概要

Calculating Heart Rate Variability from ECG Data from Youth with Cerebral Palsy During Active Video Game Sessions

Published: June 05, 2019
doi:

概要

This protocol describes a method for calculating Heart Rate Variability (HRV) from electrocardiogram (ECG) waveforms. Waveforms from continuous heart rate (HR) recordings during active video game (AVG) sessions were used to measure the aerobic performance of youth with cerebral palsy (CP).

Abstract

The aim of this study was to generate a method for calculating heart rate variability (HRV) from electrocardiogram (ECG) waveforms. The waveforms were recorded by a HR monitor that participants (youth with cerebral palsy (CP)) wore during active video game (AVG) sessions. The AVG sessions were designed to promote physical activity and fitness (aerobic performance) in participants. The goal was to evaluate the feasibility of AVGs as a physical therapy (PT) intervention strategy. The maximum HR (mHR) was determined for each participant and the Target Heart Rate Zone (THRZ) was calculated for each of three exercise phases in the 20 min AVG session: (warm-up at 40-60% mHR, conditioning at 60-80% mHR, and cool down at 40-60% mHR). Each participant played three 20 min games during the AVG session. All games were played while sitting on a bench because many youth with CP cannot stand for extended periods of time. Each game condition differed with participants using hand icons only, hand and feet icons together or feet icons only to collect objects. The objective of the game (called KOLLECT) is to collect objects to gain points and avoid hazards to not lose points. Hazards were used in the warm-up and cool down phases only to promote slower, controlled movement to maintain HR in the target heart rate zone (THRZ). There were no hazards in the conditioning phase to promote higher levels and more intense physical activity. Analytic methods were used to generate HRV (selected time-domain and frequency-domain measures) from ECG data to examine aerobic workload. Recent applications of HRV indicate that short-term measurements (5 min bouts) are appropriate and that HRV biofeedback may help improve symptoms and the quality of life in a variety of health conditions. Although HR is a well-accepted clinical measure to examine aerobic performance and intensity in PT interventions, HRV may provide information of the autonomic system functions, recovery and adaptation during AVG sessions.

Introduction

Cerebral palsy (CP) is the most common physical disability of childhood1. CP is caused by a neurologic insult to the developing brain and is associated with motor impairments such as muscle weakness, spasticity, deconditioning, and decreased motor control and balance2,3. CP is a non-progressive condition but with age, children become less physically active and more sedentary compared to their peers with typical development (TD) mostly because of the increased demands of growth on their compromised neuromuscular and musculoskeletal systems4.

Youth with CP usually receive physical therapy (PT) services to improve functional mobility and promote physical activity and fitness (e.g. aerobic and muscular endurance)2. Oftentimes, there is limited access to PT services and community resources to achieve and sustain these PT goals5,6.Active video games (AVGs) may be a feasible strategy in activity-based PT interventions in clinic, home or community settings7,8. Commercial AVGs have limited flexibility to adapt game play and meet the specific needs and PT goals for youth with CP9. However, customized AVGs provide flexible gaming parameters to challenge youth with CP while promoting physical activity and fitness10.

Our team has developed a customized AVG (called KOLLECT) to examine youth exercise responses (e.g., physical activity and aerobic fitness). The game uses a motion sensor to track youth motion during game play. The goal of the game is to 'collect' as many objects as possible for a high score and to avoid the hazards to avoid losing points. Objects may be collected with hand and/or feet icons as determined by the therapist in the flexible game parameters.

Designing activity-based PT interventions that dose physical activity intensity to promote aerobic fitness is critical for youth with CP11. Custom AVGs may be an effective strategy to dose intensity and engage youth in physical activity to promote fitness10. Heart rate (HR) monitors are often used in clinical PT practice to determine aerobic performance and activity intensity. Therefore, HR monitors will help determine feasibility of AVGs in dosing physical activity intensity to promote aerobic fitness9. ECG data generated from a HR monitor can be used to calculate heart rate variability (HRV). Analytic methods were used to generate HRV from ECG data to examine aerobic workload. Recent applications of HRV indicate that short-term measurements (5 min bouts) are appropriate and that HRV biofeedback may help improve symptoms and the quality of life in a variety of health conditions32,33,34. The application of short-term HRV measures is an appropriate means of assessing cardiovascular function during AVG sessions. Given that HRV is derived from the R-R interval of an ECG, we used selected time-domain and frequency-domain measures. Time-domain measure of HRV quantify the amount of variablility in the interbeat intervals which represents the time between successive heartbeats. We used the AVNN (average NN interval), RMSSD (root mean square of successive differences), SDNN (standard deviation of NN interval), NN50 (number of NN intervals >50 ms) and PNN50 (percentage of NN intervals). Frequency domain measures estimate the distributionof absolute or relative power into possibly four frequency bands, we specifically addressed on two bands, low frequency (LF) power and high frequency (HF) power along with the LF/HF ratio. Although HR is a well-accepted clinical measure, HRV may be useful because it provides information about autonomic system function, recovery, adaptation, and provides an estimate of aerobic workload during an AVG session28.

