A Pilot Study on the Effectiveness ofHeart Rate Variability Biofeedback onHealthy Subjects

(1)

DEGREE PROJECT, IN MEDICAL ENGINEERING , SECOND LEVEL STOCKHOLM, SWEDEN 2015

A Pilot Study on the Effectiveness of Heart Rate Variability Biofeedback on Healthy Subjects

SOFIE SJÖDAHL

KTH ROYAL INSTITUTE OF TECHNOLOGY

(2)

(3)

(4)

Master of Science Thesis in Medical Technology (HL202X), 30 ECTS

2015:014

A Pilot Study on the Effectiveness of Heart Rate Variability Biofeedback on

Healthy Subjects

Sofie Sjödahl

Approved

2015-03-13

Reviewer

Kaj Lindecrantz

Supervisor

Farhad Abtahi

Examiner

Mats Nilsson

(5)

Abstract

Heart Rate Variability (HRV) is a medical term describing the heart’s natural varying time difference between heartbeats (called NN-intervals). A higher HRV i.e. a larger variability between NN-intervals is connected to healthiness, and lower HRV to unhealthy states.

Biofeedback (BF) is a method that can detect and send physiological signals back to the user on a screen. Ev- ery individual has a resonant frequency, and when breathing at this frequency, the interplay between blood pressure and respiration causes HRV to increase momentarily.

HRV biofeedback aims at increasing HRV, by measuring heart rhythm and respiration to guide the user in resonant breathing.

This thesis had the objective to investigate the effectiveness of one 20 minutes long biofeedback session with resonant frequency breathing. The hypothesis was that the time frame would be longer than two hours. It was carried out with 12 healthy volunteers, who participated in a biofeedback session with an Android application, and afterwards 5 minutes long ECG measurements were made every half hour for two hours. A control session was held with the same participants to give the trial more scientific strength.

The result showed that a 20 min resonant breathing biofeedback session can elevate Standard Deviation of NN intervals (SDNN) significantly (p < 0.05), 2 hours after the biofeedback session. The conclusion was that the hypothesis cannot be rejected, but the result is too weak to strengthen it much. Further research is needed to draw more conclusions about the time frame of HRV elevations in healthy people.

(6)

Examensarbete i Medicinsk Teknik (HL202X), 30 ECTS

2015:014

En pilotstudie av effektiviteten av Heart Rate Variability Biofeedback med friska deltagare

Sofie Sjödahl

Godkänt

2015-03-13

Recensent

Kaj Lindecrantz

Handledare

Farhad Abtahi

Examinator

Mats Nilsson

(7)

Referat

Heart Rate Variability (HRV) är en medicinsk term som beskriver hjärtats naturliga variation av tidsskillnad mellan hjärtslag (tiden mellan två hjärtslag kallas för ett NN-intervall). Ett högre HRV, dvs. större variation mellan NN-intervallen, är kopplat till hälsa, och lägre HRV är kopplat till ohälsa.

Biofeedback (BF) är ett redskap som kan mäta och visa användarens fysiologiska signaler på exempelvis en skärm.

Alla människor har en individuell resonansfrekvens, och vid andning i denna frekvens kan HRV höjas tillfälligt. HRV biofeedback är ett medel till för att höja HRV, genom att under mätning av hjärtrytm och andning guida användaren i andning i resonansfrekvensen.

Det här examensarbetet hade som syfte att undersöka effektiviteten av en 20 minuter lång HRV biofeedbackses- sion. Hypotesen var att tidsspannet för vilket HRV förblir högre än normalt är längre än två timmar. Försöket ut- fördes med 12 friska frivilliga testpersoner som använde en androidapplikation för HRV biofeedback, efter vilken EKG mättes i 5-minuterssekvenser varje halvtimme i två timmar. Deltagarna figurerade även som kontrollgrupp för att styrka experimentet.

Resultatet visade att en 20-minuters session av HRV biofeedback kan medföra ett höjt Standard Deviation of NN intervals (SDNN) signifikant (p < 0.05), med verkan två timmar efter sessionen. Slutsatsen var att hypotesen inte kan förkastas, men att resultatet är för svagt för att stärka den nämnvärt. Vidare forskning behövs för att dra fler slutsatser om tidsspannet för ökning av HRV hos friska personer.

(8)

(9)

Acknowledgements

I would like to thank my supervisor Farhad Abtahi for the help and support. Warm thanks are also directed to all who participated in the trial; it wouldn’t have been possible without you. Last, I thank my family for their never ending love and support.

Thank you!

(10)

(11)

List of Figures

2.1 Respiration curve (top) and HRV (bottom), at the resonant frequency . 5

2.2 A fictitious ECG graph . . . 6

2.3 An example of a tachogram . . . 6

2.4 Example of a PSD plot . . . 7

2.5 Illustration of the Pan Tompkins Method . . . 9

3.1 The measurement device . . . 12

3.2 Body electrode placement . . . 12

3.3 The biofeedback application interface . . . 13

3.4 The game interface . . . 13

4.1 Histogram of NN intervals . . . 21

4.2 SDNN box plots . . . 21

4.3 HF power control session box plot . . . 22

4.4 Mean HR control session box plot . . . 22

4.5 PSD just before biofeedback (3a) . . . 23

4.6 PSD 120 minutes after biofeedback (3e) . . . 23

4.7 Blood pressure box plots . . . 24

4.8 Resonant frequency compared to participant height and weight . . . 24

4.9 Absolute BF effect compared to coffe drinking and exercise. . . 25

4.10 Relative BF effect compared to coffe drinking and exercise. . . 26

(14)

