Neural and Vascular Responses to Fused Binocular Stimuli:
A VEP and fNIRS Study

Abstract

Purpose.: The aim of our study was to investigate the correlation between neural and hemodynamic responses to stereoscopic stimuli recorded over visual cortex.

Methods.: Test stimuli consisted of a static checkerboard (checks) and dichoptic static random dot (RD) presentations with no binocular disparity (ZD) or with horizontal disparity (HD). Hemodynamic responses were recorded from right and left occipital sites using functional near-infrared spectroscopy (fNIRS). Visual evoked potentials (VEPs) were recorded over three occipital sites to the onset of the same stimuli.

Results.: Early components, N1 and P2, were sensitive to HD, suggesting that an enhanced N1-reduced P2 complex could be an indicator of binocular disparity and stereopsis. VEPs to checks and ZD stimulation were similar. fNIRS recordings showed changes in hemodynamic activation from baseline levels in response to all stimuli. In general, HD elicited a larger vascular response than ZD. Oxyhemoglobin concentration (HbO) was correlated with the VEP amplitude during the checks and HD presentations.

Conclusions.: We report an association between neural and hemodynamic activation in response to checks and HD. In addition, the results suggested that N1-P2 complex in the VEP could be a neural marker for stereopsis and fNIRS demonstrated differences in HbO. Specifically, checks and HD elicited larger hemodynamic responses than random dot patterns without binocular disparity.

Introduction

Depth perception is based on the amalgamation of monocular cues, such as lighting, accommodation, shading, blur, perspective, and motion parallax, and binocular cues, such as convergence and stereopsis. Of these, stereopsis is the principal measure of binocular vision. Multiple studies have used electrophysiologic and psychophysical methods to investigate stereoscopic processing. ^1–4 Aspects, such as visual persistence, hemispheric dominancy, position within the visual field, temporal frequency, and depth reversal rates, affected depth perception. ^5–9 Visual evoked potential (VEP) studies have implicated the N1 (or an early negative wave) and P3 components to be elicited in the visual cortex by stereo and depth-inducing stimuli. ^10–15 In accordance with these results, functional magnetic resonance imaging (fMRI) has shown definitively that areas involved in stereoscopic processing extend from the dorsal occipital ^16–18 through to the posterior parietal areas. ^16,19–21 The relatively new technique of functional near-infrared spectroscopy (fNIRS) can measure changes in blood flow in response to neural activation as absolute or relative concentrations of oxyhemoglobin (HbO) and de-oxyhemoglobin (Hb) levels. fNIRS relies on the principle that amplitude modulated light of two different wavelengths from the near-infrared and visible spectrum will penetrate the scalp and skull through to brain tissue. fNIRS also can be used easily in conjunction with VEPs. In previous studies, we have shown, using fNIRS, early visual processing in response to simple visual stimuli. ^22,23 In our study, we used fNIRS and VEPs, respectively, to quantify and correlate the hemodynamic and neural responses to stereopsis. The aims were 3-fold: to identify neural markers within the VEPs in response to stereopsis over the primary visual cortex, to observe absolute changes in hemodynamic activation in response to static stereograms, and to correlate VEPs to hemodynamic activation to understand if the neural and vascular components underlying simple and complex visual stimuli were linearly coupled. 

Methods

Observers

Five observers took part in the fNIRS study. Three additional observers were used for the VEP study. All who were recruited either were university students or employees with an age range of 18–32 years. All observers also were tested for suitability by measuring stereo (with the TNO stereo test) and visual acuity (with Bailey Lovie LogMAR charts). The requirements were that the stereo acuity was 60 seconds (”) of arc or better and the visual acuity was 6/6 (0.0 LogMAR) or better in each eye. Observers wore appropriate refractive correction if required and had no history of any visual disorders. The experiments were approved by the local Ethics committee and, in accordance with the Declaration of Helsinki, all participants gave their informed consent. 