The purpose of this study was to examine the feasibility of using AVG strategies to promote physical activity and fitness. A second purpose was to present the AVG data collection protocol and the methodology to calculate HRV from ECG data obtained via a HR monitor. These measures and this protocol may prove relevant to clinicians to monitor and dose PT intervention sessions.

Protocol

  Institutional Review Board approval was obtained. All youth provided written assent and parents provided consent prior to participation.

1. AVG data collection sessions

  1. The AVG game session
    1. In this study, have youth with CP participate in an AVG session which is comprised of three 20 min games. See Table 5 for Youth Demographics. It was expected that a total of 30 games would be played; however, 29 games were completed because one subject only played 2 games in his AVG session.
    2. Have the subjects wear a HR monitor throughout the session to record HR and ECG responses.
    3. In the AVG session, have youth play each AVG while seated on a bench with feet flat on the floor and knees and hips flexed to 90 degrees (90/90 sitting) for postural support and stability.
    4. Use the following three gaming conditions for collection objects: 1) hand icons only; 2) feet icons only; and 3) both hand and feet icons. Use a counterbalanced order between subjects. Choose these three conditions to determine which is more effective in promoting physical activity and fitness and not too demanding to cause early, undue fatigue.
      NOTE: Each game was designed using the phases of exercise prescription: warm-up, conditioning and cool-down. [Please see Table 1]. Additionally, there was a rest phase before game play began to document baseline HR and a recovery phase after game play to document time to return to baseline HR.
    5. Allow subjects a rest period between games for HR to return to baseline level.
  2. Calculating HRV from ECG Data
    1. Organize data into 5 min time intervals to ensure comparable data for each phase. Therefore, there were 6 phases defined for these calculations: 1) Rest; 2) Warm-up; 3) Conditioning 1 (first 5 min); 4) Conditioning 2 (second 5 min); 5) Cool-Down (5 min) and 6) Recovery. Dividing the conditioning phase into two 5 min phases allows examination of subject aerobic performance in shorter intervals to account for fatigue due to deconditioning 12 (Table 4).
    2. To properly calculate HRV measures for each segment of a subject's session, perform R-peak detection on the raw ECG signal12,13. Use the raw signal to avoid manipulations that could skew the data.
    3. To process the data, obtain the start times of each recording session and convert from 'datetime' variables (MM/DD/YYYY HH:MM:SS.SS) to seconds. None of the sessions occurred across two days which allowed the MM/DD/YYYY portion to be ignored during these calculations. Acquire the start time of the game of interest from the timing table to locate each game session within the electrocardiogram (ECG) file; this time was converted to seconds after it had been extracted from the timing file. The timing file contained start times for each phase of the game as well as the end of the recovery period (Table 2).
    4. Calculate the rest period as the 5 min prior to the game start and the Recovery Phase as the 5 min after the end of the Cooldown Phase. Once these times were obtained, obtain the location (S) of the game phase of interest within the ECG file through the following equation:
      Equation 1  (1)
      where Phase is set to either Rest, Warmup, Conditioning 1, Conditioning 2, Cooldown, or Recovery; the time was divided by 1/Frequency to account for the ECG sample rate. The HR monitor had a sampling rate of 250 Hz and therefore contained a measure every 4 ms.
      1. Change this number by altering the sampling rate with the first prompt from the Peak_Detection.m program to account for the use of alternate recording devices. Choose which 5 min segment to work with while running the peak detection program. This was done via a prompt to the user. Set the end time to 5 min after the start time and take the frequency of the recording device into consideration.
    5. Once the 5 min section had been chosen, calculate a threshold for peak-detection based upon the average and standard deviation of the waveform.
      1. Set the threshold as Equation 2but this can be increased in the program if the data are uniform to reduce false-positive detection from T peaks which are higher than their corresponding R peaks. Examples of these false positives can be seen in Figure 1.
      2. Along with a minimum height for the R peak, assign a minimum distance between peaks to minimize the detection of incorrect peaks around the desired R. Set this value to 75 which corresponded to 0.3 s between peaks or 200 beats per min (bpm) (this value changes with frequency). The value of 200 bpm is higher than any HR achieved by the subjects in this study and can be changed based on the population being studied.
    6. Once the threshold was calculated, let the program run through the waveform and attempt to discern all the R's for RR interval and HRV calculations. Generate a preliminary plot so that the user could review it for irregularities such as those shown in Figure 1 or Figure 2.
      1. Correct these irregularities manually by editing the Detection variable which contains the microvolt (µV) reading of the peak in column 1 and the location in the current game session (s/0.004) in the second column. In most cases the proper R peaks can easily be found by zooming into the problem location as seen in Figure 1. Many data sessions are fairly uniform as shown in Figure 3 and will therefore only require a few corrections. Some cases, however are fairly messy and require more time to review and obtain proper R locations.
      2. If the fluctuations in the waveform make it excessively difficult to properly locate a peak, ignore small segments ~1-2 s and attribute to ectopic beats which are not used in HRV calculations12.
    7. After the R's have been located, run the HRV_Measures program. Calculate RR intervals first as they are the basis of the HRV measures used in this study12.
      1. Obtain a matrix of intervals and ignore any interval greater than 1.5 s (40 bpm) as it was due to the aforementioned ectopic beats being removed from the calculations. Save these RR intervals for further calculations and verification of data. Use these intervals to calculate the Root Mean Square of the Successive Differences (RMSSD) with the following equation:
        Equation 3 RMSSD = (2)
        Where N = Number of RR Intervals (R-R)i = Interval between neighboring QRS Peaks (R-R)i+1 = Interval between subsequent set of peaks
    8. Choose this variable as it has been shown to be efficacious on intervals ranging from 1 min to 24 h in length13,14,15,16,17 and can therefore be used to assess these 5 min intervals in the game phases. Along with RMSSD, obtain the Standard Deviation of NN intervals to measure changes in HR throughout the phase14,16,18.
    9. Use the RR intervals to calculate NN50, the number of intervals that differ from the previous interval by more than 50 ms12 which has also been used on intervals ranging from one min to 24 h16,17,19,20,21.
      1. Calculate the NN50 variable via a simple count function that checked whether or not the difference between consecutive RR interval lengths was greater than 50 ms. Once NN50 was obtained in this manner, divide by the total number of intervals to calculate pNN50 which is the percentage of intervals that differ by more than 50 ms. This calculation allowed the measured data to be compared across subjects, games, and even sessions of varying lengths as it is a unit-less variable13,14,16,17.
    10. Calculate mean RR interval length for each phase and subject as a separate HRV measure16,17,19,22,23,24. Use this measure to calculate Average HR by dividing the mean RR interval by 60 s. Both of these measures are easily comparable across game sessions to observe the trend of the subject's activity16,17,19,22,23,24.
    11. Once these measures were calculated, calculate the Low Frequency and High Frequency Power Spectral Density (PSD) for both the raw ECG of the 5-min interval and the RR interval matrix by obtaining PSD from Fast-Fourier transforms13,14,17,19,25. All these data were then stored in a table, an example of which is shown in Table 4.