List of Tables

2.1 Selected HRV measures . . . 7

3.1 The trial protocol . . . 14

3.2 Participant demographics . . . 15

3.3 Resonant breathing frequency criteria . . . 15

4.1 A summary of the HRV data aquired . . . 20

4.2 Blood pressure results . . . 25

A.1 SDNN data for all subjects . . . 37

A.2 Mean HR data for all subjects . . . 38

A.3 LF power data for all subjects . . . 38

A.4 HF power data for all subjects . . . 39

A.5 Total power data for all subjects . . . 39

A.6 Correlation p-values for resonant breathing frequency comparisons . . . 40

A.7 Correlation p-values for BF effect comparisons . . . 40

x

(15)

Acronyms

ANOVA Analysis of Variance AR Autoregressive

BF Biofeedback

ECG Electrocardiography FFT Fast Fourier Transform GUI Graphical User Interface HF High Frequency

HR Heart Rate

HRV Heart Rate Variability

LF Low Frequency

PSD Power Spectral Density RSA Respiratory Sinus Arrhythmia

SDNN Standard Deviation of Normal-to-Normal intervals TEB Thoracic Electrical Bioimpedance

VLF Very Low Frequency

(16)

(17)

Chapter 1

Introduction

Biofeedback for Heart Rate Variability is a method aimed at increasing health in various areas and it has the capacity to be used in a much larger scale as a healthcare treatment in institutions as well as at home.

Heart Rate Variability is a term describing the varying time between consecutive heart beats. It is known that a higher variability is connected to healthiness and a lower variability is connected to different unhealthy states and conditions, such as myocardial infarction and asthma [1].

The heart’s beating rate is determined by different influences, among which blood pressure and respiration are the main ones. Respiration affects the heart rhythm by increasing it at inhalation and decreasing it at exhalation. Each person has an individual resonant breathing frequency where the interplay between blood pressure and heart rate are in sync, with 180 degrees phase shift, at which HRV is maximized.

Various trials have shown that HRV can increase momentarily when partici- pating in a session of resonant frequency breathing [2]. It has also been shown to help patients improve their condition in various diseases such as depression [3] and asthma [4]. There have however been no studies showing any long term effect on healthy people or how long the acute effect lasts.

A former KTH thesis has developed a HRV biofeedback application for Android using Java. The application receives signals from a Bluetooth device that measures ECG and Impedance (Z) over the upper body, and displays the breathing curve and HRV to the user in a GUI (Graphical User Interface) [5] [6]. Another KTH thesis resulted in an alternative game-influenced interface in the application, to make it more enjoyable for the user [2].

1.1 Objective

The objective with this thesis was to find out for how long time the increased HRV from a single 20 minutes long resonant breathing biofeedback session lasts in general in healthy people, assuming that there is an acute effect. A part of the objective

(18)

CHAPTER 1. INTRODUCTION was also to investigate if blood pressure seems to be lowered by that session, and to review the demographics of the participants to see if there was any correlation between lifestyle and results from the biofeedback, as well as evaluating the BF application in terms of usability and suggest possible improvements.

Hypothesis: The effect on HRV from a 20 minutes biofeedback session lasts longer than two hours

1.2 Method

The objective was accomplished by carrying out biofeedback trials with 12 healthy subjects using the application developed in the former thesis [5]. A literature study was carried out in order to set up the testing protocol and plan for analysis. The group of participants also figured as a control group. The control session was carried out with ECG measurements every half hour for two hours. After that, the group of volunteers underwent a HRV biofeedback training session, and afterwards identical ECG measurements every half hour for two hours.

Signal processing was carried on the obtained data and measures reflecting HRV was extracted. The results from the biofeedback session and the control session were evaluated with statistical testing to ensure the result was significant.

1.3 Thesis Outline

Chapter 2 contains a presentation of the background of HRV, resonant breathing and biofeedback, giving a base to understand the project. It also contains information about signal processing, HRV measures and the state of the art. Chapter 3 describes the methods used to conduct the study, the trial protocol, how the trials were performed and how the information was processed and analyzed.

Chapter 4 displays the results of the study and the objectives as well as whether the hypothesis was rejected or not. Chapter 5 contains a discussion about the results, an evaluation of this project, an assessment of the utilized software and suggested future work. The conclusion is stated in chapter 6.

Appendix A contains detailed data from the trials, as well as data from the correlation comparisons of effectiveness of biofeedback to habits and comparison of resonant frequency to height and weight of the participants. Appendix B contains the full questionnaire handed to the participants before trials.

2

(19)

Chapter 2

Theory

This chapter introduces the theory behind the project, starting with the physiology behind HRV and how to measure it. The terms resonant breathing frequency and biofeedback are explained; what they are and how they work. The state of the art is presented to give an idea about how far the research about HRV has come today.

Furthermore, the data handling that was used in the project is introduced, including signal processing, extracting of measures and statistical analysis.

2.1 Heart Rate Variability

Heart Rate Variability is a medical term describing the varying time difference between consecutive heart beats, creating a varying heart rhythm. The heart rhythm is largely controlled by various influences of the autonomic nervous system via the sinoatrial node. The autonomic nervous system is divided into the sympathetic part (“fight and flight”) by release of epinephrine and norepinephrine and the parasympathetic part (“rest and digest”) by release of acetylcholine by the vagus nerve. The resulting impact is a result of the stimuli from both autonomic systems [1].

One of the influences on heart rate is the baroreflex system, which reacts on blood pressure changes, by feedback from stretch sensors in the aorta and carotid arteries. When blood pressure increases, the baroreflex causes heart rate to decrease which in turn means that a smaller volume of blood is pumped through the system, and blood pressure falls. A falling blood pressure triggers the baroreceptors and heart rate goes up again. In this way, heart rate and blood pressure changes in loops, triggered by each other [7]. Another influence on heart rate is respiration; heart rate increases with inhalation and decreases with exhalation [8] [9]. This respiratory influence on the sinoatrial node is called Respiratory Sinus Arrhythmia (RSA). One of the functions of RSA is controlling the gas exchange rate in the alveoli [10].