Visual Stimuli

Three types of test stimuli were used consisting of a static checkerboard and two static, red-green random dot (RD) presentations. The checkerboard presentation (checks) matched the ISCEV ²⁴ standard for small check sizes (check width 15 minutes ['] of arc, and >97% contrast). The luminance of the black checks was less than 1 cd/m² and the luminance of the white checks was 66 cd/m². A luminance-matched grey screen (33 cd/m²) served as the control presentation for the static checkerboard. 

The RD display consisted of 2000 randomly placed red and green squares (3 × 3 mm) subtending 10.3′. Both RD presentations were viewed dichoptically, each through red and green filters. In the zero disparity (ZD) condition the red-green dots when fused appeared as a “flat” surface and, in the horizontal disparity (HD) condition, the red-green dots when fused produced horizontal disparity in bands across the screen, which appeared as horizontal sinusoidal corrugations. From the peak to the trough of the sinusoidal pattern, the maximum disparity was 11′. At the test distance of 1 meter, the visual field was 20.7° × 15.4°. Both RD presentations used black screens as control stimuli. The observers were asked to fixate on a small cross, presented at the center of the screen. 

VEP Recordings

VEPs were recorded using the Brain Vision recorder (Brain Products Ltd., Munich, Germany) with arrangement of Ag/AgCl electrode (actiCAP System; Brain Products Ltd.) based on the International 10–20 System of Electrode Placement. Recordings from occipital locations O₁, O_z, and O₂ were compared to fNIRS recordings. Data were sampled at 1 KHz, and band-pass filtered between 0.15 and 100 Hz. Horizontal and vertical electro-oculograms (EOGs) also were obtained to correct for blink artifacts. All electrode impedances were maintained below 10 KΩ. VEPs were averaged in response to 100 stimulus presentations. Each duty cycle lasted for 1000 ms (test stimulus presentation of 250 ms followed by a blank screen presentation for 750 ms). 

fNIRS Recordings

A two-channel fNIRS oximeter (OxiplexTS ISS, Inc., Champaign, IL) was used to measure absolute changes in hemoglobin concentrations. Eight sources were modulated at 110 MHz, and emitted light at 834 and 692 nm (four sources for each wavelength). The sensor was a flat flexible pad that housed a detector and four emitter pairs at fixed distances from each other (1.93–3.51 cm). The attenuated light after passing through layers of scalp and cortex was detected. The fixed distances between each of the emitter-detector pairs were used to calculate the slopes of averaged light intensity (DC), modulated intensity (AC), and the phase. These values were converted to absolute concentrations of HbO and Hb using a modified version of the Beer-Lambert Law. Total hemoglobin concentration (THb) was calculated to be the summation of HbO and Hb. We report from only the left and right occipital locations O₁ and O₂ as those locations overlying the midline produced nonsignificant fNIRS responses. ²² Each observer was presented with a 10-minute trial for each of the three stimuli. Each trial consisted of 10 one-minute long segments (30 seconds of test stimulus presentation followed by 30 seconds of control stimulus presentation). Observers were instructed to perform a simple button-press task whenever the presentation changed to sustain their attention. 

Data Analysis

For the VEP analysis, the data were analyzed using the BrainVision analyzer (Brain Products Ltd., London, UK). Each trial (1000 ms) was averaged within each observer and a grand average was calculated. Implicit times and amplitudes were calculated for four components, namely P1 (50–150 ms), N1 (120–200 ms), P2 (170–250 ms), and N2 (270–330 ms). For the fNIRS analysis, one-minute segments were averaged within each observer and across observers. The amplitude values recorded during the last 16 seconds of the test and control stimulus presentations for each one-minute segment were used. Comparisons for each chromophore were made between conditions (control and test presentations), stimuli, and locations across observers. SPSS 16 was used to perform repeated measures ANOVAs on the VEPs and fNIRS data separately. Greenhouse-Geisser correction epsilon was used in cases where the Mauchly's sphericity test for normality was violated to correct the degrees of freedom. Subsequent post-hoc analysis with pairwise comparisons was done wherever necessary using Bonferroni corrections. In general, HbO responses were the most stable of the three chromophores. Therefore, a one-tailed bivariate Pearson's correlation of coefficient was used to correlate the HbO and VEP responses. 