2. Acquire ECG Data from the Patient

  1. Prepare the HR monitor chest strap and Bluetooth module for application to the subject.
    1. Ensure that the Bluetooth module has been fully charged (3 h) using the charge cradle.
    2. Plug the module into the data computer via the charge cradle and open the config tool. Enter a name for logging purposes.
    3. Select the HR device, click the Time tab and select Set Date/Time to sync the module to the correct time and date. The device can now be removed from the charge cradle.
    4. Moisten the conductive areas (beige) on the HR monitor chest strap by placing a hand in water and rubbing the conductive areas.
    5. Place the HR monitor Bluetooth module into the chest strap with the conductive surfaces of the module lined up with those of the chest strap: it will click into place.
    6. Press and hold the button on the module until the lights flash. The module is now on and recording.
    7. Apply the HR monitor chest strap (with Bluetooth module) to the player with the module aligned with the left mid-axillary line and the strap just under the pectoral muscles. Once properly positioned, tighten the device so that it will not move during the session but is not uncomfortable for the player.
  2. Acquire a signal and view the live feed.
    1. Plug the connector into the USB port of computer that will be used to view the data.
    2. Open the Live View program and enter Setup Mode by clicking the icon with the wrench and screwdriver.
    3. Choose a player from the list if appropriate or add a new subject with the New button in the bottom left corner of the screen.
    4. Enter Subject information as desired for identification purposes (name, age, gender, height, weight).
    5. Click on the Hardware tab and select the current subject.
    6. Click Assign in the bottom of the tab and select the current device (listed as 01 if no other devices are present). Then click assign in the pop-up box.
    7. Click on the Team tab. Highlight the subject and then click the right arrow button to place the player on Team A.
    8. Click on the Deployment tab and then move the newly created team to the first tab.
    9. Open the Live Mode tab by clicking the blue Wi-Fi symbol in the top left corner.
    10. Use the Live Mode tab to monitor HR, respiratory rate, and posture of the subject in real time.
      NOTE: Signal strength, battery power, and confidence of the measures can also be viewed.
    11. Record accurate timing (MM/DD/YYYY HH:MM:SS) of the start and end of each session and phase for processing.
  3. Download the ECG Data from the HR monitor.
    1. Remove the strap from the player at the end of the session and remove the Bluetooth Module from the chest strap.
    2. Place the module in the charge cradle and plug it into a computer with the software program installed.
    3. Open the log.
    4. Select the device from the dropdown menu. All sessions currently on the device are displayed with dates and times.
    5. Uncheck the box that says Use Default Save Location and chose a new save location.
    6. Click Save. A progress bar will then appear. Saving may take up to an hour depending on the length of the session.
    7. Rename the date, once it has been saved.