HRV has been a known indicator since 1965 when Hon and Lee saw that HRV decreased before fetal distress, before there was any notable change in heart rate [11]. HRV has for long been used in the clinics as a monitoring tool, for example in the earlier mentioned scanning for fetal distress and neonatal supervision [12] and for

(20)

CHAPTER 2. THEORY sudden cardiac death in Chronic Heart Failure patients [13]. It is also used as a risk marker after a myocardial infarction [1] and of overtraining [14]. Furthermore, a low HRV has been connected to both cardiovascular and non-cardiovascular pathologies, like stress and related conditions [15]. This leads to the conclusion that a high HRV is healthy [1], or rather an indicator of emotional and physical resilience [10].

2.2 Resonant Breathing and Biofeedback

Since low HRV is associated with unhealthiness, training sessions have been suggested to provide means of increasing HRV and not only monitoring it. The training usually includes deep diaphragmatic breathing to make the body relax and increase HRV. Many studies suggest that each individual has a resonant frequency (often around 0.1 Hz) at which the breathing is in sync with the heart rate variability.

Figure 2.1 shows how respiration and HRV are in sync with 180 degrees phase shift, while breathing at the resonant frequency. Breathing at this individual frequency maximizes HRV [16] [8] and also gas exchange in the lungs is maximized [10].

The explanation to why there is an individual breathing frequency, seems to be found in the earlier mentioned baroreflex system. The blood pressure change takes a few seconds, while the heart rate changes in less than a second. When a person breathes at the resonant frequency, the stimulus for increased heart rate comes from both the respiratory system and baroreceptors at the same time, causing the heart rate to oscillate with high amplitude, in tune with respiration. A larger mass of blood in the system leads to a slower response to change the blood pressure.

This means that taller people have a longer response time, and therefore a lower resonant frequency. When engaging in HRV biofeedback at the resonant frequency, the baroreflex is exercised and over time rendered more efficient [7].

Biofeedback is a tool for helping users become more aware of their body and help them gain more control, by measuring physiological changes. A few examples of quantities that can be measured are heart rate, blood pressure, muscle activity and skin temperature. The user is notified of changes by for example an interface on a screen, where they can see immediate variations when changing voluntary factors such as breathing or flexing muscles [17].

Biofeedback sessions with regular breathing at the resonant frequency has shown to improve various disorders such as asthma [18] and hypertension [19]. Trials with healthy patients have shown to increase their HRV in an acute timeframe but more research need to be done to establish if there are any long time effects [15].

2.3 State of the Art

As of today it is concluded that HRV biofeedback can increase HRV in the acute timeframe on healthy people [15] but it is not investigated for how long this effect lasts. Studies which have shown long term effects on patients with diseases have not shown any change in the HRV measures.

4

(21)

2.4. INFORMATION EXTRACTION

Figure 2.1. Respiration curve (top) and HRV (bottom), at the resonant frequency

There is a device utilizing resonant breathing on the market, approved by FDA, with the purpose to lower blood pressure by guided breathing. It does not use biofeedback, but instead a continuous frequency-alternating sound to guide the breathing at the user’s resonant frequency [20].

2.4 Information Extraction

There are many means to measure intervals between heart beats; one of them is to identify the interval between heart beats in an Electrocardiogram (ECG) recording.

ECG is a measurement of the electrical activity of the heart occurring at the different phases in the cardiac cycle. The peaks and troughs in the cardiac cycle are named P, Q, R, S, T, where the P-wave represents the depolarization of the atria, the QRS- complex represents the depolarization of the ventricles and the T-wave represents the repolarization of the ventricles, see Figure 2.2 [1]. Since HRV aims to mark the time between the sinoatrial node rhythm, the P-wave would be the best marker.

Unfortunately the signal to noise ratio is often too low for the P-wave, which is why the R-peak is usually used as fiducial point instead, and it has been concluded that the R-R intervals represent the sinoatrial activity rather well [12].

The notation NN (Normal-Normal) interval is sometimes used instead of RR to emphasize that it is actual sinoatrial triggered contractions that are used in the calculations, since not all R-peaks are from normal contractions [1]. Therefore, the NN notation will be used in this report from now for indicating a series free from artifacts. When plotting HRV, a so called tachogram is obtained, showing the beat nr or time span on the x-axis and the NN intervals on the y-axis, see Figure 2.3.

The technique used to measure respiration in this thesis is Thoracic Electrical Bioimpendance (TEB). By injecting current in electrodes on a person’s thorax and using sensing electrodes on the other side, the changes in impedance of the thorax can be measured. The impedance changes due to respiration; inhalation increases air volume in the lungs and lowers the conductivity. The expanding of the chest also prolongs the distance between the electrodes. Both factors combine to increased impedance. Similarly, during exhalation, the impedance decreases [21].

(22)

CHAPTER 2. THEORY

Figure 2.2. A fictitious ECG graph

Figure 2.3. An example of a tachogram

2.4.1 HRV Parameters

To obtain useful information from HRV, several derived quantities exist. These are split into domains, of which the main ones are time domain and frequency domain. Some of the popular measures in these domains are displayed in Table 2.1 for an overview. The most used measures in time domain is SDNN, which stands for Standard Deviation of NN intervals, and is an estimate of the overall HRV (Equation 2.1, where N=number of intervals, RR=RR-interval in seconds) [16].

