Questionnaire and Projective Color Association in Physiological Responses to Different Emotional Charge

Procedure The focus and procedure of the study will be explained to all participants. Their physiological responses (cerebral blood flow, brain electrical activity, heart rate variability, and skin conductance response) to visual word and picture stimuli will be measured in the Multimodal and Functional Imaging Laboratory (MAFIL). The measurement lasts about 35 minutes, total time spent in MAFIL is approx. 60 minutes. After the measurement of physiological responses, participants are asked to complete the online CA method application and the online questionnaire within the next 24 hours. Both the CA method application and the online questionnaire are focused on capturing the participants' perception of word and picture stimuli using different measurement methods, i.e., the colour association and questioning method. On average, this task takes approximately 20 minutes for each method. It should be noted that the different time period for measuring physiological reactions to and the perception of stimuli by participants is given by necessity to capture only brain activation connected with the perception of the stimuli, and not with answering online questionnaire and CA method application.

In the case of all kinds of measurement, i.e. physiological responses, CA method, and questionnaire, the respondents are exposed to the same 145 stimuli. They are 115 words (three sets of 35 words with positive, neutral, and negative content) and 30 pictures (two sets of 15 photos with positive and negative content). Combining stimuli with different content ensures greater variability of reactions, which supports the robustness and relevancy of the results. The content of the stimuli was evaluated according to the national norm of the CA method for words and the International Affective Picture System for pictures. It should be noted that participants are exposed to 21 stimuli twice during the measurement of physiological responses to check the stability of physiological reactions. The order of the stimuli is identical for all participants, and the stimuli combine words and pictures regardless of their content.

Participants The sample will consist of 101 individuals aged 18-64 years, of whom 23 will repeat the entire procedure twice to allow the robustness check of the results in time. Participants will be recruited by a market and media research agency.

Measurement Physiological responses In the study, physiological responses to word and pictorial stimuli will be measured and compared. Four measurement modalities will be used to record data: functional magnetic resonance imaging (fMRI), surface skulk electroencephalography EEG, heart rate activity, and galvanic skin response (GSR) representing electrodermal activity of the body.

Functional magnetic resonance imaging (fMRI) will be used for the quantification of cerebral blood flow. Measurement will be carried out with Siemens Prisma 3T MRI scanner (Siemens Medical Solutions USA, Inc.) with 64 channels head-neck coil, Syngo version VE11c, accompanied by standard Anatomical T1 MPRAGE scan. Multi-echo MB-EPI fMRI acquisition based on CMRR EPI sequences with TR=700 ms and TE=15/34/53 ms at 1570 scans. GRAPPA PAT factor 2 and slice acceleration (MB factor) set to 6 are used. The dimension of the recorded voxel is 3x3x3 mm with a slice thickness of 3 mm. The number of axial slices is 48 (transverse) in the inplane FoV=192 mm and the setting of the flip angle is 45°.

Time-frequency analysis will be performed from indirect measurements of neural EEG activity. The measured parameter will be the power of oscillations in each frequency band in the scalp EEG measured in each sensor. Non-invasive brain electrical activity will be measured by high density EGI using Net Amps GES400 series amplifier and 256 HydroCel Geodesic Sensor Net (GSN) on the skull surface with 1kHz sampling frequency.

Heart rate variability (HRV) will be measured from an indirect measurement of the response of the autonomic nervous system. The measured parameter will be the period of the cardiac cycle read from the ECG curve. ECG data will be acquired as a supplementary channel during EEG data acquisition (EGI system) with 1 kHz sampling frequency. The ECG channel will be measured as a bipolar lead with one electrode placed under the left clavicle and the other electrode on the left chest side.

Electrodermal activity will be measured as an indirect measurement of the response of the sympathetic autonomic nervous system. The measured parameter will be the galvanic skin response (GSR). GSR was measured between the index and middle fingers of the dominant hand, using the bipolar BrainAmp ExG MR (Brain Products GmbH) with a sampling frequency of 5kHz.

An MRI compatible LCD monitor (BOLD screen MR 24 from the Cambridge research system) will be used as an interface to display word and pictorial stimuli during the measurement of physiological responses to them. The subject will see the image on the MRI scanner due to a mirror mounted above the head coil holes supplied by the MRI scanner manufacturer. Each of the 145 stimuli will be displayed for 5 seconds to the subject under investigation.

Pre-processing of physiological data For subsequent statistical processing, the measured data will be adjusted and pre-processed. It makes it possible to use them for the evaluation of the potential of the questionnaire and the CA method to reflect the physiological and neurophysiological responses of the organism to stimuli.

The measured fMRI data will be realigned (motion correction) according to the middle echo signal. Optimally, multi-echo (ME) images using weighted average based on tSNR (Temporal Signal-to-Noise Ratio) and TE (Echo Time) will be combined. Spatial normalization is adopted in the standard anatomical template (MNI space) and spatial smoothing is performed with the Gaussian kernel, FWHM=5mm. The quality check will be performed by movement analysis, based on the frame-wise displacement metric. In addition, a positive valid data check is performed by spatial coverage of the brain based on the mask explorer 2.12 tool. Pre-processing is performed in the SPM12 software.