3. Data Analysis and Calculation of Heart Rate Variability Measures

  1. Prepare files for processing .
    1. Name ECG files as 'KOLLECT_Subject#_AVG4' (e.g., KOLLECT_01_AVG4.csv').
    2. Generate a timing table in comma separated variable (.csv) format to draw timing data from during data processing. See Table 1 for an example of the correct format.
    3. Import the Date-time data from the .csv file and right click on the name of the newly created variable and change it to 'Timing.mat'.
  2. Preliminary R peak detection.
    1. Open and run Peak _ Detection . m.
    2. Enter the frequency of ECG recording device when prompted by the program.
    3. Enter the player number for the data to be analyzed when prompted.
      NOTE: Some players did not complete active video game 4 (AVG4) and therefore only players 1-10 are used for this study. Other numbers will provide an error message.
    4. Enter the number of the game to be analyzed (1, 2, or 3) when prompted.
    5. Enter the phase to be analyzed (Rest, Warmup (WU), Conditioning (Con), Rest, or Recovery).
      1. Enter an offset in minutes if desired, or enter 0 for no offset.
    6. Select the magnifying tool and select an area of the plot that is output to create a window with a width of approximately 2,000 (s/0.004) and a height that will show the full waveform as shown in Figure 3. Zoom in or out if the window is not easily inspected visually.
    7. Visually inspect the graph to evaluate if the peaks detected are correctly labeled. See Figure 1 for example of incorrectly detected and missed peaks caused by irregular ECG data (Figure 2).
  3. Peak Correction
    1. Correct the incorrectly detected or missing peaks by locating the Detection variable and double clicking in the workspace.
    2. Utilize the Data Cursor tool on the plot of the ECG waveform to obtain the x and y coordinates of the incorrect peak; X (time*frequency) is the first column in Detection.mat and Y (Voltage) is the second column (Figure 3).
      1. Right click the text box that appears and click Select Cursor Update Function.
      2. Select TooltipUpdate.m from the folder containing the files used for this analysis. This will allow the tooltip to display more exact values.
    3. If the point is a false positive, remove it from the array by clicking on its row in the Detection.mat variable and pressing Control and the Minus key. An example of false positive detection can be seen in Figure 3.
    4. Edit incorrectly marked peaks that are adjacent to unmarked peaks, as shown by the two T peaks marked as R in Figure 1, by changing their values to match that of the unmarked peak.
    5. Obtain the value of the missed peak can be obtained with the Data Cursor tool.
    6. Add additional rows to Detection. mat using control and the plus key for peaks missed due to low voltage levels.
    7. Enter the values in numerical order to avoid negative values during the calculation process (i.e., add the peak located at 11000 between the peaks at 10908 and 11167) (Figure 5).
    8. Ensure that values are entered correctly before continuing through the full session as numbers are occasionally clipped off when entered.
    9. Repeat step 2.3 until all peaks have been checked and/or corrected.
      NOTE: Some files have limited variability in waveform amplitude and are quicker to check, as seen in Figure 4 while others are more variable and may require closer zoom to accurately locate peaks during visual inspection.
  4. Obtain HRV measure calculations.
    1. Save the original plot generated from Peak_Detection.m for later reference.
    2. Run HRV_Measures.m to generate the correctly labeled plot. A sample of corrected data is shown in Figure 6.
      1. Change the plot title by using Insert | Title on the plot window and changing it to the desired title.
      2. Check the window for output, the program will notify the user of the location incorrectly entered data if any exists.
    3. Save the variable named interval.
    4. Open the variable entitled HRV from the Workspace window to view Mean RR (ms), Average HR (bpm), RMSSD (ms), SDNN (ms), NN50 (count), pNN50 (%), low frequency (LF)/ high frequency (HF) (ECG), LF/HF RR, Low Frequency Power RR, and High Frequency Power (RR)). Save the h values of these variable to a table such as the one shown in Table 4.
    5. Repeat Sections 3.2 – 3.4 for all other segments, sessions, and subjects that need analysis.

Representative Results

This method provides data for use in analyzing the effect that a newly developed method has on the subject's Heart Rate Variability (HRV). It does this by locating the R portion of the QRS waveform of a subject's ECG data, as shown in Figure 6, and by calculating various HRV values from it. If the HR monitor is making proper contact with the subject, the data will be uniform, substantially reducing the need for corrections (as seen in Figure 4).

Thresholds should be set to handle messy and irregular data as depicted in Figure 1 and Figure 2. If the data are sufficiently variable due to momentary changes in the HR monitor skin contact, the initial analysis may incorrectly label peaks as shown in Figure 3. This error can be rectified by manually correcting values or entering extra data points as explained in Section 3 of the protocol. Altering the threshold levels and minimum time between peaks can also help to clean up the detection values and produce an adjusted plot like Figure 6 from Figure5.

Once the data have been obtained and analyzed for discrepancies, they can be used to calculate HRV values for statistical analysis. The analysis of ECG data can be used to quantify observations made during sessions for evaluation purposes.

Figure 1
Figure 1. Representative graph of continuous HR (y-axis) in μv) across time (x-axis in s) for subject one game 3 during the warmup session representing 'messy' data. Messy data: In this section R peaks are smaller than the T portion of the waveform. This can cause issues with peak detection. 

Figure 2
Figure 2. An example of some electrocardiogram (ECG) irregular waveform patterns. Irregular Waveform Patterns: Changes in contact with the subject due to movement can cause voltage variations reducing uniformity of the waveform. Please click here to view a larger version of this figure.