SDN N = v u u

t 1

N −1

N −1

X

i=1

(RRi+1− RR)² (2.1)

The time domain measures do not require time consuming computations, but neither do they contain any information about the relation between sympathetic and parasympathetic influences.

In frequency domain, PSD (Power Spectral Density) analysis is common, con- taining information about the frequency components in the NN-series. In short time recordings, three divisions are made according to the origin of the frequencies;

Very Low Frequency (VLF), Low Frequency (LF) and High Frequency (HF) [1].

6

(23)

Table 2.1. Selected HRV measures

Parameter Unit Description Time domain

SDNN ms Standard Deviation of NN intervals Frequency domain

LF power ms² Power in low frequency range: 0.04-0.15 Hz HF power ms² Power in high frequency range: 0.15-0.40 Hz Total power ms² Variance of all NN intervals

Figure 2.4. Example of a PSD plot divided into frequency components VLF (dark green), LF (light green) and HF (yellow)

The high frequency component is associated with the parasympathetic influence on heart rate via the vagus nerve and the low frequency component may be associated with baroreceptors that measure blood pressure [16] [22]. The very low frequency component has for a long time had unknown influence, and it will not be analyzed in this thesis. The frequency components are usually displayed is a PSD plot, see Figure 2.4. The quantity that is mostly used for comparison is the power of the frequency component, such as LF power (ms²) [12].

The PSD can be calculated in different ways, two popular methods are the discrete transform FFT and Autoregressive modeling (AR). AR assumes that future values are correlated linearly to previous values [12]. Both these methods require re- sampling, since they demand evenly sampled data and NN-intervals are by definition unevenly sampled. FFT and AR give similar results [12].

Many of the HRV parameters are dependent on the number of data points, which makes it very important that all recorded samples that will be compared have the same time duration [12].

(24)

CHAPTER 2. THEORY

2.4.2 Signal Processing

The raw ECG needs to be processed in order to extract as accurate information as possible. Filtering is used to remove noise from the ECG, and the border frequencies are usually 5-15 Hz to remove high frequency noise and low frequency disturbance from e.g. breathing [23].

The Pan Tompkins method is a reliable QRS detection method that was used in this thesis. The steps can be seen in Figure 2.5. The signal is filtered to remove noise (a) and then differentiated to highlight the QRS complex (b). After that, the signal is squared to make all data points positive. The squaring also highlights the QRS-complex even more since the squared amplification is non-linear(c). An average moving window function is used to obtain waveform information about the QRS-complex (d). Finally, a threshold is set to detect the QRS complex intervals and peak detection to find the R-peaks within the intervals (e) [24].

After obtaining the RR-interval time series, de-trending can be used again, since nonstationarities are clearer in RR than ECG. Artifact reduction is crucial since HRV measures are highly sensitive to artifacts. Ectopic beats are heartbeats that are different from normal ones, there can be a shortening of a RR-interval directly followed by a lengthened one [12]. Artifacts can also be due to movement or equip- ment problems. For handling outliers connected to the artifacts, and obtain the NN-series from the RR-series, removal or interpolation can be carried out. Removal is a fast and simple method, in which the outlier’s data point is deleted, and unfortunately the continuity of the signal is lost. Many methods with interpolation exists, which adjusts the outlier to the rest of the signal [25], for example by replacing the outlier with a mean or median value of a chosen number of neighboring intervals. There is also a method using cubic spline interpolation, but this one has showed to be inferior compared to removal or non-linear interpolation [26].

2.4.3 Statistical Analysis

To ensure that the results drawn from this thesis are statistically significant, a statistical comparison was made. The t-test is a common statistical method in order to decide if there is a significant difference to two sets of data [27]. The ANOVA (Analysis of Variance) test is another statistical test that can be used to make comparisons between more than two groups of data. It is based on comparison of variance within the groups to variance between the groups. If the variance within groups is larger than between groups, there is no significant difference between the group means [28].

8

(25)

Figure 2.5. Illustration of the Pan Tompkins Method: filtration to remove noise (a), derivation to highlight the QRS complex (b), squaring to make all data points positive and amplify the ECG frequencies (c), average moving window function (d), thresholding to detect the QRS complex intervals and peak detection to find the R-peaks within the intervals (e).

(26)

(27)

Chapter 3

Methodology

This chapter describes the process of the project, divided into the following sections:

the utilized biofeedback system, defining of the trial protocol, performing of trials and analysis of data.

The project started with a literature research as well as an analysis of the two former theses and the developed software. The literature research was conducted primarily with KTH Primo and focused on HRV biofeedback for healthy subjects and protocol setup to gain the relevant knowledge in order to perform the trials. It also included research about what methods to use when analyzing the data.

3.1 The Biofeedback System

The existing biofeedback setup consists of a device (Figure 3.1) for measuring ECG and respiration (TEB) at a sampling rate of 200 Hz. The ECG is measured with 3 electrodes, and the TEB with 4 electrodes, placed on the body as can be seen in Figure 3.2. The device communicates with an Android application on a tablet via Bluetooth. As mentioned before, the application is a result of a former KTH thesis [5] which listens for data sent from the measuring device. The application’s interface is simple with a few graphs and control options, see Figure 3.3. It contains an ECG plot, a tachogram, a respiration plot and a PSD plot and some measures such as heart rate, among other things. The plot of the respiration curve also contains a sine wave which the user is meant to match the breathing to. The sine wave changes wavelength according to the breathing frequency that is entered, which can unfortunately only be an integer. It is optional to let the program play a sound at every peak and trough of the sine wave for support.

