fNIRS Feasibility in Event Perception: fNIRS Is Proficient in Distinguishing Event Boundaries

fNIRS Feasibility in Event Perception: fNIRS Is

Proficient in Distinguishing Event Boundaries

Rogier van Koningsbruggen

University of Amsterdam 11225513

January 2020

Abstract

Human event perception is a complex higher order function and is vital to our functioning. In event related designs, the practical constraints of fMRI pose a challenge for simulating real life situations. Functional Near Infra-Red Spectroscopy (fNIRS) could offer practical advantages and options in investigating naturalistic experience. However, fNIRS is only able to target the cortex due to its limited probing depth and it has inferior spatial resolution to fMRI. Within the field of event perception, events are conceptual chunks containing meaningful information. These events are separated by Event Boundaries (EB). This study defined event boundaries by subjective annotation. We assess fNIRS’ feasibility in

distinguishing Event Boundaries (EBs) in television commercials. Perceptual categories Event Boundary, No-Event Boundary and Onset were analysed and a Linear Discriminant Analysis (LDA) predictive model was constructed to further specify fNIRS’ strength in detecting EBs. Results provide evidence in favour of the feasibility of fNIRS is event perception research, for differences in GLM fit were significant between all categories. Although reaching 89% for Onset, the LDA model wasn’t able to predict EB above chance in the EB-NoEB contrast. These results provide crucial evidence in favour of fNIRS feasibility in event perception research. Indicating the importance of defining a viable control

condition, this study concludes that fNIRS holds great potential in becoming a reliable technique in investigating higher order functions like event perception.

Introduction

When you see someone hang up the laundry, make a sandwich or move furniture the brain anticipates, predicts and (sometimes) makes you act. Knowledge of the sequences of movements, the risks, the difficulties and many more features of the events are key to a proper response. The structure of events around us is continuously predicted by the brain in order to come up with such a response. Judgments are always made about events around us, perceived in whatever modality. The study of event perception is a major subject in neuroscience. The theory is that the brain distinguishes meaningful events and subevents within these events. In an influential review paper regarding perceived event structure, Jeffrey Zacks and Barbara Tversky conclude that ‘an event’s structure includes partonomic relation, perceptual event boundaries, objective feature of object and actor motion, perceptual causal properties, statistical pattern of occurrence and goal relations’ (Zacks & Tversky, 2001). They make clear that an event is a complex term but that it should contain an objective basis.

Event Boundaries

The event’s boundaries (EBs) have been subject of study for decennia. Darren Newtson investigated event segregation and the human perception of such separation. He and his colleague reported a fair degree of agreement across observers to where these breakpoints would be (Newtson & Engquist, 1976). These points seem to be indicated mostly when a change of action or physical properties occurs (Newtson, Engquist, & Bois, 1977). So, there seems to be an objective basis underlying the subjective perception of EBs. Besides, changes in visual input only, for example the changing of a filming angle, are found not to be predictors of EB (Schwan, Garsoffky, & Hesse, 2000). Therefore, EBs seems to be defined by conceptual and perceptual changes relative to a stable scene or activity. An example taken from Swallow, Zacks, & Abrams (2009): ‘When watching a person read a book on a couch, observers might identify an event boundary when the actor changes his position from sitting to lying down and again when he closes the book, signalling a change in his goals.’ Conceptual change is clearly stated as a major EB indicator.

High-order regions like the Angular Gyrus (AG) and posterior Medial Cortex are found to be sensitive to event annotation (Baldassano et al., 2017). In these regions, pattern shifts are used to successfully distinguish the EBs from other points in time using fMRI.

How people divide perceptual information into meaningful events is accounted for by the Event Segmentation Theory (EST). It states that perceptual processing is regulated by modelled representations of ‘what is happening at the moment’, so called event models (EMs). These event models are argued to help with the storage of predictive information. EMs are updated when important situation features change, and therefore EM timings coincide with the timing of EBs. According to EST there is more comprehensive processing at the time of an event boundary. Consequently, the information at an EB will be stored better in the long-term memory (Zacks, Speer, Swallow, Braver, & Reynolds, 2007). A study by Silberstein & Nield (2008) used Steady State Topography (SST) to capture an increase in both left and right lateral frontal activation at the time of a TV commercial ‘apex’, being the main point of the advertisement. There is ample evidence that accounts for the advantageous influence of segmentation on long term memory (LTM) encoding. Research shows that movie recall deteriorates when; movies were edited/cut at EBs (Schwan & Garsoffky, 2004) and when they were disturbed by commercial breaks at EBs (Boltz, 1992). The construction of memory encoding seems to be supported by event boundaries, demanding more computational power of the brain regions involved. A deterioration in performance on other tasks demanding computational power can therefore be anticipated. Confirmatory, visual probe detection is found to be impaired at the time of EBs (Huff, Papenmeier, & Zacks, 2012).

Most compelling evidence comes from Ben-Yakov & Henson (2018) , who found that both the hippocampus and the AG account for event boundaries. Using fMRI Ben-Yakov and Henson measured the brain by means of the response to EBs. Video stimuli were temporally labelled with observed EBs. This was subjectively done by observers, different from the participants.

Combining all subjective judgments on EBs and the objective measurements of fNIRS, this study led to the promising result of a model able to capture the sensitivity of the hippocampus to Event Boundaries. Their method is in accordance with Newtson’s findings in the 80’s and the determination of event boundaries in this study will be based on Ben-Yakov & Henson their procedure.

The role of EBs in LTM encoding is demonstrated in fMRI research. In order to investigate the influence of event segmentation on LTM encoding Ezzyat & Davachi (2011) designed sentence recall paradigm. Subjects had to recall a cued sentence that either preceded the event boundary sentence in the narrative or was the event boundary sentence itself. The recall, demanding LTM encoding to bind the sentences, was found to be weakened for across event information in comparison to within event information. The long-term association between chucks of information preceding and following EBs was found to be impaired. This implies the major influence of EBs on LTM encoding.