Figure 3
Figure 3. An example of an electrocardiogram (ECG) output with an incorrectly labeled peak HR Incorrectly Labeled Peak. Near the top of the figure a spike in voltage causes part of the waveform to be detected as matching the R pattern. It can also cause nearby R patterns to be ignored due to proximity such as the one highlighted at (9924, 2074). 

Figure 4
Figure 4. Representative graph of continuous HR (y-axis) in μv) across time (x-axis in s) clean electrocardiogram (ECG) waveform. Clean Waveform: An example of a section of uniform ECG data with a relatively even waveform and voltage level.

Figure 5
Figure 5. Representative graph of continuous HR (y-axis) in μv) across time (x-axis in s) of a raw electrocardiogram (ECG) prior to cleaning. Data Prior to Cleaning: A 30 sec segment of ECG data from Subject 01 Game 3 during the conditioning phase is shown. Some peaks have been missed and some are incorrectly labeled due to high voltage variability. Please click here to view a larger version of this figure.

Figure 6
Figure 6. Representative graph of continuous HR (y-axis) in μv) across time (x-axis in s) of a raw electrocardiogram (ECG) after clearning. Data Post cleaning: The same 30 sec of ECG data from Subject 01 Game 3 after it has been properly labeled as described in Section 3 of the protocol. Please click here to view a larger version of this figure.

Phase  Time  THR Zone  Game Features 
Resting 5 min  Baseline rest  NA
Warm-up  5 min  40-60% mHR  4 objects + 4 hazards; slower speed 
Conditioning  10 min  60-80% mHR 8 objects + 0 hazards; faster speed
Cool-down  5 min  40-60% mHR 4 objects + 4 hazards; slower speed
Recovery  5 min  Baseline  rest  NA 
KEY:  THR = Target Heart Rate;  NA = Not applicable

Table 1. Active video game (AVG) game phases. KEY: Target heart rate (THR); NA (Not applicable).

Subject AVG Game Warmup Start Conditioning Start Cooldown Start Recovery Start
(MM/DD/YYYY) (MM/DD/YYYY) (MM/DD/YYYY) (MM/DD/YYYY)
(HH:MM:SS) (HH:MM:SS) (HH:MM:SS) (HH:MM:SS)
1 4 1 11/25/2015 11/25/2015 11/25/2015 11/25/2015
1 4 1 16:33:53 16:39:03 16:49:04 16:54:09
1 4 2 11/25/2015 11/25/2015 11/25/2015 11/25/2015
1 4 2 17:27:47 17:32:57 17:43:01 17:48:03
1 4 3 11/25/2015 11/25/2015 11/25/2015 11/25/2015
1 4 3 18:25:22 18:30:33 18:40:35 18:45:38
2 4 1 4/10/2016 4/10/2016 4/10/2016 4/10/2016
2 4 1 11:59:19 12:04:29 12:14:36 12:19:50
2 4 2 4/10/2016 4/10/2016 4/10/2016 4/10/2016
2 4 2 12:40:25 12:45:37 12:55:44 13:00:53
2 4 3 4/10/2016 4/10/2016 4/10/2016 4/10/2016
2 4 3 13:19:57 13:25:02 13:35:04 13:40:11
3 4 1 11/18/2015 11/18/2015 11/18/2015 11/18/2015
3 4 1 17:08:10 17:13:20 17:23:21 17:28:28
3 4 2 11/18/2015 11/18/2015 11/18/2015 11/18/2015
3 4 2 17:59:46 18:04:48 18:14:54 18:19:55
3 4 3 11/18/2015 11/18/2015 11/18/2015 11/18/2015
3 4 3 18:42:03 18:47:03 18:57:04 19:02:02

Table 2. Timing File KEY: AVG = Active video game

ID_AVG_Game AVNN (s) Avg HR (bpm) RMSSD (ms) SDNN (ms) NN50 pNN50 (%) LF / HF (ECG) LF / HF (RR) LFP (RR) HFP (RR)
03_AVG4_G1_Rest 719.875 83.347 29.827 55.604 35 8.393 1.328 0.602 0.123 0.204
03_AVG4_G1_WU 656.373 91.411 26.52 50.372 28 5.932 1.288 0.675 0.125 0.185
03_AVG4_G1_Con 1 -5 557.772 107.57 20.651 43.932 4 0.743 1.187 0.76 0.119 0.157
03_AVG4_G1_Con 6 10 532.483 112.679 27.771 33.481 9 1.599 1.244 0.809 0.118 0.146
03_AVG4_G1_Con 2 – 7 538.546 111.41 20.389 34.351 6 1.077 1.198 0.819 0.118 0.144
03_AVG4_G1_Con 3 – 8 530.761 113.045 27.756 34.26 8 1.413 1.192 0.826 0.118 0.143
03_AVG4_G1_Cool 597.019 100.499 31.806 41.96 16 3.181 1.281 0.712 0.120 0.169
03_AVG4_G1_Recovery 665.511 90.156 42.136 70.698 57 12.639 1.301 0.636 0.122 0.191
AVNN = Average NN Interval; Avg HR = Average heart Rate; RMSSD = Root Mean Square of Successive Differences; SDNN – Standard Deviation of NN Interval; NN50 = Number of NN Intervals > 50 ms; pNN50 = % of NN intervals > 50 ms; LF = Low Frequency Power; HF = High Frequency Power; LF/HF = Low Frequency – High Frequency Ratio.  bpm = beats per minute; ms = milliseconds; ECG = Electrocardiogram – which contains the QRS complex;  RR =  where R is a point associated with a peak of the QRS complex of the ECG wave and RR is the interval between successive R points; 