The addition made in the other thesis [2] includes a game tab where the user’s breathing controls the image of a moving balloon, see Figure 3.4. The user’s task is to make the balloon follow a sine wave by controlled breathing. The sine wave of the game also adjusts to the user’s entered resonant frequency.

(28)

CHAPTER 3. METHODOLOGY

Figure 3.1. The measurement device

Figure 3.2. Body electrode placement. ECG electrodes are marked as diamonds and TEB electrodes as ovals.

12

(29)

3.1. THE BIOFEEDBACK SYSTEM

Figure 3.3. The biofeedback application interface

Figure 3.4. The game interface

(30)

Table 3.1. The trial protocol

Session Description

1. Control session: Measurement 1a

Measurement 1b: 30 min after 1a Measurement 1c: 60 min after 1a Measurement 1d: 90 min after 1a Measurement 1e: 120 min after 1a 2. Finding the resonant frequency

3. Biofeedback session: Measurement 3a (baseline) Blood pressure measurement Biofeedback for 20 minutes Blood pressure measurement Measurement 3b: 30 min after BF Measurement 3c: 60 min after BF Measurement 3d: 90 min after BF Measurement 3e: 120 min after BF

3.2 Trial Protocol

The number of test subjects was set at 12 since this is a pilot study, and according to Julious [29], 12 is a desirable sample size since it’s enough participants to ensure high enough precision in mean and variance, and at the same time a realistic number of people to gather for a pilot study. Since there are no previous studies, the details of the protocol were established by a pre-trial with 3 participants. It was concluded that within 2 hours, the effects of biofeedback are likely to have faded, and the protocol was decided to contain measurements during two hours after biofeedback training, with 30 minutes between the measurements. The established trial protocol can be seen in Table 3.1. All the recordings were sampled in durations of 5.00 minutes, a standardized recording time, to ensure an appropriate comparison [1].

The criteria for participants in this trial were overall healthiness, age between 20 and 65 years and no chronic diseases or conditions that affect heart rhythm or blood pressure. Participants that were pregnant, lactating or having an active medical implant were prohibited to take part in the study. Participant demographics such as gender, age, weight, height and habits, can be seen in Table 3.2. The gender distribution was 50 % females and 50 % males. The full questionnaire can be seen in Appendix B.

Instead of having a control group, a session with control measurements were performed before the biofeedback session. This ensures that any change in HRV after biofeedback is indeed due to the biofeedback and not simply normal variations over time.

14

(31)

3.3. CONDUCTING THE TRIALS

Table 3.2. Participant demographics

Subject Age [years] Weight [kg] Height [cm] Gender Tobacco Coffee Exercise

1 24 70 162 Female None Daily More than twice a week

2 23 51 166 Female None Daily More than twice a week

3 26 63 179 Male None Daily Once a week

4 24 59 173 Female None Sometimes Twice a week

5 25 70 180 Male None Daily Twice a week

6 30 83 182 Male None Sometimes Less than once a week

7 31 84 187 Male None Daily More than twice a week

8 30 80 172 Male None Sometimes Once a week

9 24 85 174 Female None Never Less than once a week

10 51 61 172 Female None Never More than twice a week

11 33 73 168 Male None Daily Less than once a week

12 24 55 161 Female None Sometimes Once a week

Median 25.5 70 172.5

Mean 28.8 69.5 173

Std 7.8 11.8 8

Table 3.3. Resonant breathing frequency criteria

Parameter Unit

LF power ms²

Total power ms²

LF peak height s²/_Hz

LF peak singularity Ordinal scale 1-10 Smoothness of tachogram Ordinal scale 1-10

3.3 Conducting the Trials

This section describes how a typical trial session was carried out. The potential participant was given a consent letter with all information about the trial and was given time to consider the participation. If deciding to participate, he or she was sent a form regarding demographics and habits to fill in (seen in Appendix B), and the session was scheduled. The participant was asked not to ingest any caffeine or other stimulating substance within two hours before the session, and no food within one hour before.

At the beginning of the session, more detailed information about the study was presented and questions were answered. The participant signed a form claiming that their participation was indeed voluntary and could stop at any moment with-

(32)

CHAPTER 3. METHODOLOGY out the need to give a reason. Electrodes were placed on the upper body and leg of the participant according to the picture on the measurement device (Figure 3.2). The whole trial was split into 3 sessions, as decided in the protocol. The first session consisted of two hours of control measurements, 30 minutes between each measurement. During this time the relaxed participant was sitting down and reading, writing or talking.

The second session was aimed at finding the resonant frequency, f = nr of breaths per minute. To determine the resonant frequency, Paul Lehrer et al [8] suggests to record respiration and ECG when the subject is asked to breathe at different frequencies (often 5.5, 6, 6.5 times per minute) for two minutes. After the recordings the breathing frequency is established to be the one where as many as possible of the measures in Table 3.3 are maximized. In this trial, only frequencies that were integers were tried since the android application with which the biofeedback was carried out, only allows such. For finding the resonant breathing frequency, a simple metronome application was used, giving a sound at the start of each inhalation and exhalation. f =7 was tested first, and then the pace was lowered with one step each time, to 6, 5 and if needed, 4. This order was chosen so as not to shock the participant, rather bringing the breathing pace down slowly. f =7 was practiced a little longer to get the participant used to the technique. The frequency that maximized most of the parameters in Table 3.3 was chosen. Two of the parameters Lehrer mentions were not available in this trial, and in that case, Lehrer instructs to use the remaining ones. If the decision between two frequencies was very hard, the participant was asked which one felt better. No one was forced to breathe in a frequency that felt very uncomfortable.