Results

VEP Recordings

Four components were identifiable within the majority of VEP waveforms. Figure 1 shows the averaged VEPs in response to checks (blue line), ZD (red line), and HD (black line) at locations O₁, O_z, and O₂. Table 1 shows summarized mean (± SEM) values for N1 and P2 implicit times and amplitudes. P1 and N2 components were variable and did not show significant differences between stimuli. 

Figure 1. 

 

Group averaged VEPs in response to checks (blue), ZD (red) and HD (black) at locations O₁, O_z, and O₂ (n = 9). Arrows: mean locations for P1, N1, P2, and N2 across all three stimuli.

Table 1. 

 

Implicit Times and Amplitudes of VEP Components N1 and P2

	Implicit Times (ms) Mean ± SEM						Amplitude (μV) Mean ± SEM
	N1			P2			N1			P2
	O1	OZ	O2	O1	OZ	O2	O1	OZ	O2	O1	OZ	O2
HD	179 ± 23	155 ± 6	162 ± 5	261 ± 13	259 ± 12	259 ± 11	−7.5 ± 1.4	−8.9 ± 1.5	−8.0 ± 2.0	5.97 ± 1.56	6.0 ± 1.5	6.9 ± 1.4
ZD	143 ± 5	141 ± 4	139 ± 9	235 ± 6	237 ± 7	234 ± 6	−5.6 ± 1.2	−6.4 ± 1.4	−5.9 ± 1.5	9.67 ± 1.80	9.6 ± 1.7	9.7 ± 1.5
Checks	128 ± 6	114 ± 6	125 ± 7	215 ± 9	198 ± 17	224 ± 11	−5.5 ± 1.8	−7.9 ± 2.1	−4.7 ± 2.1	8.62 ± 0.8	8.6 ± 1.3	8.5 ± 1.2

The main effects of stimulus for N1 implicit times (F[2, 12] = 11, P < 0.05), P2 implicit times (F[2, 9] = 13, P < 0.01), and P2 amplitude (F[2, 9] = 4, P < 0.05) were significant. Specifically, P2 was seen to be relatively delayed in HD and ZD conditions than for checks (P < 0.05). Furthermore, ZD elicited a larger P2 than HD at all three occipital locations (P < 0.05). 

Subsequent post-hoc pairwise comparisons revealed that the N1 occurred much later in response to HD than for checks at O_z and O₂ (P < 0.05). This component also occurred later for ZD relative to the response to checks at O₁ (P > 0.05). Additionally, when the data from locations were pooled together, the N1 was later for HD than for checks (P < 0.05). A larger N1 component also was observed for HD than for ZD at O_z (P < 0.05). Pairwise comparisons revealed that the P2 occurred later in response to HD than for checks at location O_z (P < 0.05). 

fNIRS Recordings

Increases in HbO and THb concentrations were observed in response to all three stimuli at both locations. As expected, Hb concentration dropped during stimulation. It took 10 to 12 seconds for the HbO response to rise and reach a plateau. This was maintained until approximately 5 seconds following offset when values returned to baseline levels. Figure 2 shows the HbO, Hb, and THb responses after averaging within each observer and across observers in response to checks, ZD, and HD. Table 2 shows mean ± SEM values for change in HbO, Hb, and THb concentrations. 

Figure 2. 

 

Averaged change in HbO (red line), Hb (blue line), and THb (green line) concentrations (left column, O₁; right column, O₂) from black/grey screen presentation. White area shows the period of test stimulus presentation. Grey shaded area shows the period of control stimulus presentation (grey/black screen). Top, middle, and bottom rows show the changes in hemoglobin concentration in response to checks, ZD, and HD, respectively (n = 5).