Table 3. Heart Rate Variability (HRV) Data for Subject 03 Game 01

Table 4. Descriptive Statistics of Heart Rate Variability Measures for Various Phases of Exercise for Each Game  Please click here to download this table.

Gender GMFCS Level Clinical Diagnosis Movement Disorder Dominant Side Height (cm) Weight (kg) BMI (kg/m2) BMI percentile
boy 2 diplegia dystonia right 161.20 47.60 18.32 17.00
boy 3 diplegia spasticity left 141.17 49.20 24.70 95.00
boy 2 left hemiplegia spasticity right 165.80 50.50 18.40 13.00
boy 3 diplegia spasticity right 154.30 57.00 23.90 83.00
girl 2 left hemiplegia spasticity right 161.20 60.30 22.86 71.00
girl 2 left hemiplegia spasticity right 146.40 40.80 19.00 30.00
girl 2 right hemiplegia spasticity left 154.60 64.00 26.80 85.00
girl 3 left hemiplegia spasticity right 166.10 61.20 22.20 42.00
boy 2 left hemiplegia spasticity right 168.10 49.70 17.60 51.00
boy 3 diplegia spasticity right 135.00 29.80 16.00 43.00
KEY:   GMFCS= Gross Motor Function Classification System;  BMI= Body Mass Index

Table 5. Patient demographics

Discussion

Ten youth with CP participated in this study (mean + SD) [ age (yrs) = 15.53 ± 3.57 ; height (cm) 154.8 ± 12.6; weight (kg) 50.69 ± 11.1; body mass index (BMI) 50.46 ± 29.2; mHR 9 bpm) = 186.8 ± 12.4]. Please see Table 5 for patient demographics.

There are some considerations for use of HR monitors and the associated measures of HR and HRV which relate to modifications and troubleshooting. Two issues that are apparent, regardless of the technology employed to acquire the data are: 1) motion artifacts and 2) ectopic beats. The problems that arise from motion artifacts and ectopic beats are typically addressed through post-processing activities subsequent to the acquisition of the RR interval12,13,18,22,26. Troubleshooting post-processing manipulations require consideration of the temporal fluctations in HR which highlight respiratory sinus arrythmias as well as calculation of the normalized HRV values so differentiations can be made between physiologically and mathematically mediated changes in HRV13,27,29.

Limitations in HRV measurements were initially identified with the application of spectral analysis techniques (i.e., frequency domain measures)13,27,29. There are physiological considerations which include the respiratory sinus arrythmias, cardiovascular drift, hydration status and environmental factors (e.g., temperature, heat, cold, altitude) that are associated with day-to-day variations in HR27,29. Mathematical considerations involve time domain measures (e.g., SDNN, r-MSSD, pNN-50 index) as well as the recent inclusion of non-linear dynamic analysis techniques13,27,29. To correctly interpret the various HRV measures we need to consider whether the body is in a state of rest or stress. Typically we expect parasympathetic influences when the body is rested which increased variability in the responses and results in higher HRV while during stress we expect sympathetic influences which reduce variability and have lower HRV measures. The limitations in HRV measurements can influence the accuracy autonomic balance hypothesis is associated with the LF/HF ratio. This hypothesis assumes that the sympathetic nervous system and parasympathetic nervous system are in competition to regulate SA node firing. The authors note tha the LF/HR ratio needs to be interpreted with caution while noting the context of obtaining information as well as reviewing the LF and HF values. Concerning the application of LF/HF ratio to AVG games in short-term episodes of HR measurements and HRV, a high LF/HF ratio may indicate higher sympathetic activity that may be observed when meeting a challenge that requires effort and increases the sympathetic nervous system activation35.

It is important to use optimal measures to determine aerobic performance and capacity in youth with CP to examine appropriate intervention dosing and effectiveness6,11. Clinical standards of care most often include measuring HR to determine intervention dosing (intensity)6,11. However the inherent variability in HR measures make it difficult to determine actual workoad in aerobic training12,13,22,27. Therefore, this methodology of calculating HRV from ECG data from a HR monitor provides a more accurate measure to assess intervention outcomes27,28. Also, the HRV measures provide new information on the autonomic nervous system responses, adaptation and recovery during the AVG exercise12,13,29,34,35. We posit that application of HRV measures during short durration exercise may provide information on the improvement of the physiological systems based on work by Kerppers and colleagues with a short duration32.