The third and last session started with a 5 minute baseline measurement, and then a blood pressure measurement with Wrist Blood Pressure Monitor R7 made by Omron Healthcare CO, Ltd. After that, the participant was engaged in 20 minutes of HRV biofeedback with resonant breathing. The participant was able to try both the game interface and the normal interface, and was asked to choose which one they wanted to use. He or she was instructed to match the breathing to the sine wave, and to be aware of dizziness symptoms in which case they should breathe less deep. After the biofeedback, blood pressure was measured again, followed by ECG measurements repeated every half hour for two hours.

At the end, the participant was asked how the whole trial felt and what could be changed for a better experience. It is worth noting that, since this test took at minimum 5 hours, most participants had a major meal in between the sessions, but never less than one hour before any measurement. In some cases the sessions were divided over two different days, due to scheduling reasons.

3.4 Analysis of Data

The ECG was measured with a frequency of 200 Hz, which should be enough for measuring HRV on healthy people, according to Kamath et.al [12]. A QRS detector

16

(33)

3.4. ANALYSIS OF DATA

based on a code by Hooman Sedghamiz [30] was used to pre-process the ECG signal and find the R-peaks. The code is using the Pan Tompkins method for QRS- detection, described in the Theory section 2.4.2. Before running through the Pan Tompkins method, the algorithm does a band pass filtering between 5 and 15 Hz exclude high frequency noise and low frequency muscle and respiration signals. The decision rule for finding R-peaks is based on a threshold that adapts to the signal.

After running through the whole recording, the program checks that no peaks were missed. The ECG plots were also visually supervised to make sure no peaks were missed, and no false peaks marked. Threshold values and minimum peak distance were changed from sample to sample to adjust for missed peaks.

A Matlab based program called HRVAS by John Ramshur [31] was used to process the RR-data and extract the HRV parameters from the RR-series. In all cases possible, segments with no artifacts were chosen. But when needed, an RR- interval was considered an outlier when it differed more than 30% from the previous interval. 20% was proposed as limit by Pradhan [25] as being the limit for normal intervals, but this proved to exclude normal intervals in the recordings so the limit was increased to 30%. The correction method used for outliers was removal, which is judged a simple and efficient correction method [26].

Detrending of the obtained NN-series was carried out with the Wavelet Packet, which is a method that leaves much of the wanted signal components while erasing low frequency nonstationarities [32]. The method used for frequency analysis was an autoregressive method with model order 16, which is the least suggested order according to Boardman [33].

Since according to theory, the resonant frequency depends on the amount of blood in the person’s system [7], the relationship between participant height and resonant frequency respectively participant weight and resonant frequency was investigated by making a correlation test. Another correlation test was made to see if exercise or coffee drinking had any correlation to effectiveness of the biofeedback session. The effectiveness was calculated in two different ways: absolute effectiveness (SDNN of 3e subtracted by SDNN of 3a) and relative effectiveness (SDNN of 3e divided by SDNN of 3a).

Coffee drinking was graded as:

0. Never 1. Sometimes 2. Daily

Exercise frequency was graded as:

0. Less than once a week 1. Once a week

2. Twice a week

3. More than twice a week

(34)

3.5 Statistical Analysis

In order to make a decision whether to reject or not reject the hypothesis (that the effect on HRV from biofeedback lasts longer than two hours), a statistic test was conducted. Assuming that the data was normally distributed, the ANOVA test was used on the control groups (1a-1e) and the biofeedback groups (3a-3e) of the parameters. The statistical testing of blood pressure change after BF was conducted with the same method.

18

(35)

Chapter 4

Results

4.1 HRV Results

Several HRV measures have been investigated to see if there was any significant change before or after biofeedback. The means and standard deviations of common HRV measures are shown in Table 4.1. Also mean Heart Rate (beats per minute) is displayed since a significant change was found in the control sessions in mean HR.

Data for each parameter and all subjects and can be seen in Appendix A.

The only parameter with a significant difference (p < 0.05) after biofeedback compared to before is SDNN. Only the baseline (3a) and 120 min after (3e) had significantly different SDNN means, no other combination of comparisons of session 3. Figure 4.1 shows a typical histogram of NN intervals before and 120 min after biofeedback, respectively. This illustrates an increased variability after biofeedback since the length of the intervals is less concentrated around the mean value, and more spread out.

Figure 4.2 shows SDNN of the control session and the biofeedback session recordings. SDNN of the control session is not increasing with significance. SDNN for the biofeedback session only had a significant increase 120 minutes after the biofeedback, it may well be that the effect remains for more than two hours afterwards.

Thus, the hypothesis cannot be rejected.

HF power increased significantly between measurement 1a and 1e control in the control session (Figure 4.3), and mean heart rate decreased significantly between 1a and 1e (Figure 4.4). Figure 4.5 and 4.6 shows typical PSD plots from one of the subjects just before (3a) and 120 minutes after biofeedback (3b), respectively.

Even though LF, HF and total power seems to be increasing, it turned out not to be significant when comparing all subjects.

