%, monkey F: 27%). Thus, each repetition contained 18 interleaved trials (2 stimulus modality × 9 heading angle), and each stimulus condition was typically repeated 20 times, leading to more than 540 trials. In the microstimulation experiment, microstimulation was applied in half trials that were randomly interleaved with the nonstimulated trials. In either trial, reward was always delivered with respect to the ground truth of the real heading stimulus. For the trials with 0° heading, rewards were delivered randomly.

Electrophysiology Recording and Neural Activity Perturbation.

Surgical preparation.

A recording chamber was surgically implanted for each macaque. The chamber was sealed with dental acrylic and embedded with three plastic screw bases designed to secure a recording grid. The grid was manufactured as a thin circular plate (≈ 5 mm thick) featuring a honeycomb array of 0.5 mm diameter guide holes spaced 0.8 mm apart between centers. This hexagonal arrangement provided a standardized coordinate system for precise electrode positioning throughout the recording sessions.

After selecting a target guide hole, a hand-held miniature drill was used to create a vertical passage through the dental acrylic and skull until reaching the dura mater. A custom-fabricated stainless steel guide tube was then advanced to pierce the dura. Each guide tube consisted of two concentrically welded stainless steel cannula of different diameters, with an integrated grounding wire for electrophysiological recordings. The guide tube was served to protect the electrode tip during insertion, facilitating penetration through the dura and any fibrotic tissue. Following penetration, electrodes were advanced into the cortex using a hydraulic micromanipulator (FHC Inc., United States).

Electrophysiological recording.

Both recordings and electrical microstimulation were performed using the AlphaLab SnR multi-channel neural electrophysiology system (Alpha Omega, Israel). Under the centralized control of the TEMPO system, neuronal activity during behavioral tasks was analyzed in real time using MATLAB (MathWorks, United States). For single-unit recordings, we primarily used 16- and 24-channel linear array electrodes (V-Probe, Plexon, United States) with impedances ranging from 100 to 1,200 kΩ. Neuronal data acquired via multi-channel recording were spike-sorted offline using Kilosort 4.0 (64) and subsequently analyzed in MATLAB. In microstimulation experiments, tungsten electrodes with impedances around 500 kΩ were employed.

VPS mapping.

Area mapping was made through cross-validation between structure MRI and physiological properties. Along each electrode penetration, baseline changes of electrophysiological signals were carefully monitored to determine transition patterns of gray and white matter. Recordings and microstimulation were made in the posterior portion of the lateral sulcus (AC: 1 ~ 3 mm) according to locations defined in previous studies (2, 13).

Electrical microstimulation.

Prior to microstimulation, we assessed the functional homogeneity of multi-unit (MU) activity near each candidate site to examine the consistency of neuronal tuning. For each site, tuning curves were measured by presenting heading stimuli (vestibular or visual optic flow) to characterize directional selectivity in the horizontal plane. Eight directions with 45° apart were sampled. The electrode was then moved vertically in a 100 μm step to map tuning curves for three consecutive sites. A correlation analysis was performed across the three sites to evaluate local functional clustering. Sites exhibiting consistent vestibular or visual clustering were preferentially selected for microstimulation. After passive tuning curve mapping, the animals performed the heading discrimination task, in which microstimulation was randomly applied on 50% of trials throughout the 1,500 ms stimulus period. We used constant-current, negative-leading biphasic pulse trains (300 Hz) at two intensities: a low 30 μA to examine a “sufficient” effect, and a larger 100 μA to examine a “necessary” role.

Reversible chemical inactivation.

Muscimol was prepared with a concentration of 10 mg/mL, respectively. They were accurately injected (~4 μL) into the targeted areas at a speed of 0.15 μL/min using the hand-made injectrode. The injectrode was a hollow cannula containing an electrode, allowing both drug-delivery and neural recording. Because the half-life elimination of the drug is about 12 h for muscimol, we therefore tracked the time course of pharmacological effects by assessing the animals’ behavioral performance at 0, 24, and 48 h post-injection.

Peripheral GVS.