The DLPFC

The DLPFC is strongly related to LTM encoding. Ranganath & D’Esposito (2005) gave rise to the substantiated idea that the DLPFC plays a crucial role in organizing information in the working memory (WM). Ranganath proposed the idea that the DLPFC plays a role in LTM-processing though its function in WM-processing. Further research extended the understanding of the DLPFC role in LTM. Intuitively, we remember items better on the long run if these items are used more intensively when they are in the WM. More importantly, the DLPFC was found to be predictive for LTM performance (Blumenfeld & Ranganath, 2006). Ranganath successfully integrated his findings to substantiate the crucial role of the DLPFC in LTM encoding. Using neuroimaging techniques like fMRI, activity in the dorsolateral PFC was found to be predictive of organisation strategies. Strong evidence is reported in studies that show that specifically the dorsolateral PFC is activated when a ‘chunking-strategy’ is maintained (Bor, Cumming, Scott, & Owen, 2004; Bor, Duncan, Wiseman, & Owen, 2003). Chunking is cognitive strategy that tries to optimise WM encoding by consciously assessing meaning to a set of features

like a series of numbers or a set of perceptual features like in events. The determined role of the DLPFC in LTM and chunking strategies in WM encoding give rise to the idea that the DLPFC is a suitable cortical region to investigate EBs.

BOLD from video

fMRI research has shown that BOLD can be used in a video viewing paradigm, to specifically address activation to a certain condition (Danek, Öllinger, Fraps, Grothe, & Flanagin, 2015; Lawrence et al., 2006). Besides LTM, emotional responses to television commercials are of great interest to anyone studying advertising strategies. Shen & Morris (2016) provided evidence that BOLD signal in fMRI from frontal gyri can be decoded to emotional processing while viewing Television Commercials (TVCs).

fNIRS

In the current study, functional near-infra-red spectroscopy (fNIRS) is used. Like fMRI, fNIRS assumes that hemodynamic responses follow up on neural activation. As fMRI creates voxels, fNIRS creates channels between a transmitter and receiver (see figure 1). Although fNIRS exclusively captures superficial cerebral blood flow due to its low penetrability of 3 centimetres, the BOLD-assumption of neural activity is similar to fMRI. Nevertheless, fNIRS is cheaper than fMRI and has many practical advantages. Yet, the ‘poor-man’s-fMRI’ is not able to reach subcortical regions and its functions in cognition and memory. fNIRS uses near-infrared light (700-900 nm) and the light scattering properties of hemoglobin to measure BOLD responses. Infrared light absorption by hemoglobin depends on its degree of binding with oxygen. Oxygenated (O2Hb) and deoxygenated hemoglobin (HHb) have different absorption spectra. The change in transmitted and received light through the channel is registered by fNIRS. It’s probing depth is around 3 centimetres (cm) and only the cortex can therefore be targeted. fNIRS studies show the correlations with cerebral blood flow to be stronger for oxy-Hb than for deoxy-Hb (Malonek et al., 1997; Strangman, Culver, Thompson, & Boas, 2002). This study focusses on the O2Hb channels.

Table 1: The brain imaging techniques fMRI, EEG and NIRS compared: pros and cons. Obtained from Liu, Pelowski,

Pang, Zhou, & Cai, 2016.

LTM is already investigated targeting superficial layers of the cortex using EEG (Khader & Rösler, 2011). In his influential review Wolfgang Klimesch (1999) comes to a similar conclusion; EEG power relates to cognitive and memory processes. Despite that both EEG and fNIRS have an inferior spatial resolution to fMRI, the evidence that superficial layers of the cortex are successfully scrutinized for LTM encoding offers perspective for the applicability of fNIRS in investigating these layers. The hippocampi and prefrontal cortex circuity are known to be involved in cognitive control, crucial for memory function (Simons & Spiers, 2003). Complementary to EEG, the use of NIRS enabled researchers to investigate changes in cortical activation patterns with a better spatial resolution than using EEG alone (Balconi, Grippa, & Vanutelli, 2015). Integrating these findings, bottom-up/subcortical originating memory encoding is expected to influence frontal activity, and thus fNIRS measurements as well.

Figure 1: Representation of fNIRS measurement. The source (transmitter) transmits near infra-red light. The light

detected by the detector (receiver) will differ in intensity due to the absorbing properties of hemoglobin in the blood.

Crucially, fNIRS has practical advantages over fMRI and EEG. NIRS systems are portable and can be used with a battery. NIRS can therefore be used in ‘day-to-day environments’ like domestic environments or offices. Decision making and its neural facets can be investigated narrowing the gap between experimental and real-life experience. The challenge of translating the experimental paradigms to normal life situations, which is big for EEG and fMRI paradigms, can be mitigated at least partly by fNIRS. Table 1 shows that NIRS is cheaper, more convenient and acceptably accurate in both time and space when compared to the expensive fMRI and the spatially inaccurate and inconvenient EEG. These advantages give reason to investigate if fNIRS is suitable for investigating event perception. The dependency of LTM-encoding on EBs is already investigated by the high-end version of fNIRS: fMRI. Nevertheless, these studies focused on subcortical structures. The current study, fNIRS channels were formed over the dorsolateral, ventrolateral, dorsomedial and ventromedial PFC. Difficult to grasp objectively, and therefore greatly studied, understanding subjective naturalistic experience remains a key goal of neuroscience. Predicting the subjective by objective means is the goal which has interest from both the academic as the business world. Provided literature substantiates that capturing event perception in naturalistic experience is possible. Combining evidence from pre-frontal cortex studies on LTM encoding in event segregation, plus the recent compelling evidence of the hippocampus accounting for EBs and the promising beneficial properties of fNIRS give rise to the question if

functional near infra-red spectroscopy (fNIRS) is able to distinguish subjective perceptual categories in naturalistic experience. And if so, can a model be trained to predict these categories solely from fNIRS data?

This study targets the dorsolateral, dorsomedial, ventrolateral and ventromedial PFC. As the studies stated above suggest, the effect of EBs and their involvement in memory encoding on PFC activity is substantiated. Therefore, event boundaries are hypothesised to be distinguishable from No-EB timepoints: defined as the timepoints between two EBs. The aim of the current study is to investigate if fNIRS is able to distinguish and predict event-boundaries in TV-commercials (TVCs). The pervasive aim will be the contribution of scientific evidence to the applicability and feasibility of fNIRS brain imaging in event perception research.

This study combines functional Near-Infra-Red-Spectroscopy (fNIRS) and a paradigm consisting of brand logo’s and corresponding commercials to investigate the applicability of this new technique. In this study, subjects watched television commercials. fNIRS data corresponding the pre-annotated EB timings in the commercials will be analysed to see if fNIRS data holds predictive information about EBs.

The TVC Onset timing (the start of the clip) is expected to show the most activity in all O2Hb channels, and especially the DLPFC channels (2,8,28,38). EB timings are believed to show greater activity than no-EB timings but less than Onset timings (see figure 2). The sensitivity of the fNIRS data to the perceptual categories are expressed by the beta weight from the GLM fit. Group differences in fit between perceptual categories are expected (see figure 3). Then, a predictive model will be constructed over all channels.