4.2 Other Comparisons

Results of the blood pressure measurements are displayed in Table 4.2 with box plot representations in Figure 4.7. There was a decrease after biofeedback, but not

(36)

CHAPTER 4. RESULTS

Table 4.1. A summary of the HRV data aquired from all the recordings (Control session 1a-e, biofeedback session 3a-e)

Control session Biofeedback session

Measures 1a 1b 1c 1d 1e 3a 3b 3c 3d 3e

SDNN [ms]

Mean 62.56 72.01 84.08 87.11 88.47 61.78 75.31 75.65 74.24 86.66

Std 21.75 22.79 24.18 30.17 33.15 9.2 24.31 18.96 15.66 21.02

p 0.0942 0.0392

Mean HR [bpm]

Mean 79.33 75.11 71.87 71.1 67.99 74.68 72.89 73.28 70.18 69.84

Std 9.79 5.72 8.66 7.23 6.13 7.41 7.73 8.52 5.56 6.65

p 0.0084 0.4182

LF power [ms²]

Mean 618.9 749.8 861.9 1046.3 1321.1 499.9 872.5 767.8 658 989.5

Std 586.7 653.6 833.8 1015.2 1482.1 288.1 689.4 573.8 462.7 954

p 0.4379 0.3805

HF power [ms²]

Mean 239.8 387.5 483.8 562.2 659.7 293.3 344.7 330.4 378.7 560.9

Std 160.6 232.3 347.7 343.9 509.8 200.9 279.9 161.1 183.8 349.3

p 0.0395 0.0816

Total power [ms²]

Mean 903.1 1210.3 1416.3 1701.7 2070.5 835.9 1293.9 1168.8 1096.3 1636.4 Std 704.8 858.6 1127.4 1383.1 1946.3 356.1 941.2 707.3 587.2 1208.6

p 0.2231 0.2001

20

(37)

4.2. OTHER COMPARISONS

Figure 4.1. Histogram of NN intervals before and 120 minutes after biofeedback

Figure 4.2. SDNN box plots from the control session and biofeedback session

(38)

CHAPTER 4. RESULTS

Figure 4.3. HF power control session box plot

Figure 4.4. Mean HR control session box plot

22

(39)

Figure 4.5. PSD just before biofeedback (3a)

Figure 4.6. PSD 120 minutes after biofeedback (3e)

significant in either systolic nor diastolic blood pressure.

A comparison between weight and resonant frequency, respectively height and resonant frequency are seen in Figure 4.8. The correlation test as well as the plots suggest that there is no correlation. Neither was there any significant relationship between how well the participant responded to biofeedback (absolute or relative) and coffee drinking, nor how often they exercise (Figures 4.9 and 4.10). The p-values from all the correlation tests can be seen in Appendix A, Tables A.6 and A.7.

(40)

CHAPTER 4. RESULTS

Figure 4.7. Blood pressure box plots

Figure 4.8. Resonant frequency compared to participant height and weight

24

(41)

Table 4.2. Blood pressure results [mmHg], before biofeedback (1) and after biofeed- back (2)

Subject Systolic 1 Systolic 2 Diastolic 1 Diastolic 2

1 100 96 50 54

2 102 99 58 56

3 105 105 71 70

4 105 102 68 65

5 110 114 71 67

6 116 113 80 72

7 115 119 78 72

8 115 110 69 72

9 124 111 73 64

10 115 114 73 69

11 115 115 77 71

12 121 108 69 67

Mean 111.9 108.8 69.8 66.6

Std 7.5 7.0 8.4 6.1

p 0.31052 0.30246

Figure 4.9. Absolute BF effect compared to coffe drinking and exercise. Coffee drinking is graded as: 0=never, 1= sometimes and 2=daily. Exercise frequency is graded as: 0=less than once a week, 1=once a week, 2=twice a week and 3= more than twice a week.

(42)

CHAPTER 4. RESULTS

Figure 4.10. Relative BF effect compared to coffe drinking and exercise. Coffee drinking is graded as: 0=never, 1= sometimes and 2=daily. Exercise frequency is graded as: 0=less than once a week, 1=once a week, 2=twice a week and 3= more than twice a week.

26

(43)

Chapter 5

Discussion

The results suggest that a 20 minutes long biofeedback session can affect HRV in the acute timeframe, since SDNN was significantly increased. The increase progressed slowly, and was detectable two hours after the biofeedback session, which was the last measurement in this trial. Hence, no indication of how long that change lasts could be established. The participants had no prior training to resonant frequency breathing, so it was not as efficient as a well prepared session would have been, but still the SDNN elevation was significant.

A possible explanation for the detected change in HF power and mean HR in the control session is that the participants’ presumable initial nervousness for the trial decreased as the trial progressed. Since HF power reflects the parasympathetic influence of the heart rate, an increase in HF power would mean the person gets more relaxed. A decrease in mean heart rate suggests the same. 10 out of 12 trials were performed all in one day, starting with the control session, which can be an explanation to why the change is not seen in the biofeedback session; then they were already relaxed.

The blood pressure seemed to decrease after biofeedback, but not significantly.

Since resonant frequency breathing has proven to lower blood pressure [19], the explanation why those results didn’t show in this thesis could be because the participants’ blood pressures were already within the healthy interval, and thus didn’t get affected much. It could also be that it takes more than one 20 min session to get an effect.

5.1 Evaluation of Trial

An advantage of this trial was that a control session was performed, to ensure any change after biofeedback was actually due to the biofeedback. The number of subjects was also supposedly high enough to ensure reliable statistics.

There are however also drawbacks of the trial, a few are listed below:

• Only integers could be used as resonant frequency, due to shortages in the

(44)

CHAPTER 5. DISCUSSION software. Hence, the frequencies used in the biofeedback might not be the true resonant frequencies, but rounded to closest integer.

• The software was also very unreliable and sometimes the respiration curve was so uneven participants had trouble focusing on the breathing.

• Participants had only a short time to calm down before starting the measurements, and most of them had never done ECG recordings before. This may have had an inpact on the control session.

• The age span in the trial was not ideal, with a mean value of 28.8 years, which means that this trial cannot be expected to represent the whole adult span.

• Blood pressure was only measured instantaneously, and since blood pressure varies naturally with time, a more exact measure would have been a mean value of many consecutive measurements.

• A better method for compensating for artifacts than could have been used, than simple removal. That would compensate better for any peaks that were actually normal but judged by the program to be outliers.