Noted here are the important applications we have made relative to the existing applications of HR monitoring and HRV measures during exercise performance. This methodology allows the user to extract RR intervals and HRV measures from ECG waveforms during gaming physical activities in youth with CP. The method is currently tailored towards AVG sessions in a specific game but could easily be adapted to other protocols and ECG devices for future experiments. In cases where the data are uniform and the ECG recording device is well fitted to the subject, this protocol will allow for quick data processing with minimal input from the user. However, in the case of non-uniform data with large variances in signal amplitude the protocol will require user input to correctly label missed peaks and to remove false positives from the data set. In the future this method may be improved with a more robust detection method to reduce user aid for peak detection and correction (e.g., non-linear dynamic analysis techniques29).

Throughout execution of the protocol, it is essential that the following critical steps are performed. It is important to ensure a high level of signal confidence throughout data collection sessions to reduce the processing and peak correction time required. This can be improved by ensuring that the ECG recording device is making proper contact with the subject prior to each session. It is also important to keep the conductive contacts moist during the sessions which can be done by rewetting the recorder prior to each session. As well, after the data are collected, post-processing activities need to address the methodological considerations with time domain measures, frequency domain measures, non-linear dynamic analyses as well as calculating normalized HRV values to distinguish between physiologically derived and mathematically mediated changes in HRV12,13,29.

Considerations for future work include application of HRV measurements for children and adults involved in physically challenging activities of different intensities and body positions6,7,8,9,10,17,23,26,29, cognitively challenging games and mental workload24,25,26,27, virtual and simulation type experiences, assessment of overtraining23,31, quality of sleep assessments13,26,27,31, chronic fatigue, physical exhaustion and combat readiness31 as well as the vagal connection between HR and the brain regarding prosocial behavior30.

開示

The authors have nothing to disclose.

Acknowledgements

The authors thank the participants and their families for their time and effort expended for participation in the study. As well, the authors acknowledge Dr. Yichuan Liu and Dr. Hasan Ayaz for their assistance with the timing calculation of the HR monitoring and Dr. Paul Diefenbach for development of the KOLLECT Active Video Gaming software. Funding for this work was provided by Coulter Foundation Grants #00006143 (ONeil; Diefenbach, PIs) and #00008819 (ONeil; Diefenbach, PIs).

Materials

BioHarness Bluetooth Module (Electronics sensor)  Zephyr 9800.0189 Detects Heart Rate, Resiration Rate, Posture, and Skin Temperature.
BioHarness Chest Strap Zephyr 9600.0189, 9600.0190 Sizes Small XS-M, Large M-XL
BioHarness Charge Cradle & USB Cable Zephyr 9600.0257 Used to Transfer Data from the Module to a Computer for Analysis.
BioHarness Echo Gateway Zephyr 9600.0254 Allows for Realtime Viewing of Subject's Heart Rate.
MATLAB R2016a Mathworks 1.7.0_.60 Used for All Programming.