Table 2. 

 

Changes in HbO, Hb, and THb Concentrations at Occipital Locations (Mean ± SEM)

			Checks	ZD	HD
Change in chromophore concentration from control stimulation (μM)	HbO	O1	0.92 ± 0.06	0.64 ± 0.03	0.77 ± 0.05
		O2	0.62 ± 0.05	0.61 ± 0.03	0.73 ± 0.03
	Hb	O1	−0.30 ± 0.03	−0.26 ± 0.01	−0.35 ± 0.02
		O2	−0.23 ± 0.03	−0.18 ± 0.02	−0.26 ± 0.02
	THb	O1	0.58 ± 0.04	0.37 ± 0.02	0.51 ± 0.04
		O2	0.35 ± 0.03	0.39 ± 0.04	0.45 ± 0.02

A three-factor ANOVA model with factors of condition (ON/OFF), stimulus (checks/ZD/HD), and location (O₁/O₂) was used. A summary of the results for the main effects and post-hoc analyses are presented in Table 3. All three chromophores showed larger concentrations at O₁ than at O₂ (P < 0.05). Also, responses to HD were larger than those to ZD (P < 0.05). Post-hoc pairwise comparisons between conditions revealed significantly higher amplitudes during the test stimulus presentation than the control stimulus presentation across locations, stimuli, and chromophores (P < 0.05). Pairwise comparisons between locations showed that the response at O₁ was higher than that at O₂ only for the checks stimulus for HbO and THb concentrations (P < 0.05). On the other hand, the change in Hb concentration at O₁ was larger than that at O₂ for ZD and HD presentations (P < 0.05). Pairwise comparisons between stimuli revealed that for HbO and THb concentrations, larger amplitude levels were observed for checks than for HD at O₁ (P < 0.05). Additionally for THb, the response to HD was larger than ZD at O₁ and larger than for checks at O₂ (P < 0.05). 

Table 3. 

 

Summary of Statistical Analyses on fNIRS Measures

	Hemodynamic Response
	HbO	THb	Hb
Stimulus	HD > ZD*	HD > ZD*	HD > ZD*
Location	O1 > O2*	O1 > O2*	O1 > O2*
Condition	Test > Control*	Test > Control*	Test > Control*
Stimulus × Location	*	*	NS
Stimulus × Condition	NS	NS	*
Location × Condition	*	*	*
Stimulus × Location × Condition	For checks : O1 > O2* For O1 : checks > HD*	For checks : O1 > O2* For O1 : checks > HD* For O1 : HD > ZD* For O2 : HD > checks*	For HD and ZD : O1 > O2*

 HbO and THb: Significant main effects of stimulus – for HbO, F(2, 154) = 7, P < 0.05 and for THb, F (2, 135) = 6, P < 0.05; location – for HbO, F(1, 79) = 7, P < 0.05 and for THb, F(2, 135) = 10, P < 0.05; condition – for HbO, F(1, 79) = 584, P < 0.05 and for THb, F(1,79) = 408, P < 0.05; interactions between stimulus and location – for HbO, F(2, 150) = 5, P < 0.05 and for THb, F(2, 155) = 3, P < 0.05; location and condition – for HbO, F(1, 79) = 11, P < 0.05 and for THb, F(1, 79) = 5, P < 0.05; and stimulus location and condition – for HbO, F(2, 151) = 7, P < 0.05, and for THb, F(2, 144) = 5, P < 0.05.

 Hb: Significant main effects of stimulus – F(2, 152) = 3, P < 0.05; condition − F(1, 79) = 283, P < 0.05; and location F(1, 79) = 11, P < 0.05; and significant interactions between stimulus and condition – F(2, 135) = 5, P < 0.05, and location and condition F(1, 79) = 17, P < 0.05.

*  A significance of less than 0.05.

