5.8 RESULTS, INTERPRETATION AND DISCUSSION
5.8.24 TREND ON CORRELATION OF TTH DURATION WITH OCCURRENCE
This analysis was performed to establish the correlation between the frequencies of occurrence of TTH pain with its duration of stay (Fig. 5.33).
Representation
1. - - dotted line Median.
2. –Average (continuous line).
0 0
Avg. duration
Avg. frequency Avg. frequency Avg. frequency Avg. frequency Avg. frequency BaseLine
Median Median Median Median Median
Median Median Average
Median Median Median Median
1 month 3 months
Period
6 months 12 months
5 10 15 20 25
5 10 15 0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15
Techniques EMGa EMGav EMGv
FIG. 5.33
Correlation of TTH: duration with occurrence.
142 CHAPTER 5 CHRONIC TTH ANALYSIS BY EMG AND GSR BIOFEEDBACK
3. Bubbles (O)—Average of all subjects.
4. + individual subject plot.
The baseline data for the EMGav, EMGv, and EMGa was more toward the average of frequency and duration with a few exceptions for the subjects where the data lay in the high duration and high fre- quency zone. Initially in all the groups, the subjects or their averages were in the zone of high occur- rence of TTH for higher durations.
The baseline data for EMGav was mostly in the high duration and high frequency zone, showing that the subject group consisted of individuals suffering from the most frequently occurring severe pain.
After applying the different trend models such as linear, logarithmic, exponential, polynomial, and power model, we found the best fitted trend in the logarithmic model. Hence the logarithmic model trend was analyzed. The mathematical modeling of the logarithmic model is given as follows:
(Tables 5.23–5.25).
EMGv:After applying the different therapies, the data started moving toward the quartile of low duration and low frequency. This trend continued and, at the end of the year, the majority of data came under the zone of low frequency and low duration with a few exceptions for four subjects in the range of higher frequency or duration, which proves EMGv was not as efficient as the other therapies.
EMGav:There was high rate of convergence of data toward the lower quartile of low duration and low frequency in the initial months and this continued until the end of the year. The data came under the average values for most of the subjects.
Table 5.23 Trend Lines Model
Model formula PeriodTechniques(ln(Avg. Frequency) + intercept) Number of modeled observations 349
Number of filtered observations 61
Model degrees of freedom 30
Residual degrees of freedom (DF) 319
SSE (sum squared error) 4589.77
MSE (mean squared error) 14.388
R-Squared 0.500008
Standard error 3.79315
P-Value (significance) <.0001
A linear trend model is computed for average of duration given natural log of average of frequency. The model may be significant at P.05. The factor period may be significant atP.05.
Table 5.24 Analysis of Variance
Field DF SSE MSE F P-Value
Period 24 2472.0111 103 7.15877 <.0001
Techniques 20 762.25159 38.1126 2.64891 .000196
143 5.8 RESULTS, INTERPRETATION AND DISCUSSION
Table 5.25 Individual Trend Lines
Panes Color Line Coefficients
Row Column Techniques P-Value DF Term Value StdErr t-Value P-Value
Duration BaseLine EMGv .281848 26 ln(Avg. Frequency) 4.39243 3.99682 1.09898 .281848
intercept 2.63734 8.03539 0.328216 .745377
Duration BaseLine EMGav .0011665 25 ln(Avg. Frequency) 6.40729 1.74851 3.66444 .0011665
intercept 1.01322 3.45925 0.292901 .772015
Duration BaseLine EMGa .650741 25 ln(Avg. Frequency) 1.53178 3.34276 0.458237 .650741
intercept 12.1876 6.46665 1.88469 .0711494
Duration 1 month EMGv .0034553 23 ln(Avg. Frequency) 4.07031 1.24899 3.25889 .0034553
intercept 0.96489 2.25989 0.426964 .673376
Duration 1 month EMGav .0031575 21 ln(Avg. Frequency) 6.75871 2.02788 3.3329 .0031575
intercept 2.70307 3.99063 0.677354 .505576
Duration 1 month EMGa .331464 25 ln(Avg. Frequency) 2.64055 2.66612 0.99041 .331464
intercept 7.0711 4.69983 1.50454 .144971
Duration 3 months EMGv .0025702 23 ln(Avg. Frequency) 2.6856 0.794167 3.38165 .0025702
intercept 2.21206 1.40782 1.57126 .129779
Duration 3 months EMGav .0191972 21 ln(Avg. Frequency) 2.95765 1.16602 2.53654 .0191972
intercept 2.9906 2.11083 1.41679 .171209
Duration 3 months EMGa .0463431 21 ln(Avg. Frequency) 4.12013 1.94597 2.11726 .0463431
intercept 15.6719 3.34793 4.68107 .0001277
Duration 6 months EMGv .30771 22 ln(Avg. Frequency) 0.853519 0.817353 1.04425 .30771
intercept 4.02216 1.23483 3.25726 .0036096
At the end of the year, the majority of the data came under the zone of low frequency and low duration with exceptions for two subjects in the range of higher frequency or duration, which proves that this therapy is less efficient than the other therapies.
Further analysis of subjects suffering with chronic TTH showed that at the end of the year, the majority of the data came under the zone of low frequency and low duration with the exception of only one subject in the range of higher frequency or duration.
It has been found that EMGa converged the most diverged data more effectively than that of EMGv and EMGav.
It is clear from the analysis that three confounding factors, i.e., intensity, duration, and frequency, greatly influence the characteristics of the EMG and GSR signals and thus the performance of pattern recognition systems. A massive amount of information is necessary to encapsulate and describe the complexity and variability of surface EMG and GSR signals. To translate the vast and complex infor- mation in EMG and GSR signals into useful control signals for prosthetic devices for identifying neuromuscular diseases, data storing, and sharing, big data are needed.
The IoT can help in remote patient monitoring of subjects with chronic or long-term stress. It can help in tracking the subject’s medication orders and the location of subjects admitted to hospital or under treatment, and send information to caregivers. With the help of big data and IoT, EMG and GSR data have been made available online and there are now at least 33 datasets with surface EMG collected from 662 subject sessions[32].