Figure 3: Expected GLM fit over categories. Mean predictor fit (ß) from GLM will be analysed over O2Hb channels and TVCs to obtain the overall effect of perceptual category.

Method

Subjects

The 63 healthy volunteers (28 M, 35 F, mean age = 23.7, range =19-53) were paid for participating. The subjects were randomly assigned to one of six versions of the experiment.

Recordings

To conduct this and other experiments, a fNIRS – EEG setup was used. For this particular study only the fNIRS data was used. The NIRS system was the Artinis Brite-24 (Artinis B.V, Elst, Netherlands)2

with a 50 hertz sampling rate. The NIRS signals were acquired through OxySoft. Because of the EEG sampling rate and the simultaneous sampling of fNIRS data, the fNIRS data was resampled to 125 Hz. The standard wavelengths 760 and 850 nm were used. These wavelengths are scattered by almost all biological tissue, but hemoglobin has the feature of absorbing them. The degree of deoxygenated (HHb) and oxygenated (O2Hb) hemoglobin are obtained by the Brite-24 using the modified Lambert-Beer Law (Sassaroli & Fantini, 2004).

Figure 4: The canonical fNIRS function (Nishiyori, 2016). The blue shaded block indicates the stimulus duration. The dotted line is the baseline from which concentration changes are plotted in red (O2Hb) and blue (HHb). On the y-axis the change in

concentration and on the x-axis the time (s).

The positions of the transmitters and receivers were mirrored for the two hemispheres (see Appendix 1). A schematic view of the transmitters and receivers used to target the dorsolateral (see

fig. 3), ventrolateral, dorsomedial and ventromedial PFC is given in figure 4. The optode positions were manually chosen and realised in the Artinis fNIRS neoprene caps sizes M and L2

(https://www.artinis.com/brite#brite-fnirs-technique-section). Head circumference was measured to determine the suitable cap size (<57 cm  size M, >57 cm  size L). The absolute distance between transmitters (Tx) and receivers (Rx) was 3 cm. For optode positioning the EEG location FP1 and FP2 were used as starting point for orienting the rest of the 2x12 template (12 channels per hemisphere).

Annotation

For annotation, the method introduced by Ben-Yakov & Henson (2018) was used. Because current observers were able to pause and rewind the TVCs the procedure in their study of subtracting 900 ms of each annotated timing to control for reaction time was skipped. In their 2018 study, boundaries were a minimum of 4300 ms apart (threshold). Timings closer to each other than 4.3 seconds were averaged, and this new value was treated as the resulting EB. Supporting this, research on vascular responsivity using optical imaging shows a reduction in responsivity to temporally close, and short lasting (500 ms) stimuli lasting 4 seconds (Cannestra, Pouratian, Shomer, & Toga, 1998). Integrating this evidence, in this study the

Figure 3: DLPFC location Picture from the Journal of Affective Disorders: “Accelerated intermittent theta-burst stimulation in major depressive disorder. Focus on the reward system” by Romain Duprat3

temporal difference between EB and no-EB timings were controlled to be minimally 4 seconds. Because the no-EB timings were defined as the median of sequential EB timings, EB timings were averaged using Ben-Yakov & Henson’s method if occurring less than 8 seconds apart. The onset of the TVC is believed to be a major event boundary and is therefore analysed separately. The Onset timing was defined as the time that a TVC started playing on the screen.

Experimental paradigm

The behavioural paradigm, programmed with EventIDE (Okazolab, Delft, Netherlands http://www.okazolab.com/), consisted of 3 blocks. In blocks 1 and 3 the participants were presented brand logo’s together with words and in block 2

TVCs were presented (see appendix 2 for the whole paradigm). For this study, we focussed on the TVC block (see figure 5 for the detailed schema of the study specific paradigm). Blocks of 5 TVCs were divided by self-imposed breaks the participants were instructed to attentively view the commercials. A fixation cross appeared between all TVCs. 25 TVCs were shown corresponding with the brands of the brand logo blocks. The total duration of the TVCs block was kept within a 20 second difference over versions. A total of 150 TVCs were used (see appendix 3).

Figure 5: Schematic representation of the paradigm. Brand Logos Post is schematically identical to Brand Logos Pre. In bold the duration of the Inter-Stimulus-Interval (ISI) is described.

Figure 4: A schematic view of the fNIRS optode placement on the left side of the head. The absolute distance between the transmitter (Tx8) and the receivers (Rx5, Rx6, Rx7 and Rx8) is 3 cm. The channels are indicated by the red lines. See Appendix 1 for real pictures.

Rx5

Analysis of fNIRS data

Analysis was done in Matlab 2007b (The Mathworks, inc. Matick, USA). Raw fNIRS signals were smoothed by applying a 5th order cubic Savitzky-Golay filter with a frame length of 12 seconds

(https://nl.mathworks.com/help/signal/ref/sgolayfilt.html, Boston, USA). Next, the smoothed signal was filtered with a Chebyshev Type 1 lowpass filter (passband freq. 2 Hz, stopband frequency 5 Hz, Passband ripple 1dB, stopband attenuation 20dB) (https://nl.mathworks.com/help/signal/ref/cheby1.html, Boston, USA). For analysis, a stick model timeseries was generated for each perceptual category (EB, NoEB & Onset) and convoluted with the canonical BOLD function. The red line in figure 2 illustrates the general idea of the shape of the function. The resulting convolved model with 3 predictors was regressed against the standardized and mean-centred timeseries using a glmfit, resulting in 3 betas per O2Hb channel. Further statistics were performed in R (version 3.6.2) & Rstudio (RStudio Team 2015, Boston, USA, www.rstudio.com). O2Hb channels were included in a two-way repeated measure analysis of variance (rm-ANOVA) of the betas (Malonek et al., 1997; Strangman et al., 2002). A Bonferroni corrected pairwise-t-test was conducted between categories. Also, the effect of channel on the category difference was obtained by conducting a Bonferroni corrected pairwise-t-test between channels. A Linear Discriminant Analysis (LDA) was performed over the same channels (see fig. 4) to see if binomial classification between perceptual categories (contrasts EB/NoEB-Onset & EB-NoEB) was feasible with the fNIRS data. Therefore, data was split up in a train (70%) and validation (30%) set. For the EB/NoEB-Onset contrast EB and NoEB were considered as a single category. Literature on LDA points out it’s high classification accuracy in BOLD response on naturalistic stimuli (Lix & Sajobi, 2010; Mandelkow, De Zwart, & Duyn, 2016).