GVS was delivered using a C64 Pro electrical stimulator (Quanlan, China). The cathode and anode terminals were connected to 8-mm Ag/AgCl electrode pads (3 M), which were positioned over the mastoid processes behind each ear. Prior to electrode placement, the skin was treated with lidocaine spray to minimize any discomfort or tactile sensation caused by the stimulation. Conductive gel was applied to ensure good electrical contact. The stimulation intensity ranged from 1.0 to 2.5 mA. GVS experiments consisted of two main components: 1) GVS was administered during the monkey’s performance of a heading discrimination task, allowing simultaneous assessment of behavioral modulation and intracranial recording of neural activity in the targeted brain regions. 2) GVS was combined with functional ultrasound imaging (fUS), wherein the monkey passively received GVS without performing any cognitive task while the target brain area was scanned. Three distinct electrical stimulation conditions were applied in sequence: two direct current (DC) conditions with reversed electrode polarities and one alternating current (AC) condition. Each stimulation trial lasted 6 s.

Functional ultrasound imaging.

We combined functional ultrasound imaging (fUSI) with GVS using an ultrasound imaging system (Iconeus, France). The fUS methodology generally followed established protocols (65, 66). Briefly, we transmitted tilted plane waves at a pulse frequency of 55,000 Hz (with 2° increments from −10° to +10°) and compounded the resulting images to generate Doppler maps at a frame rate of 500 Hz. We used a 192-channel linear ultrasound probe (15.6 MHz; spatial resolution: 100 × 100 µm; slice thickness: ∼400 µm). Prior to scanning, a square cranial window was created above the target brain region and a recording chamber was installed. During imaging, the chamber was filled with ultrasound gel or saline, and the probe was fully immersed. A stepper motor advanced the probe in 1 mm steps, providing an effective field of view of 22.1 mm and a maximum imaging depth of ∼1.5 cm—sufficient to cover the VPS area of interest.

Data Analysis.

Direction discriminative index (DDI).

To estimate the strength of directional tuning in passive heading task, DDI was computed as:

$$\mathrm{DDI}=\frac{R_{\mathit{max}}-R_{\mathit{min}}}{(R_{\mathit{max}}-R_{\mathit{min}})+2\sqrt{SEE/N-M}},$$

where R_max and R_min represent the mean firing rates in the directions with strongest and weakest activities respectively; SSE denotes the sum-squared error around mean responses; N is the total number of trials; and M is the number of stimulus directions. The DDI ranges from 0 to 1, with higher values indicating stronger directional selectivity.

Multiple linear regression analysis.

A multiple linear regression analysis was used by including a number of regressors:

$$\mathrm{Y}=\mathrm{\beta }0+\mathrm{\beta }1\mathrm{X}1+\mathrm{\beta }2\mathrm{X}2+\mathrm{\beta }n\mathrm{X}n+\mathrm{\epsilon },$$

where Y represent the actual dependent variable value. β₀ means the intercept of the overall regression line and β_n means partial regression coefficient. The coefficients for each independent variable were estimated using the least squares method based on the observed values of the dependent variable. X denotes the regressors value such as hemisphere, DDI and cluster index of stimulated site. $\mathrm{\epsilon }$ is a random error term.

Choice probability analysis (CP).

CP was computed using receiver operating characteristic (ROC) analysis for individual neurons in area VPS under different stimulus modalities. For each neuron, response in each trial was first normalized by z-score (subtract mean) across the whole trial in each stimulus condition (certain heading and modality). These responses were then grouped across heading conditions to compute a grand CP. Specifically, trials were grouped into two categories according to subjects’ choice. ROC curves were then constructed, and the area under the curve was defined as the neuron’s grand (CP). Significance was assessed by a nonparametric permutation test. A CP value significantly greater than 0.5 indicates a correlation between neural activity and the animals’ perceptual decisions, in the preferred direction of the neuron (stronger response correlates with choice in the preferred direction and vice versa for weaker response). A CP value of 0.5 indicates no significant correlation. A CP value significantly smaller than 0.5 also indicates significant correlation, yet in an unexpected direction (e.g., stronger response correlates with choice in the anti-preferred direction).

Functional ultrasound imaging analysis.

The imaging planes acquired through ultrasound scanning resemble coronal sections but exhibit discernible geometrical distortions. Our primary objective was therefore to identify, within the continuous set of ultrasound planes, the one that most clearly and completely depicted the VPS area. Following scanning, we computed the Pearson correlation coefficient between the Doppler signal in each voxel and the GVS waveform. These correlation values were overlaid as a heat map onto the background structural image—which was itself derived from the time-averaged Doppler signal. To minimize false positives, only regions containing at least 10 contiguous voxels exceeding the correlation threshold were displayed.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Text file data have been deposited in ION319 (10.12412/BSDC.1763025883.30001) (67).