Introduction
Binocular vision allows the visual system to fuse each of the 2D images from our retinas by using the difference between them to estimate relative depths. This is termed binocular disparity and is a powerful cue that enables us to achieve depth perception. Research has utilised random dot stereo-pairs or stereograms (RDS) to isolate the effect of binocular disparity, which triggers the perception of depth. RDS stimuli present a pair of images, one to each eye, which when viewed binocularly produce a fused strong percept of depth (Julesz
1971). This percept can be dramatically enhanced by the use of a dynamic stimulus. Consequently, there are individuals who are unable to resolve static RDS to perceive depth, but successfully perceive dynamic RDS (Fujikado et al.
1998; Watanabe et al.
2008). Similarly, behavioural research has shown better psychometric performance when using dynamic RDS compared to static RDS (Allison and Howard
2000; Norman et al.
2006). Yet, static, dynamic RDS and numerous other types of stereoscopic stimuli have been shown to similarly activate large areas of the visual cortex (Parker
2007).
Previous physiological and neuroimaging literature with macaques and humans has identified depth perception to be associated with increased activity in multiple parieto-occipital regions (Herpers et al.
1981; Likova and Tyler
2007; Livingstone and Hubel
1988; Norcia and Tyler
1984; Preston et al.
2008; Shikata et al.
1996; Tanabe et al.
2005; Uka and DeAngelis
2004) compared to the primary visual cortex (Cumming and Parker
1997; Neri et al.
2004; Prince et al.
2002). This high-level visual processing undoubtedly involves numerous cortical areas. These include Brodmann area 19 (Baecke et al.
2009; Fortin et al.
2002; Iwami et al.
2002; Nishida et al.
2001; Thiyagesh et al.
2009), V3a (Backus et al.
2001; Bridge and Parker
2007; Chandrasekaran et al.
2007; Cottereau et al.
2014; Fang and He
2005; Georgieva et al.
2009; Gilaie-Dotan et al.
2002; Goncalves et al.
2015a; Tsao et al.
2003), intraparietal sulcus (IPS) (Baecke et al.
2009; Buckthought and Mendola
2011; Durand et al.
2009; Fang and He
2005; Negawa et al.
2002; Tsao et al.
2003), dorsal V4 (Brouwer et al.
2005; Iwami et al.
2002; Negawa et al.
2002; Tsao et al.
2003), V5 (Brouwer et al.
2005; Chandrasekaran et al.
2007; Cottereau et al.
2014; Fortin et al.
2002; Freeman et al.
2012; Negawa et al.
2002; Orban et al.
2006; Welchman et al.
2005), V6 (Cardin and Smith
2011), and the lateral occipital complex (LOC) (Brouwer et al.
2005; Cottereau et al.
2014; Freeman et al.
2012; Read et al.
2010; Welchman et al.
2005). Dynamic RDS have been reported to elicit responses in similarly numerous regions of interest (ROI) with pronounced activation to depth in the superior parietal lobe (SPL), inferior parietal lobe (IPL) and intraparietal sulcus (IPS) within Brodmann area 7 (Iwami et al.
2002; Thiyagesh et al.
2009), pericalcarine area (Gonzalez et al.
2005), V5 (Smith and Wall
2008), V6 (Cardin and Smith
2011), and the parietal–occipital junction (Paradis et al.
2000; Tyler et al.
2006). Brodmann area 19 has also been associated with dynamic depth perception, specifically area V3a (Iwami et al.
2002; Paradis et al.
2000) and the fusiform gyrus (Gonzalez et al.
2005; Iwami et al.
2002). Due to the size of Brodmann area 19, and the spatial limitations of our neuroimaging technique, we focussed on this area aiming to capture a neurovascular response to depth perception.
The current study utilises functional near infrared spectroscopy (fNIRS), a non-invasive optical neuroimaging technique that has not been used previously to investigate dynamic depth perception. NIRS measures changes of concentrations of blood oxy- ([HbO]) and deoxy-haemoglobin ([HbR]), monitoring the haemodynamic response (HDR) of neuronal stimulation (Villringer et al.
1993). Our own and previous research has used fNIRS to successfully characterise the cortical HDR to simple visual stimuli proving it to be a reliable neuroimaging technique (McIntosh et al.
2010; Toronov et al.
2007; Ward et al.
2015; Wijeakumar et al.
2012b). To expand this, we employ a complex visual stimulus, which uses binocular disparity to induce depth perception and involves high-level neural processing. By employing fNIRS with healthy young adults we measured absolute changes of [HbO] and [HbR] in response to dynamic depth perception.
Discussion
The current study used fNIRS to measure the HDR associated with depth perception in response to a dynamic depth stimulus. Reliable visual stimulation effects were seen in the right parieto-occipital hemisphere wherein there was a characteristic increase of [HbO] and decrease of [HbR] during presentation of the test stimulus (horizontal disparity, induced depth percept), compared to the control stimulus (zero disparity, perceived as ‘flat’). Our data concurs with and expands on BOLD evidence that relates predominantly to [HbR] (Fabiani et al.
2014; Mehagnoul-Schipper et al.
2002; Näsi et al.
2010; Rees et al.
1997). In the parieto-occipital cortex, we report coupling in cerebral oxygenation with a mirrored image between [HbO] and [HbR] during depth perception (Fig.
2c). An additional significant finding was that of a hemispheric dominance effect with the right hemisphere producing statistically significant changes in [HbO] compared to the left which produced a bimodal HDR (Fig.
2d). Our results contribute to the literature supporting a right hemisphere bias in depth perception (Baecke et al.