Results

The main interest of this study was whether perceptual categories in continuous naturalistic experience could be distinguished with fNIRS.

Figure 6 shows the main results of this study. The

performed GLM resulted in a fit per predictor ß per subject per TVC per channel. Beta’s were averaged over channels and TVCs to obtain the mean fit of the perceptual categories. The within subject variance was taken into account by computing a repeated measures ANOVA. An overall difference between categories was found (ANOVA, F (1.14,70.79) = 35.42, p = <0.0001). Using Bonferroni adjusted Pairwise-t-test, Onset (mean = 37.7, SD = 49.7) was found to differ significantly from both EB (mean = -6.08, SD = 25.9, p =< 0.0001) and NoEB (mean = -12.09 SD = 25.9, p = <0.0001) over all channels. Likewise, EB and NoEB were significantly different over all channels (p = <0.05). To obtain which channels hold valuable information about perceptual category a pairwise comparison was conducted. Table 2 lists the channels in which the effect of perceptual category was significant. A visual representation is given in

figure 7.

A second interest of this study was the feasibility of a fNIRS based model in predicting perceptual category. The Linear Discriminant Analysis model contained the oxygenated O2Hb channels. The model was trained over 70% of the data and validated over 30%. Two binomial classification models were built; EB vs NoEB and EB/NoEB vs Onset. The first model classified EB / NoEB but wasn’t able to do this above chance level (47%). The predictive power didn’t improve over DLPFC channels (2,8,28,38) when only using the Oxygenated fNIRS channels (see Appendix 4). Nevertheless, classification accuracy of 89% was found when considering EB and NoEB together as one class and Onset as another. The performed permutation significance test based on cross validation yielded a non-significant p-values for the EB-NoEB contrast, but a significant p-value (0.001) for the EB/NoEB-Onset contrast.

Figure 6: Comparison of perceptual categories Event Boundary, No Event Boundary and Onset. Mean predictor fit (ß) from GLM was analysed over O2Hb channels and TVCs to obtain the overall effect of perceptual category. *, p<0.05 ****,p<0.0001

Table 2: Effects of category within channel Results from pairwise comparison between O2Hb channels are shown. Channel number, name, F-statistic and significance are shown for significant channels only.

Channel name F-statistic Significance * = p<0.05 ** = p<0.01 *** = p<0.001 2 Rx1 – Tx1 O2Hb 8.372 ** 6 Rx2 – Tx1 O2Hb 20.137 *** 8 Rx2 – Tx3 O2Hb 19.961 *** 10 Rx2 – Tx4 O2Hb 8.755 *** 12 Rx3 – Tx2 O2Hb 10.88 ** 14 Rx3 – Tx3 O2Hb 20.515 *** 16 Rx4 – Tx3 O2Hb 12.052 ** 18 Rx4 – Tx4 O2Hb 15.241 ** 20 Rx3 – Tx5 O2Hb 28.002 *** 22 Rx4 – Tx5 O2Hb 40.545 *** 24 Rx5 – Tx6 O2Hb 4.241 * 30 Rx7- Tx8 O2Hb 6.785 ** 33 Rx7 – Tx10 O2Hb 5.648 ** 40 Rx8 – Tx10 O2Hb 24.756 *** 42 Rx6 – Tx9 O2Hb 19.354 *** 44 Rx8 – Tx9 O2Hb 25.264 ***

Figure 7: Visual representation of the effect of perceptual category within the O2Hb channels. Green indicates a significant effect (see table 2 for specifications on the effects). Red indicates no significant effect was found.

Discussion

Shining light on fNIRS’ feasibility to predict perceptual categories in naturalistic experience, present findings provide promising evidence for fNIRS application in naturalistic experience studies. As expected, the three perceptual categories (Event Boundary, No-Event Boundary and stimuli Onset) were found to be distinguishable using fNIRS. To test the strength of these differences, binomial classification models were used. Contrary to the findings of Ben-Yakov & Henson (2018) and our hypothesis, our data underlying the significant difference between categories EB and NoEB was not strong enough to predict category above chance. Yet, combining EB and NoEB and contrasting it to Onset resulted in a classification accuracy of 89%.

Perceptual Categorization

From present evidence it can be concluded that activity in the dorsolateral, ventrolateral, dorsomedial and ventromedial PFC holds information about perceptual changes and that this can be measured with fNIRS. This ties with former research on event-related perception. Others have shown that the DLPFC has a great share in LTM encoding, involved in event segmentation (Ezzyat & Davachi, 2011). In compliance with earlier research by Zacks et al. (2007), LTM is thought to substantiate the difference between EB and NoEB conditions in the present study. This idea settles with the Event Segmentation Theory (EST) which states that at the time of an event boundary, comprehensive processing and superior LTM encoding are attained. Tx10 Tx9 Tx6 Tx8 Tx7 Tx1 Tx3 Tx2 Tx5 Tx4 Rx7 Rx8 Rx6 Rx5 Rx1 Rx2 Rx4 Rx3

The observers who annotated the EB timings did this subjectively, but with the same definition and example (see introduction or Swallow et al., 2009). As the study by Baldassano et al. (2017) suggested, higher order regions in the frontal cortex are found to be sensitive to event annotation. Results (see table 2 and figure 7) suggest that a widely distributed region is sensitive to perceptual categories.

Nevertheless, the found effects are believed to be substantiated greatly by the difference between Onset and the other categories. Onset, defined as the start of the TVC, is a strong but coarse Event Boundary. The perceptual change is major, where the perceptual change in the contrast EB vs NoEB is subtle. The prefrontal cortex is already known to be involved in many higher order functions like event perception. All in all, the current study wasn’t able to further specify the regions involved in event segregation of naturalistic experience.

Classification

The classifying results hint that classification of perceptual categories in naturalistic experience using fNIRS is possible, yet difficult. In this study fNIRS wasn’t able to predict subjectively annotated perceptual categorization (EB vs NoEB). The linear discriminant classification was performed with the O2Hb channels as predictor variables (see Appendix 4). The accuracy did not improve by restricting the predictor variables to be channels 2, 8, 28 and 38 only (DLPFC).