Correlation between VEPs and HbO Response

Large VEPs were associated with large vascular responses. Root mean square (RMS) values of the HbO response (HbO_RMS) were calculated for a time window of the last 16 seconds of the test stimulus presentation for each of the one-minute segments for each observer at each location. For the VEPs (VEP_RMS), each trial was split into 10 sections where each section was an average of 10 segments. RMS values were calculated for a time window between 50 and 330 ms after pattern onset for each section and for each observer at each location (10 RMS values × 5 observers for O₁, O₂, and O_z). The correlation between VEP_RMS and HbO_RMS for checks and HD was significant at O₁ and O₂ (checks at O₁ r ² = 0.45, P < 0.005; checks at O₂ r ² = 0.30, P < 0.05; HD at O₁ r ² = 0.46, P < 0.001; HD at O₂ r ² = 0.36. P < 0.01, Figs. 3a–3d). Furthermore, VEP_RMS at O_z was significantly correlated with HbO_RMS averaged across O₁ and O₂ (O₁ + O₂/2) for checks (r ² = 0.57, P < 0.001) and HD (r ² = 0.60, P < 0.001, Figs. 3e, 3f). None of the correlations achieved significance for ZD. 

Figure 3. 

 

VEP_RMS was correlated with HbO_RMS for checks (a, b) and HD (c, d) at O₁ and O₂. The VEP_RMS at O_z was correlated with HbO_RMS averaged across O₁ and O₂ ([O₁+O₂]/2) for (e) checks and (f) HD.

Discussion

VEPs and fNIRS measure different aspects of brain function. Neural activation of the cortex, which occurs within 50 to 100 ms after stimulus onset, is represented by the former. On the other hand, fNIRS measures the slow vascular changes as a consequence of neural activation with a latency of several seconds. 

Evoked Potentials to Static Pattern Onset Stimulation

Typically, VEPs have complex waveforms with several components. Early components (>200 ms) are said to be involved in the primary stages of visual processing of any stimulus, and are known to originate from magnocellular and parvocellular pathways. ^25–28 Some studies also have shown associations between the early P1 and non-stereoscopic depth. ^29–31 Omoto et al. used figures where smaller squares made up bigger squares to create 2D and 3D images with a concave or a convex effect, and observed an enhanced P1 peak over the temporal-parietal-occipital regions for the latter. ³⁰ However, in our current study, no significant differences in P1 were observed across stimuli suggesting that the underlying early pattern processing mechanisms were similar for each of the three stimuli. 

An early negative VEP component has been implicated in studies examining the neural correlates of depth perception using monocular cues, like perspective, ¹⁴ and binocular cues, such as chromostereopsis. ¹⁵ These studies report source generators of depth processing to be located along the occipital and parietal junctions. Sahinoglu et al. in 2002 used dynamic RD stereogram presentations with convergent disparities to identify an N1 peak (200–400 ms) whose peak latency and amplitude changed with increasing disparity. ³² Many other studies also have reported a similar effect on the early negativity. ^33–39 In our current study, an enhanced N1 peak was observed in response to the HD condition, which appeared as a surface corrugated in depth, compared to the ZD condition, which was a fusible flat surface. An enhancement of the N1 component also has been implicated strongly in studies investigating attentional mechanisms. ^40,41 However, Cauquil et al. in 2006 reported no effect of attention on the N1 component in a paradigm where 2D and 3D perspective cues (monocular) were presented. ¹⁴ In agreement with the results from our study, an enhanced N1 was observed in response to 3D rather than for 2D stimulation. Thus, we support the suggestions made in previous studies of depth perception that N1 might be a marker for neural circuitry involved in depth processing. In our study, the P2 peak amplitude was reduced in response to HD. In contrast, Omoto et al., who used non-stereoscopic stimuli with concave and convex surfaces to induce the perception of depth, reported enhanced P1 and P2 amplitudes. ³⁰ Our results also may be interpreted as showing that N1-P2 complex could collectively represent stereoscopic depth processing. 