EEG data represent a time-frequency decomposition of the data (time-frequency analysis) separately for each electrode. The data will be 3-dimensional, where the time dimension has boundaries of 0-5s, the frequency dimension has boundaries of 4-40 Hz, and the third dimension is represented by individual sensors. First, considering the amount of data, compression of the EEG recordings will be used. The frequency dimension will be divided into 33 bins (logarithmic) and the time axis is divided into 5 bins (1 second each). The third dimen-sion will not be compressed. Thus, each cell will contain a number that is indicative of a particular 1s segment in a given electrode, for a given narrow frequency band. EEG pre-processing will be performed in MATLAB software, Brain Vision Analyzer 2.0 toolbox. Gradient and pulse artifacts will be removed using the IIR Butherworth band-pass at 1-40 Hz. Additionally, the ocular artifacts will be removed using the ICA decomposition corresponding to the ocular artifact. Subsequently, back-reconstruction of the signal without artifactual components will be per-formed. The recordings will be segmented according to the moment of stimulus appearance with segment length of 5 seconds (segment 0 to 5 seconds relative to the stimulus). The 52 sensors will be cut from the analysis because the sensors are located on the cheeks, just above the eyes, around the ears, and the last row on the neck. For each segment, the time frequency analysis will be calculated using the cwt function in MATLAB R2017a and the absolute value of the Cont coefficients is taken. The total signal power (Ptot) will be calculated for each segment. Segments that have Ptot>1.5*median(Ptot) will be marked as artifacts. The threshold will be set arbitrarily to reasonably filter out segments with poor signal. The average will be calculated over each second, which corresponds to matrix compression in the time domain. The result will be a matrix for each segment that is 33x5x204x166, where there are 33 frequency bins, 5-time bins, 204 electrodes, and 166 stimuli.

The ECG data will be preprocessed in Brain Vision Analyzer 2.0 (Brain Products) for MATLAB software. Gradient artifacts will be denoised using IIR 1-20Hz bandpass filtering with semi-automatic R-wave detection. Data are segmented according to the onset of stimulus occurrence. The segment length will be 6 s. The average length of the R-R interval (hereafter R-R) will be calculated for each segment. For each proband separately, a correction for the R-R drift will be performed during the task. A third-order polynomial function is modelled on the R-R wave-form, and the estimated slow change trend will be subtracted from it. Subsequently, R-R will be converted to heart rate (HR) in units of pulses per minute.

Galvanic skin response data will be also pre-processed in MATLAB software, LEDALAB Toolbox (http://ledalab.de). Gradient artifacts and residual gradient artifacts will be removed using a median filter with a complementary IIR pass filter with a cut-off frequency of 1Hz (median over 20 samples).

CA method The CA method (Colour Association Method) is a combined projective technique based on the principles of two recognized psychological concepts: the Lüscher´s colour test and word associations. The CA method links the knowledge of the basic principles of associations in human consciousness with the measurement of colour preferences and combines their benefits. Focusing on association mechanisms that are almost identical across individuals allows its use for all individuals regardless of the level of their knowledge or rational thinking.

Questionnaire Finally, participants will be asked to express their perception of word and pictorial stimuli using a questionnaire. For each of the 145 stimuli, they will be asked to 'evaluate how the word / image [displayed] affects you using the numerical scale 1-7, where the number 1 represents a completely negative perception and the number 7 represents a completely positive perception'.

Statistical analysis In this study, two main approaches to statistical analysis will be applied. General linear models will be used to analyse the MRI data (using the SPM12 software). The models explore brain activity depending on the positivity of the stimuli approximated by questionnaire responses and CA method categories. Group analyses will be calculated with SPM12 random effect models controlling for gender and age. One sample t-test, paired t-test, and flexible factorial model (ANOVA) will be used to evaluate the data according to individual goals. A threshold level of p=0.05 (FEW corrected) will be used for all results.

Statistical analysis of the relationship between CA method, questionnaire categories, heart rate, and skin conductance will be performed on aggregated data. First, the mean heart rate and skin conductance will be calculated for the categories of the most negative, neutral, and positive stimuli for each individual. Subsequently, these aggregated values will be analysed by the nonparametric Kruskal-Wallis rank test and Dunn's test of multiple comparison in order to reveal whether/which categories embody statistically significant differences com-pared to others. These tests will be used in the case of the ordinal character of the data or the violation of assumptions for parametric tests; otherwise, ANOVA and Turkey test will be used. The same procedure will be used for the analysis of the mutual relationship between the CA method and the questionnaire. Moreover, this analysis will be further replenished by correlation analysis employing the Pearson or Spearman correlation coefficient depending on the characteristics of analysed data.