This surprising result could be explained by methodological inconsistency. Due to practical problems, the method for choosing the right cap size was not as robust as it should be. The procedure was altered within the months of experimenting and sometimes inadequately worked out. This could have resulted in too much variation in optode positioning between subjects. Unfortunately, the desired regions might not be to the targeted by the same channel for every participant. In addition, fNIRS’ has a decent, but limited spatial resolution of approximately 1 cm (Quaresima & Ferrari, 2019). These factors are thought to account for the lack of sensitivity of the DLPFC to EBs in the current study.

Too, there are limitations to the annotation approach that can account for this surprising finding. First of all, the number of observers

annotating the EBs was relatively low is this study. By definition, subjectivity between observers has a lot of variation. An increase in observers and therefore annotation points can bolster the strength of the annotation and bundle subjectivity to approach an objective definition of an EB. Second and most importantly, the NoEB timings were defined as the timepoints between two EBs. Based on a 4 second reduction in vascular responsivity (see Cannestra et al., 1998), EBs were averaged if temporally closer than 4 seconds as long as EBs met this criterion. Nevertheless, the hemodynamic response to brief stimuli is believed to sustain for a temporal window of 4 to 6 seconds (Bandettini, Wong, Hinks, Tikofsky, & Hyde, 1992). Still, the chosen threshold of 4 seconds can be substantiated by the need for many annotations for they would massively decrease with a higher threshold. The optimal ratio is yet to be found but increasing the number of observers will improve the strength of annotation unquestionably. Another possible way to test the significance of fNIRS sensitivity to EBs is by random permutation on GLM fitting of the predictor. This clever permutation method is described by Ben-Yakov & Henson (2018). This way, the control condition NoEB and potential data loss due to averaging timepoints (as described) can be omitted. The EB timings are expected to show highest beta weights for the binary convoluted predictor. By fitting the predictor randomly to different timings in the time-series data, the beta is expected to be lower in comparison to the ‘in-tact’ EB timings fit. In conclusion, the control condition can be defined differently and options like the permutation method described above should be investigated for fNIRS specifically.

A high classification accuracy (89%) in predicting Onset vs EB/NoEB shows that changes in bottom-up input can not only be captured by fNIRS but can also be predicted by a discriminant analysis. Although highly anticipated, fNIRS is able to predict heavy changes in perceptual input like the Onset of a television commercial. This seemingly obvious finding, combined with the significant group differences between EB and NoEB, does give rise to the notion that perceptual categories can be predicted with fNIRS, but a better control condition needs to be developed. From this standpoint more evidence has to build on the annotating procedure and NoEB definition

before concluding on fNIRS feasibility to predict event boundaries.

It is important to note that study paradigm consisted of more than watching commercials. Block 1 and 3 (see appendix 2) were included to answer questions about brand perception. No effect is expected from viewing the logo’s and words in the first block. But, to fully exclude an effect on the event perception in TVCs, future research should use a paradigm in which subject view naturalistic clips only.

Additionally, TVCs had different numbers of annotations. Only TVCs with more than one EB and thus minimally one NoEB were analysed. But most TVCs contained a lot of event boundaries (10-20). Due to the persistent nature of hemodynamic responses, fast occurring events are likely to contain signal change from previous events. The discussed random permutation analysis can solve this problem by omitting the definition of a control trial. Some TVCs had only two perceived event boundaries which were far apart, in this case the EB and NoEB are thought to be more distinguishable as NoEB is defined as the median timepoint between the two EBs. To first acquire a proof of concept, TVCs with sparse, temporally diffuse annotated EBs can included only in future research.

Implications

Combing present findings and profound evidence of the hippocampus holding predictive information on EBs leads to the suggestion that subcortical based information on event perception is propagated to frontal regions. The close relation between event boundaries and LTM encoding is thought to substantiate this propagation. fNIRS, lacking the probing depth to reach subcortical structure, can therefore pick up this information at superficial layers of the cortex. In order for future naturalistic experience research to benefit maximally from the advantageous properties of fNIRS, this suggestion has to be further investigated.

EBs are crucial in naturalistic experience because they are key point of subliminal attention. The current study used television commercials to simulate such experience. The information about where these key points in naturalistic experience

would be, are of incredible value to understand human perception. Also, the feasibility of fNIRS to capture the differences between EB and NoEB lead to great prospect for this technique in business related science like (neuro)marketing. Especially because of its convenient and suitable properties in investigating behaviour in naturalistic environments like a supermarket, or when playing a sport. In contrast to fMRI, with fNIRS subjects can freely move during an experiment. Subsequently, fNIRS paves the way for progress in narrowing the gap between paradigm and real life.

Various academic fields profit greatly from the growing body fNIRS literature. With its decent ratio between temporal and spatial resolution, fNIRS has shed new light on neurodevelopment, psychiatric conditions, neurorehabilitation and cognitive impairment (Pinti et al., 2018). Most compelling are the practical advantages fNIRS brings to the field of neuroscience. fMRI, one of the most popular neuroimaging techniques to date, has obvious limitations in naturalistic experience simulation. fNIRS can be used to investigate the hemodynamic response of children, adults, elderly, mentally/physically disabled. In spite of these advantages, fNIRS cannot obtain anatomical information and has a low penetration depth. Still, and especially in the field of studying naturalistic experience, fNIRS is a robust technique and it holds great potential for future research. The current study has shown that perceptual categories in naturalistic experience can be distinguished with fNIRS, although its predictive power seems to rely considerably on data analysis strategies. We have provided valuable insight in fNIRS’ capabilities and discussed improvements on these strategies, in order to bolster forthcoming research.

Conclusion

At this stage of understanding, fNIRS is considered a viable method for investigating major perceptual stimuli. In order for fNIRS to be feasible in predicting subtle perceptual changes like event segmentation, more research has to be done. The described methodological limitations of the current study can help improve future work on the enigmatic objectification of subjective perceptual naturalistic experience.