参考文献

  1. Winter, S., Autry, A., Boyle, C., Yeargin-Allsopp, M. Trends in the prevalence of cerebral palsy in a population-based study. Pediatrics. 110 (6), 1220-1225 (2002).
  2. Fowler, E., et al. Promotion of physical fitness and prevention of secondary conditions for children with cerebral palsy: Section on Pediatrics Research Summit Proceedings. Physical Therapy. 87 (11), 1495-1510 (2007).
  3. Rosenbaum, P., Paneth, N., Leviton, A., Goldstein, M., Bax, M. A report: The definition and classification of cerebral palsy: April 2006. Developmental Medicine & Child Neurology. 49 (s109), 8-14 (2007).
  4. Hanna, S., et al. Stability and decline in gross motor function among children and youth with cerebral palsy aged 2 to 21 years. Developmental Medicine & Child Neurology. 51 (4), 295-302 (2009).
  5. Rimmer, J., Rowland, J. Health promotion for people with disabilities: Implications for empowering the person and promoting disability-friendly environments. American Journal of Lifestyle Medicine. 2 (5), 409-420 (2008).
  6. Feehan, K., et al. Factors influencing physical activity in children and youth with special health care needs: A pilot study. International Journal of Pediatrics. , (2012).
  7. Fehlings, D., Switzer, L., Findlay, B., Knights, S. Interactive computer play as motor therapy for individuals with cerebral palsy. Seminars in Pediatric Neurology. 20 (2), 127-138 (2013).
  8. Sandlund, M., Dock, K., Hager, C., Waterworth, E. Motion interactive video games in home training for children with cerebral palsy: parents’ perceptions. Disability & Rehabilitation. 34 (11), 925-933 (2012).
  9. Howcroft, J., et al. Active video game play in children with cerebral palsy: Potential for physical activity promotion and rehabilitation therapies. Archives of Physical Medicine and Rehabilitation. 93 (8), 1448-1456 (2012).
  10. Bilde, P., Kliim-Due, M., Rasmussen, B., Petersen, L., Petersen, T., Nielsen, J. Individualized, home-based interactive training of cerebral palsy children delivered through the Internet. BMC Neurology. 11, 32 (2011).
  11. Kolobe, T., et al. Research Summitt III proceedings on dosing in children with an injured brain or cerebral palsy. Physical Therapy. 94 (7), 907-920 (2014).
  12. Schipke, J., Pelzer, M., Arnold, G. Effect of respiration rate on short-term heart rate variability. Journal of Clinical and Basic Cardiology. 2 (1), 92-95 (1999).
  13. Ernst, G. Heart rate variability. Heart Rate Variability. , 1-336 (2014).
  14. Francis, J., et al. Association between symptoms of depression and anxiety with heart rate variability in patients with implantable cardioverter defibrillators. Psychosomatic Medicine. 71 (8), 821-827 (2009).
  15. Mendes, R., et al. Is applying the same exercise-based inpatient program to normal and reduced left ventricular function patients the best strategy after coronary surgery? A focus on autonomic cardiac response. Disability and Rehabilitation: An International Multidisciplinary Journal. 36 (2), 155-162 (2014).
  16. Muralikrishnan, K., Balakrishnan, B., Balasubramanian, K., Visnegarawla, F. Measurement of the effect of Isha Yoga on cardiac autonomic nervous system using short-term heart rate variability. Journal of Ayurveda and Integrative Medicine. 33 (2), 279-283 (2012).
  17. Yadav, R. K., Gupta, R., Deepak, K. K. A pilot study on short term heart rate variability & its correlation with disease activity in Indian patients with rheumatoid arthritis. Indian Journal of Medical Research. 136 (4), 593-598 (2012).
  18. Thuraisingham, R. A. Preprocessing RR interval time series for heart rate variability analysis and estimates of standard deviation of RR intervals. Computer Methods and Programs in Biomedicine. 83 (1), 78-82 (2006).
  19. Alamili, M., Rosenberg, J., Gögenur, I. Day-night variation in heart rate variability changes induced by endotoxaemia in healthy volunteers. Acta Anaesthesiologica Scandinavica. 59 (4), 457-464 (2015).
  20. Pal, G., et al. Preference for salt contributes to sympathovagal imbalance in the genesis of prehypertension. European Journal of Clinical Nutrition. 67 (6), 586-591 (2013).
  21. Telles, S., Raghavendra, B. R., Naveen, K. V., Manjunath, N. K., Kumar, S., Subramanya, P. Changes in autonomic variables following two meditative states described in yoga texts. Journal of Alternative and Complementary Medicine. 19 (1), 35-42 (2013).
  22. Kičmerová, D. . Methods for Detection and Classification in ECG Analysis. Doctoral thesis. , (2009).
  23. Murai, K., Hayashi, Y. Evaluation of mental workload for ship handling using physiological indices. , 604-608 (2009).
  24. Taelman, J., Vandeput, S., Spaepen, A., Van Huffel, S. Influence of mental stress on heart rate and heart rate variability. Heart. 29 (1), 1366-1369 (2009).
  25. Durantin, G., Gagnon, J. F., Tremblay, S., Dehais, F. Using near infrared spectroscopy and heart rate variability to detect mental overload. Behavioural Brain Research. 259, 16-23 (2014).
  26. Buchheit, M. Monitoring training status with HR measures: Do all roads lead to Rome?. Frontiers in Physiology. 5, (2014).
  27. Achten, J., Jeukendrup, A. Heart rate monitoring: Applications and limitations. Sports Medicine. 33 (8), 517-538 (2012).
  28. Amichai, T., Katz-Leurer, M. Heart rate variability with cerebral palsy: Review of literature and meta-analysis. NeuroRehabilitation. 35, 113-122 (2014).
  29. Billman, G., Haikuri, H., Sacha, J., Trimmel, K. An introduction to heart rate variability: Methodological considerations and clinical applications. Frontiers in Physiology. 6, (2015).
  30. Beffara, B., Bret, A., Vermeulen, N., Mermillod, M. Resting high frequency heart rate variability selectively predicts cooperative behavior. Physiology & Behavior. 164, 417-428 (2016).
  31. Fogt, D., Cooper, P., Freeman, C., Kalns, J., Cooke, W. Heart rate variability to assess combat readiness. Military Medicine. 174, 491-495 (2009).
  32. Kerppers, I. L., Arisawa, E. A. L., Oliveira, L. V. F., Sarmpaio, L. M. M., Oliverira, C. S. Heart rate variability in individual with cerebral palsy. Archives of Medical Science. 5, 45-50 (2009).
  33. Giggins, O. M., Persson, U. M., Caulfield, B. Biofeedback in Rehabilitation. Journal of Neuroengineering and Rehabilitation. 10, (2013).
  34. Shaffer, F., Ginsberg, J. P. An overview of heart rate variability metrics and norms. Frontiers in Public Health. 5, 258 (2017).
  35. Shaffer, F., McCarty, R., Zeir, C. L. A healthy heart is not a metronome: an integrative review of the heart’s anatomy and heart rate variability. Frontiers in Psychology. 5, 1040 (2014).

Play Video

記事を引用
Landis, C., O’Neil, M. E., Finnegan, A., Shewokis, P. A. Calculating Heart Rate Variability from ECG Data from Youth with Cerebral Palsy During Active Video Game Sessions. J. Vis. Exp. (148), e59230, doi:10.3791/59230 (2019).

View Video