Vector of Excursion: Waveform vs Sunnybrook

This section represents the waveform -parameter results for both movements, smile and eyebrow lift. As shown in Table 5.1 and discussed already in Subsection 4.4.3, the waveform investigates the direction of the vector of excursion. The vector of excursion is found to be a crucial factor when a human is detecting asymmetry from another human’s face, as detailed in Subsection 2.4.2. The waveform results are compared against the current clinical measure, Sunnybrook Facial Grading Scale, described in Subsection 2.3.3.

Figure 5.6 shows the smile movement’s cross-correlation results over all pa-tients and against Sunnybrook results. The figure has three lines of different pink shades. Each shade represents a channel from the prototype’s midextension; channel 4 is the most central one, channel 5 is lateral to the channel 4, and channel 6 the most lateral sensor of the mentioned ones. Each line is computed by comparing the contralateral channel-pairs by computing cross-correlation for every repetition, and over all 17 patients. For every repetition, the highest coefficient is extracted. The value shown in Figure 5.6 is the average value of the specific patient’s ten repetitions’

highest coefficients. Every value is visualized; in other words, no outliers are excluded.

In the same Figure 5.6 there are three lines or dashed lines representing the Sunny-brook values for the test patients. The figure is organized according to increasing Sunnybrook total score, that translates to increased health in respect to facial palsy.

The Sunnybrook scores are scaled between 0 and 1 for visualization purposes; the score is originally between 0 and 100. Sunnybrook is an unweighted regional scale, and the total score is visualized with continuous blue line with label "SB total". The total score contains the resting symmetry and the dynamic movements’ symmetry by regions, and the synkinesis evaluation. As only dynamic movements are measured by the prototype, and synkinesis may be detected depending on the location and intensity, the "SB voluntary" score is plotted as well. It is the score of the dynamic voluntary movements. Finally, as the channels picked for this figure represent smile, the "SB smile" curve visualizes the smile score that is embedded in the "SB voluntary"

score. To summarize, the level of generality decreases from SB total to SB vountary, and again to SB smile. It is acknowledged, that the Sunnybrook scale is designed to be observed via the total score. The partial scores are visualized here as a curiosity.

Figure 5.6Highest cross-correlation coefficients for smile movement over all patients.

Similarly, Figure 5.7 visualizes the eyebrow lift movement’s cross-correlation results over all patients and against Sunnybrook results. The results are computed in the same manner as smile results in Figure 5.6 except that the channels are now from the upmost extension; channel 1 is the most central channel and channel 2 is its lateral neighbor. Also, one Sunnybrook result differs from smile results; the "SB EBL" which is the Sunnybrook score for the eyebrow lift movement. That value corresponds to "SB smile" in Figure 5.6.

Figure 5.7 Highest cross-correlation coefficients for eyebrow lift movement over all patients.

Figures 5.6 and 5.7 show the prototype measurement results againt the relevant Sunnybrook scores over all patients and in the decreasing order of palsy level. In order to evaluate how well the prototype results correlate with the current clinical method we can inspect the two aforementioned figures. Figure 5.6 shows that the patient with the most severe palsy grading with Sunnybrook (patient number 12) has lower prototype results in all channels than the healthiest patient (patient number 17). However, the patients between these extremities have fluctuating pro-totype results. This is true for the eyebrow lift figure, too. Therefore, with naked eye it is challenging to evaluate the entity; how well do the prototype results and Sunnybrook scores correspond? To offer necessary insight to compare the prototype and Sunnybrook results, Figures 5.8-5.11 are provided.

Figure 5.8 visualizes the prototype smile results for the channels 4-6 against Sunnybrook scores. The data in the figure is same as in Figure 5.6 but in different format; here, apair plotis used. On x-axis, there is a slot for each relevant channel, and on y-axis there is a position for the picked Sunnybrook scores. The most general Sunnybrook score, the total, is on the upmost row whereas the most specific Sunny-brook value is at the lowermost row; the generality level decreases when going down the y-axis. The pair plot then scatters the prototype results against Sunnybrook values; each channel against every Sunnybrook score.

Figure 5.8 Pair plot of highest smile cross-correlation coefficients against Sunnybrook values.

Figure 5.9 offers even higher level representation of the data; it lays out the correlation between the prototype measurements and Sunnybrook data. The figure is aheatmap where each element of the matrix shape format correspond to the previous pair plot.

The darker color indicates increased correlation level. The correlation coefficient is also numerically visible in the heatmap. If inspected channelwise, the channel 4 data shows weak to moderate correlation with the Sunnybrook data. Channel 5 measured values do not correlate, or correlate moderately. Channel 6 measured values correlate on moderate level. Each channel correlates stronger with the most general Sunnybrook score, the SB total score. When the generality of the Sunnybrook score decreases, so does the correlation coefficient. In general, channel 4 has the highest correlation with Sunnybrook values, and its neighbor channel 5 the smallest.

Figure 5.9 Heatmap showing the correlation of prototype results and the Sunnybrook values for smile movement.

Similarly, Figures 5.10 and 5.11 visualize the correlation of prototype eyebrow lift results against Sunnybrook scores. The following figures are ordered in the same manner as the two previous ones for smile results. The differences are the choice of channels, for eyebrow lift the channels 1 and 2 are used, and the most detailed Sunnybrook score is the eyebrow lift score instead of the smile score.

Figure 5.10 Pair plot of highest eyebrow lift cross-correlation coefficients against Sunnybrook values.

Figure 5.11Heatmap showing the correlation of prototype results and the Sunnybrook values for eyebrow lift movement.

Reading the eyebrow lift heatmap from Figure 5.11 channelwise, it can be seen that for channel 1 the correlation to Sunnybrook scores are moderate. For channel 2 the correlation between the prototype data and Sunnybrook values are moderate, or even close to strong. Thus, channel 2 has slightly higher correlation with the Sunnybrook values than channel 1. The pair plot in Figure 5.10 reveals that for several patients (12 out of 17) the eyebrow lift score is the same despite the different Sunnybrook total score. This flat line shows in the correlation heat map too; the lowermost row of the heatmap has the weakest correlation. The more general Sunny-brook values, the voluntary movement and total score, have higher correlation with the protype data.

To summarize the waveform or direction of vector of excursion results over all patients, there is weak to moderate correlation between the measured data’s wave-form and Sunnybrook values for smile movements. The correlation increases with the generality of the Sunnybrook scale. The most central channel, channel number 4, has the highest correlation. For eyebrow lift movements the correlation is moderate.

The most specific Sunnybrook value has flat line in the eyebrow lift pair plot that shows as the lowest correlation in the heatmap as well.