References

Balconi, M., Grippa, E., & Vanutelli, M. E. (2015). What hemodynamic (fNIRS), electrophysiological (EEG) and autonomic integrated measures can tell us about emotional processing. Brain and Cognition. https://doi.org/10.1016/j.bandc.2015.02.001

Baldassano, C., Chen, J., Zadbood, A., Pillow, J. W., Hasson, U., & Norman, K. A. (2017). Discovering Event Structure in Continuous Narrative Perception and Memory. Neuron.

https://doi.org/10.1016/j.neuron.2017.06.041

Bandettini, P. A., Wong, E. C., Hinks, R. S., Tikofsky, R. S., & Hyde, J. S. (1992). Time course EPI of human brain function during task activation. Magnetic Resonance in Medicine.

https://doi.org/10.1002/mrm.1910250220

Ben-Yakov, A., & Henson, R. N. (2018). The hippocampal film editor: Sensitivity and specificity to event boundaries in continuous experience. Journal of Neuroscience. https://doi.org/10.1523/JNEUROSCI.0524-18.2018

Blumenfeld, R. S., & Ranganath, C. (2006). Dorsolateral prefrontal cortex promotes long-term memory formation through its role in working memory organization. Journal of Neuroscience.

https://doi.org/10.1523/JNEUROSCI.2353-05.2006

Boltz, M. (1992). Temporal Accent Structure and the Remembering of Filmed Narratives. Journal of

Experimental Psychology: Human Perception and Performance. https://doi.org/10.1037/0096-1523.18.1.90

Bor, D., Cumming, N., Scott, C. E. L., & Owen, A. M. (2004). Prefrontal cortical involvement in verbal encoding strategies. European Journal of Neuroscience. https://doi.org/10.1111/j.1460-9568.2004.03438.x Bor, D., Duncan, J., Wiseman, R. J., & Owen, A. M. (2003). Encoding strategies dissociate prefrontal activity

from working memory demand. Neuron. https://doi.org/10.1016/S0896-6273(02)01171-6

Cannestra, A. F., Pouratian, N., Shomer, M. H., & Toga, A. W. (1998). Refractory periods observed by intrinsic signal and fluorescent dye imaging. Journal of Neurophysiology.

https://doi.org/10.1152/jn.1998.80.3.1522

Danek, A. H., Öllinger, M., Fraps, T., Grothe, B., & Flanagin, V. L. (2015). An fMRI investigation of

expectation violation in magic tricks. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2015.00084 Ezzyat, Y., & Davachi, L. (2011). What constitutes an episode in episodic memory? Psychological Science.

https://doi.org/10.1177/0956797610393742

Huff, M., Papenmeier, F., & Zacks, J. M. (2012). Visual target detection is impaired at event boundaries.

Visual Cognition. https://doi.org/10.1080/13506285.2012.705359

Khader, P. H., & Rösler, F. (2011). EEG power changes reflect distinct mechanisms during long-term memory retrieval. Psychophysiology. https://doi.org/10.1111/j.1469-8986.2010.01063.x

Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: A review and analysis. Brain Research Reviews. https://doi.org/10.1016/S0165-0173(98)00056-3

Lawrence, E. J., Shaw, P., Giampietro, V. P., Surguladze, S., Brammer, M. J., & David, A. S. (2006). The role of “shared representations” in social perception and empathy: An fMRI study. NeuroImage.

https://doi.org/10.1016/j.neuroimage.2005.09.001

Liu, T., Pelowski, M., Pang, C., Zhou, Y., & Cai, J. (2016). Near-infrared spectroscopy as a tool for driving research. Ergonomics. https://doi.org/10.1080/00140139.2015.1076057

Lix, L. M., & Sajobi, T. T. (2010). Discriminant analysis for repeated measures data: A review. Frontiers in

Psychology. https://doi.org/10.3389/fpsyg.2010.00146

Malonek, D., Dirnagl, U., Lindauer, U., Yamada, K., Kanno, I., & Grinvald, A. (1997). Vascular imprints of neuronal activity: Relationships between the dynamics of cortical blood flow, oxygenation, and volume changes following sensory stimulation. Proceedings of the National Academy of Sciences of the

United States of America. https://doi.org/10.1073/pnas.94.26.14826

Mandelkow, H., De Zwart, J. A., & Duyn, J. H. (2016). Linear discriminant analysis achieves high

classification accuracy for the BOLD fMRI response to naturalistic movie stimuli. Frontiers in Human

Neuroscience. https://doi.org/10.3389/fnhum.2016.00128

Experimental Social Psychology. https://doi.org/10.1016/0022-1031(76)90076-7

Newtson, D., Engquist, G. A., & Bois, J. (1977). The objective basis of behavior units. Journal of Personality

and Social Psychology. https://doi.org/10.1037/0022-3514.35.12.847

Nishiyori, R. (2016). fNIRS: An emergent method to document functional cortical activity during infant movements. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2016.00533

Pinti, P., Tachtsidis, I., Hamilton, A., Hirsch, J., Aichelburg, C., Gilbert, S., & Burgess, P. W. (2018). The present and future use of functional near-infrared spectroscopy (fNIRS) for cognitive neuroscience.

Annals of the New York Academy of Sciences. https://doi.org/10.1111/nyas.13948

Quaresima, V., & Ferrari, M. (2019). Functional Near-Infrared Spectroscopy (fNIRS) for Assessing Cerebral Cortex Function During Human Behavior in Natural/Social Situations: A Concise Review.

Organizational Research Methods. https://doi.org/10.1177/1094428116658959

Ranganath, C., & D’Esposito, M. (2005). Directing the mind’s eye: Prefrontal, inferior and medial temporal mechanisms for visual working memory. Current Opinion in Neurobiology.

https://doi.org/10.1016/j.conb.2005.03.017

Sassaroli, A., & Fantini, S. (2004). Comment on the modified Beer-Lambert law for scattering media. Physics

in Medicine and Biology. https://doi.org/10.1088/0031-9155/49/14/N07

Schwan, S., & Garsoffky, B. (2004). The cognitive representation of filmic event summaries. Applied

Cognitive Psychology. https://doi.org/10.1002/acp.940

Schwan, S., Garsoffky, B., & Hesse, F. W. (2000). Do film cuts facilitate the perceptual and cognitive organization of activity sequences? Memory and Cognition. https://doi.org/10.3758/BF03213801 Shen, F., & Morris, J. D. (2016). Decoding neural responses to emotion in television commercials: An

integrative study of self-reporting and fMRI measures. Journal of Advertising Research. https://doi.org/10.2501/JAR-2016-016

Silberstein, R. B., & Nield, G. E. (2008). Brain activity correlates of consumer brand choice shift associated with television advertising. International Journal of Advertising.

https://doi.org/10.2501/S0265048708080025

Simons, J. S., & Spiers, H. J. (2003). Prefrontal and medial temporal lobe interactions in long-term memory.