2009; Durnford and Kimura
1971; Hirsch et al.
1995; Nishida et al.
2001; Taira et al.
2001) directly contradicting evidence relating to no depth perception lateralisation (Backus et al.
2001; Buckthought and Mendola
2011; Fang and He
2005; Lehmann and Julesz
1978; Mendola et al.
1999; Merboldt et al.
2002; Tsao et al.
2003). This hemispheric dominance controversy is no doubt fuelled by the extent of heterogeneity both in previous research and the current dataset. In Fig.
4 this variability is highlighted and can be seen as Z-transformed group averaged HDRs plotted from both parieto-occipital cortices. Previous research has similarly shown such variation, for example, Huppert et al. (
2006) report fNIRS findings with considerable inter-subject variability in the shape and timing of the HDR relating to motor activity. With respect to complex visual stimulation and fMRI, Baecke et al. (
2009) describe varied data with less than half of their participants demonstrating a right hemisphere bias in depth perception. The authors stress that the inconsistency of results relating to hemispheric dominance may be masked in smaller sample sizes. Therefore, the current results presenting absolute values of [HbO] in the right but not left parieto-occipital cortex in response to induced depth are compelling.
Although the left parieto-occipital cortex appeared to have a trend in the dataset, the lack of statistical response need not be surprising. A previous unpublished fNIRS study in a thesis by Wijeakumar (
2011) reports a similar bimodal HDR for PO3, as well as the primary visual cortex (V1), in response to dynamic RDS. We propose the variance and small sample size to potentially occlude results of depth processing in PO3. Our V1 recordings showed responses to both stimulus images, with no significant HDR to depth perception. It can be argued that V1 HDR present as bimodal signals with V1 activity relating to both of the complex stimuli, regardless of the difference between the images in binocular disparity. Although V1 contains both binocularly and monocularly activated cells and is involved in depth processing, evidence indicates this specialisation occurs higher in the visual pathway. Macaque studies have proposed individual V1 neurons are not selective for conscious processing of stereoscopic depth (Cumming and Parker
1997). It is now widely accepted that single V1 cells generate the cascade of higher-level processing where depth is fully perceived (Herpers et al.
1981; Hirsch et al.
1995; Iwami et al.
2002; Merboldt et al.
2002; Rees et al.
2002). Indeed, high resolution fMRI imaging (7T) has shown that although V1 does show cortical activation in response to binocular disparity, it is not to the same extent as V3A. This is shown to be consistent regardless of the width of disparity used, and Goncalves et al. (
2015a) conclude that activity in area V3A relates directly to the reported perceptual discrimination thresholds of binocular disparity images. High resolution imaging provides promising insights into V1 processing of stereopsis as recent work has suggested it is the deep layers of V1 that show a preference for binocular disparity (Goncalves et al.
2015b). These findings are in accordance with the current proposed feedback mechanisms between cortical areas within the visual system.
A similar study with fNIRs and static RDS found responses to both occipital and parietal–occipital cortices to depth perception (Wijeakumar et al.
2012a). However, Wijeakumar et al.’s control stimulus image was a black screen, therefore the test stimuli was a ‘novel’ response with the consequences of the first presentation always eliciting an onset response. In the current study, both control and test stimuli were identical with the exception that the test RDS had a shift in the dots that induced binocular disparity, therefore capturing the response made specifically to depth. However, both studies (Wijeakumar et al., and the current one) found a parieto-occipital HDR to depth perception. This supports the notion of shared neural correlates of depth perception between types of stimuli, as previously discussed. Additional support for this comes from a fMRI study on humans and macaques in which both static and dynamic RDS activated the same ROI (Tsao et al.
2003). Furthermore (Gonzalez et al.
2005) report similar ROI for both types of depth stimuli in recordings of subdural electrode VEPs in a 47-year old woman. Multimodal imaging has provided us with valuable insights regarding this integration between the vascular HDR and the neural electrical counterpart, i.e. neurovascular coupling (Fabiani et al.
2014; Iwaki et al.
2013). Iwaki and colleagues combined data from both fMRI and MEG and describe the parieto-occipital, intraparietal and posterior infero-temporal regions to be active during perception of 3D objects from moving dots. Further multimodal or combination neuroimaging studies will enlighten this seemingly complex relationship between the neural and vascular responses of the brain regarding depth perception.
Although this study has inherent limitations of one NIRS channel per hemisphere of recording, and a small sample, we report absolute values of [HbO] and [HbR] and therefore oxygenation, in response to depth perception in healthy young adults. An avenue of future work would be to use the same depth stimuli with EEG in order to couple the data with fNIRS, combining two approaches. We also intend to examine the HDR in participants with limited stereoacuity, e.g. amblyopia. Individuals with amblyopia may have reduced or absent stereoacuity, therefore we hypothesise that they would present with an attenuated HDR in the parieto-occipital cortex. Similar results have been presented using EEG where patients with reduced stereoacuity (mostly microstrabismus) have higher VEP thresholds to RDS, particularly in the right visual field (Skrandies
2009).
In conclusion, we have successfully recorded the HDR associated with dynamic depth perception using fNIRS. Healthy young adults showed a characteristic increase of [HbO] and decrease of [HbR] during complex visual stimulation in the parieto-occipital cortex. In line with previous neuroimaging work, we report a HDR over right hemisphere Brodmann area 19 for processing depth perception, with a large effect size. Occipital recordings fluctuated due to the complexity of both the test and control images. Within our young adult sample there was a strong coupling between [HbO] and [HbR]. Our study demonstrates that fNIRS is a suitable technique to investigate the HDR during high-level visual processing of complex stimuli.