Hemodynamic Activation in Response to Static Pattern Stimulation

Visual stimuli including checkerboards, ^{22,23,42–47} rotating shapes, ^48,49 peripheral drift illusion, ⁵⁰ and faces, ^51–55 elicit hemodynamic activation. As expected, a change in hemodynamic activation was observed in response to all stimuli relative to the baseline levels to blank screen presentations. The response characteristics of the averaged waveform (time taken to plateau and recovery to baseline levels) displayed similar trends to those previously reported in response to simple checkerboard stimulation. This held true for all chromophores measured. 

fMRI studies of stereoscopic processing have shown involvement of the dorsal region of the occipital cortex ^16,56,57 and other parts of the visual cortex. ^18,21,58,59 Comparing activated areas in response to static and dynamic stereopsis, Iwami et al. found that the dorsal parieto-occipital region was a common processing area. Area V3A is reported to produce the largest response when random dot presentation with disparity was alternated with zero disparity. ^18,60 Tsao et al. in 2003 compared activated areas in humans and macaques in response to disparity-defined checkerboards formed with random dot presentations. ²¹ They found activation in V3, V4d, and V7 of the visual cortex. We suggested that along with pattern-sensitive and binocularly-active neurons, disparity-sensitive cells must have accounted for the increased activation leading to the hemodynamic responses recorded over the occipital scalp across both hemispheres. This could be explained by increased hemodynamic activation from the disparity sensitive neurons of the visual cortex adding to the responses of neurons tuned to other stimulus features, such as contrast and color. 

Correlation between VEPs and HbO Concentration

Previously, studies combining fNIRS and electrophysiology have used different kinds of stimulation to investigate elements of neurovascular coupling. ^61–65 Some others also have tried to correlate the “fast” signal or the Event-related optical signal ⁶⁶ to VEPs. A study by Obrig et al. showed coupling between fNIRS and VEP measures in response to a reversing checkerboard stimulus. ⁶² They correlated absolute measures of chromophore concentrations with peak-to-peak amplitude values. Similar results were observed by Syre et al. when they correlated VEPs with fNIRS in response to a reversing checkerboard across three different stimulus train durations. ⁶⁷ These results were in accordance with those of Rovati et al., who correlated VEPs and fNIRS responses across different contrast levels. We explored neuro-vascular elements underlying stereopsis. The linear correlation between fNIRS and VEP measures was significant for stimuli with a range of attributes, such as contrast, sharp edges (corrugations), and disparity accounting for up to 30% to 46% of the variance in the data. These stimuli could activate pools of neurons and subsequently induce a vascular response. The flat stimulus ZD lacked disparity and internal structure (sharp corrugations) elicited increased hemodynamic responses, but these were not correlated with VEP amplitude. Finally, the VEP amplitude at O_z was correlated with the HbO responses from O₁ and O₂ for the same stimuli. The VEP at the midline ⁶⁸ is well-modeled as the vector sum of the responses from the right and left occipital cortices. A strong correlation between O_z, and the average of O₁ and O₂ is a good indication of the overall associations between neural and vascular activation. ^69,70  

In summary, specific changes were observed across both VEPs and hemoglobin concentration levels in response to onset and static stimulation. In agreement with previous studies, we suggested that the N1-P2 complex could be a marker of stereopsis in V1. A larger change in hemodynamic activation was observed for RD presentations with a 3D corrugated surface compared to presentations of a flat surface. Finally, VEP amplitudes were correlated with HbO concentration for complex stimuli across two occipital locations. It would be interesting to investigate the effects of stereopsis on neural and hemodynamic activation across occipital-parietal-temporal junctions. 

Footnotes

 Supported by Fight for Sight (SW).

Footnotes

 Disclosure: S. Wijeakumar, None; U. Shahani, None; D.L. McCulloch, None; W.A. Simpson, None