Nature Reviews Neuroscience. https://doi.org/10.1038/nrn1178

Strangman, G., Culver, J. P., Thompson, J. H., & Boas, D. A. (2002). A quantitative comparison of simultaneous BOLD fMRI and NIRS recordings during functional brain activation. NeuroImage. https://doi.org/10.1016/S1053-8119(02)91227-9

Swallow, K. M., Zacks, J. M., & Abrams, R. A. (2009). Event Boundaries in Perception Affect Memory Encoding and Updating. Journal of Experimental Psychology: General. https://doi.org/10.1037/a0015631 Zacks, J. M., Speer, N. K., Swallow, K. M., Braver, T. S., & Reynolds, J. R. (2007). Event perception: A

mind-brain perspective. Psychological Bulletin. https://doi.org/10.1037/0033-2909.133.2.273

Zacks, J. M., & Tversky, B. (2001). Event structure in perception and conception. Psychological Bulletin. https://doi.org/10.1037/0033-2909.127.1.3

Appendix 1:

Pictures of transmitter (Tx) and receiver (Rx) placement. The red marked channels were not used in analysis for the Receiver (Rx) and Transmitter (Tx) were too far apart (>30mm).

Right (with optode names) Left (with optode names)

Right (with optodes placed) Left (with optodes placed)

Appendix 2:

The entire paradigm in the bigger study. Figure 5 shows the TVC block only, because only data from this block was used this in the current paper.

Appendix 3:

List of TVCs used. The left column contains the version number. The right number the TVC-index in the following format: TVC-'brand'-'year'-'name'-'duration(sec)'

Version TVC-index 1 TVC-AppieToday-2015-BuikBillenBonus-77 1 TVC-Zalando-2012-Virus-27 1 TVC-NederlandseEnergieMaatschappij-2011-JohanDerksen-30 1 TVC-Hak-2013-RonaldKoeman-32 1 TVC-Blijdorp-2013-Olli-75 1 TVC-Heineken-2009-WalkInFridge-31 1 TVC-Bol-2011-Takelauto-25 1 TVC-Volkswagen-2011-OudVrouwtje-45 1 TVC-Eneco-2011-SamenGaanWeVoorDuurzaam-35 1 TVC-HertogJan-2016-Pilsener-24 1 TVC-Hema-2009-Rompertje-25 1 TVC-Robijn-2015-WasBijChantalJanzen-40 1 TVC-Philips-2014-Airfryer-25 1 TVC-CentraalBeheer-2012-Woonverzekering-65 1 TVC-Telfort-2013-Smartpakkers-44 1 TVC-DubbelFris-2013-MeisjeVsJongen-30 1 TVC-BelastingdienstDouane-ReizenApp-58 1 TVC-Fietsenwinkel-OnlineKopen-30 1 TVC-ASN-2016-Gewoontegedrag-35 1 TVC-NOCNSF-ZoDoenWeDat-38 1 TVC-Beslist-2015-Winterjas-30 1 TVC-KPN-2015-Beeldbellen-45 1 TVC-Marktplaats-2013-SpontaneVerkopen-25 1 TVC-DeBijenkorf-2017-VogelPauw-30 1 TVC-Gamma-2015-Verfwinkel-30 2 TVC-Essent-2018-Brand-40 2 TVC-Specsavers-ShouldveGoneTo-2014-30 2 TVC-HoyHoy-2014-MakeoverAart-35 2 TVC-Plus-2011-HollandsePrijsweken-40 2 TVC-McDonalds-2018-CadeauKalender-30 2 TVC-Gamma-2011-Lego-30

2 TVC-Ditzo-2012-JohndeWolf-35 2 TVC-KPN-2015-1eAppje-40 2 TVC-CentraalBeheer-1989-Klok-46 2 TVC-AlbertHeijn-2011-Appie-42 2 TVC-McDonalds-2014-Euroknaller-20 2 TVC-Skoda-2014-SkodaExperiment-36 2 TVC-IKEA-2016-Aandacht-60 2 TVC-Eneco-2014-ToonOmruil-40 2 TVC-OldAmsterdam-2015-Karakter-30 2 TVC-Pickwick-2010-Dutchblend-30 2 TVC-MasterCard-IgonedeJong-40 2 TVC-OHRA-2017-Viervoeters-45 2 TVC-SNS-2015-Motorcross-30 2 TVC-Chocomel-2016-ZoVersZoOP-20 2 TVC-Wildlands-KindTijger-25 2 TVC-Nuon-2012-EdEnEduardOverMijnNUON-45 2 TVC-DeFriesland-2017-BetereWereld-60 2 TVC-FBTO-2016-SchadeApp-30 2 TVC-ING-2014-WatGaatHetWorden-30 3 TVC-MiljoenenSpel-2012-PatriciaPaay-30 3 TVC-Robijn-HuizeGerschanowitz-40 3 TVC-Pricewise-PaulHaenen-30 3 TVC-Hunkemoller-2013-SylvieMeis-20 3 TVC-Yarden-2017-AliB-40 3 TVC-Bol-2010-Mummiepak-25 3 TVC-Jumbo-2018-Kerst-70 3 TVC-Calve-2010-Pietertje-41 3 TVC-Brand-2009-HetBierWaarLimburgTrotsOpIs-40 3 TVC-Defensie-2013-WerkenBijDefensieJeMoetHetMaarKunnen-35 3 TVC-DELA-2012-LeefVandaag-55 3 TVC-Heineken-TheHero-30 3 TVC-Opel-2013-ADAM-40 3 TVC-CentraalBeheer-2012-HetLaatsteBod-45 3 TVC-TMobile-2013-AliBZonderAnsjovis-40 3 TVC-Eneco-2012-Toon-32

3 TVC-Unox-2012-DeUnoxBoerenScharrelrookworst-35 3 TVC-Ziggo-2013-WifiSpots-30 3 TVC-Shell-AirMiles-34 3 TVC-ABNAmro-2017-Spaarverslimmers-45 3 TVC-Telfort-2012-LekkerLangBellen-48 3 TVC-Interpolis-2018-ThuisWacht-60 3 TVC-Knab-2015-AllesGeven-22 3 TVC-Tele2-2012-Bioscoop-35 3 TVC-BeslistNL-2015-Sportschoenen-30 4 TVC-Mentos-2017-SayHello-30 4 TVC-VanishOxiAction-2013-Vlekkenverwijderaar-45 4 TVC-Autodrop-2010-Vingerneus-30 4 TVC-STRATO-2014-Internet-30 4 TVC-Knorr-2014-WereldgerechtenBurritos-35 4 TVC-KNGF-2014-Buddyhond-30 4 TVC-Rolo-1996-Olifant-36 4 TVC-CentraalBeheer-2016-Rapper-60 4 TVC-TMobile-2011-AliBAltijdSamen-45 4 TVC-Andrelon-2014-OilAndCare-40 4 TVC-Essent-2013-ZekerDalen-30 4 TVC-Tele2-2015-OmdatHetKan-60 4 TVC-Coop-SamenMaakJeVerschil-45 4 TVC-Telfort-2013-Mobiel-54 4 TVC-Rabobank-2016-HypotheekBinnen1Week-30 4 TVC-PostNL-2012-Moeder-25 4 TVC-Nuon-NuonZonnepanelenHuren-40 4 TVC-Clipper-ManyReasons-20 4 TVC-Anderzorg-2016-DeLeven-20 4 TVC-Hak-2015-IlseDeLange-40 4 TVC-Eneco-2013-Toon4JarigContract-40 4 TVC-ABNAmro-2015-IntroductieTekst-50 4 TVC-Smint-2017-VIP-20 4 TVC-Airbnb-2015-BelongAnywhere-60 4 TVC-KarvanCevitam-2015-Go-30' 5 TVC-Zalando-2011-Naaktrecreatie-37

5 TVC-Volkswagen-2013-VolkswagenHond-51 5 TVC-BecamFinancieringen-Bouwvakker-20 5 TVC-Bonprix-2013-Cafe-30 5 TVC-AlbertHeijn-2015-Afscheid-80 5 TVC-Bol-2014-Flappie-30 5 TVC-Campina-2013-DieKomtErWel-45 5 TVC-Heineken-2002-DerRudi-40 5 TVC-Marktplaats-2015-GaErvoor-30 5 TVC-Jumbo-2014-Moestuin-55 5 TVC-Unox-2012-KnaksKinderfeestje-25 5 TVC-BeterBed-EdithBosch-20 5 TVC-Independer-2013-Vergelijken-26 5 TVC-Simyo-2014-Vriendendeal-44 5 TVC-Essent-2013-ThermostaatVanEssent-30 5 TVC-McDonalds-2017-Maestro-60 5 TVC-CarNext-EveryDetailMatters-40 5 TVC-Plus-AngryBirds-20 5 TVC-Eneco-2014-HollandseWindorigami-70 5 TVC-Flexa-Kleurtester-20 5 TVC-Hak-2016-BonenHermandenBlijker-25 5 TVC-TempoTeam-2015-TeamUp-20 5 TVC-Granditalia-2014-pastamaestro-26 5 TVC-Interpolis-2018-ThuismeesterAppREV-62 5 TVC-BeslistNL-2016-Angry-32 6 TVC-BakkerBart-2010-Krentenbolletjes-40 6 TVC-Yarden-2015-AdelheidRoosen-40 6 TVC-EyeloveBrillen-2018-ReneFroger 6 TVC-Lidl-2014-GerardJoling-45 6 TVC-KleneDrop-2017-Krakers-20 6 TVC-CentraalBeheer-2010-Muis-45 6 TVC-Gamma-2013-Baby-35 6 TVC-Fiat-1983-Uitlachen-30 6 TVC-AmstelRadler-2015-AlcoholVrij-45 6 TVC-Ford-2017-WelcomeHome-60 6 TVC-Mona-2010-MonaXLDaarWordJeBlijVan-35

6 TVC-HertogJan-Bastaard-25 6 TVC-PostNL-2017-Lippenstiftkus-22 6 TVC-Telfort-2012-AllesInEen-40 6 TVC-Jumbo-2015-BoodschappenGratis-60 6 TVC-CentraalBeheer-2013-MeerVerzekerdDanUDenkt-52 6 TVC-Eneco-2016-Toon-25 6 TVC-Tele2-2016-HappyDance-50 6 TVC-Robijn-2010-DoetDeWasBij-IlseDeLange-38 6 TVC-Videoland-OnDemand-30 6 TVC-Interpolis-FocusAutomodus-30 6 TVC-ABNAmro-2015-Thuis-45 6 TVC-Grolsch-2015-400JaarKarakter-60 6 TVC-Nuon-2017-LuisterenGeeftEnergie-30 6 TVC-Obvion-2014-Andersdenken-30 Appendix 4:

fNIRS Channels to Location. Data obtained by the red marked channels was considered unusable for the receiver (Rx) and transmitter (Tx) were too far apart (>30mm).

fNIRS Channel Location

1 Rx1 - Tx1 HHb 2 Rx1 - Tx1 O2Hb 3 Rx1 - Tx2 HHb 4 Rx1 - Tx2 O2Hb 5 Rx1 - Tx3 HHb 6 Rx1 - Tx3 O2Hb 7 Rx2 - Tx1 HHb 8 Rx2 - Tx1 O2Hb 9 Rx2 - Tx3 HHb 10 Rx2 - Tx3 O2Hb 11 Rx2 - Tx4 HHb 12 Rx2 - Tx4 O2Hb 13 Rx3 - Tx2 HHb 14 Rx3 - Tx2 O2Hb 15 Rx3 - Tx3 HHb 16 Rx3 - Tx3 O2Hb 17 Rx4 - Tx3 HHb 18 Rx4 - Tx3 O2Hb 19 Rx4 - Tx4 HHb 20 Rx4 - Tx4 O2Hb 21 Rx3 - Tx5 Hhb 22 Rx3 - Tx5 O2Hb 23 Rx4 - Tx5 HHb 24 Rx4 - Tx5 O2Hb 25 Rx5 - Tx7 HHb 26 Rx5 - Tx7 O2Hb 27 Rx5 - Tx6 HHb 28 Rx5 - Tx6 O2Hb 29 Rx5 - Tx8 HHb 30 Rx5 - Tx8 O2Hb 31 Rx7 - Tx7 HHb 32 Rx7 - Tx7 O2Hb 33 Rx7 - Tx8 HHb

34 Rx7- Tx8 O2Hb 35 Rx7 - Tx10 HHb 36 Rx7 - Tx10 O2Hb 37 Rx6 - Tx6 HHb 38 Rx6 - Tx6 O2Hb 39 Rx6 - Tx8 HHb 40 Rx6 - Tx8 O2Hb 41 Rx8 - Tx8 HHb 42 Rx8 - Tx8 O2Hb 43 Rx8 - Tx10 HHb 44 Rx8 - Tx10 O2Hb 45 Rx6 - Tx9 HHb 46 Rx6 - Tx9 HHb 47 Rx8 - Tx9 HHb 48 Rx8 - Tx9 O2Hb