• The measurements were performed with 200 Hz ECG sampling frequency, it should be at least 250 Hz for ECG measurements according to [23].

5.2 Evaluation of the Software

The software has a few problems that are already reported in one of the former theses [2]. The software application was also evaluated by the participants in this study, and many thought that the balloon interface (Figure 3.4) was not working well enough and preferred the line from the first interface (Figure 3.3) [5]. The balloon jumped too much and didn’t seem to adjust to the user’s breathing amplitude as well as the line. However, participants that did choose the balloon did it mostly out of amusement reasons, and that it was easier to remember when to inhale and when to exhale.

Most participants expressed difficulty at concentrating on the biofeedback for 20 minutes, making them bored and sleepy. Those moments that the graphics were working badly, participants even felt bad when looking at the screen, since their efforts didn’t seem to get registered. Many preferred only listening to a sound, as when finding the resonant frequency with the metronome, instead of biofeedback. Biofeedback is however a method which needs training to master, and the participants only engaged in it once.

Suggestions from participants included a more developed sound instead of graphics, for example a continuous sound that increases and decreases in frequency to guide in inhaling and exhaling. My own suggestion for the game interface is to make it more interesting by adding for example clouds of different sizes and col- ors along the dotted line, while the game counts how many clouds are caught, to encourage the users to stick to the line and keeping them more alert.

28

(45)

5.3. ETHICAL ASPECTS

5.3 Ethical Aspects

All participation in the trials was completely voluntary and all personal data has been kept anonymously. The participants could whenever they wanted, without reason, leave the trial. They were well informed about the risks and purpose of the trial, and all questions were answered.

Throughout the trial, participants were asked how they felt to ensure nothing was unpleasant. Resonant breathing can give dizziness symptoms, participants were told to be aware of those and in such cases breathe less deep. The currents running through the body during TEB measurements are so small they have no damaging effect. However, for safety reasons, there were requirements of the participant’s health status.

Participants could only take part in this study if:

• Their age was between 20 – 65 years.

• Their body mass index (weight/(height²)) was between 18 and 29 kg/m².

• They were functionally capable.

• They had no objections in performing common daily life activities as well as physical exercise.

They could not take part in this study if:

• They suffered from any chronic disease, such as diabetes, cardiovascular or pulmonary diseases.

• They might have been pregnant or lactating.

• They had nickel allergy or non-intact skin on the chest (e.g eczema, neuroder- matitis or sensitive skin).

• They had an active medical implant.

(46)

(47)

Chapter 6

Conclusion

This thesis investigated the effectiveness of a 20 minutes long biofeedback session with resonant frequency breathing, by trials on 12 volunteers. The hypothesis was that the effect from the biofeedback would last longer than two hours. The results suggest that a 20 min biofeedback session can elevate SDNN significantly, with the change taking place about two hours after biofeedback. Thus, the hypothesis cannot be rejected, but the result is too weak to strengthen it much.

The results from this trial could be a start of a bigger investigation of the time frames, to guide setting up protocols for biofeedback training for healthy people, with the goal to for example help stress levels down. The technique of resonant breathing is straightforward, and though it requires some practice to master and remain focused, the usability is high. However, further research is needed to con- clude the time window for acute HRV effects from biofeedback.

6.1 Future Research

The following are suggestions to further research in the area:

• Do a longer monitoring after biofeedback (more than two hours) to see how the parameters behave.

• Measure also baroreflex gain, which seems to be a parameter that can be affected in the long term by BF in healthy people [4].

• Measure also respiration during ECG recordings.

• Make a study on a greater number of subjects, to get better statistics.

• Make studies on the effect after different durations of BF instead of only 20 minutes.

• Make the same trials on non-healthy subjects, for example hypertension patients.

(48)

(49)

Bibliography

[1] Task force of the European society of Cardiology (correspondence: M. Malik).

Heart rate variability, standars of measurement, physiological interpretation and clinical use. European Heart Journal 17, 354-381. (1996)

[2] K. Lindskog. Development and Evaluation of HRV Biofeedback System. Mas- ter’s Thesis at KTH. (2014)

[3] M. Siepmann, V. Aykac, J. Unterdörfer, K. Petrowski, M. Mueck-Weymann.

A Pilot Study on the Effects of Heart Rate Variability Biofeedback in Patients with Depression and in Healthy Subjects. Appl. Psychophysiol. Biofeedback, vol 33, no. 4, pp 195–201. (2008)

[4] P. M. Lehrer, E. Vaschillo, B. Vaschillo, S. E. Lu, D. L. Eckberg, R. Edelberg et al. Heart Rate Variability Biofeedback Increases Baroreflex Gain and Peak Expiratory Flow. Psychosom. Med., vol 65, num 5, ss 796–805. (2003)

[5] A. Berndtsson. Heart Rate Variability Biofeedback Application for Android.

Master’s Thesis at KTH. (2013)

[6] F. Abtahi, A. Berndtsson, S. Abtahi, F. Seoane, K. Lindecrantz. Develop- ment and Preliminary Evaluation of an Android Based Heart Rate Variability Biofeedback System. Engineering in Medicine and Biology Society (EMBC), 2014 36th Annual International Conference of the IEEE (pp. 3382-3385). IEEE.

ISO 690 (2014)

[7] P. M. Lehrer, E. Vaschillo, The future of Heart Rate Variability, Biofeedback volume 36, issue 1, pp 11-14. (2008)

[8] P. M. Lehrer, B. Vaschillo, T. Zucker. Protocol for Heart Rate Variability Biofeedback Training. Associaton for applied psychophysiology & biofeedback vol 41 issue 3. (2013)

[9] P. M. Lehrer, E. Vashcillo, B. Vaschillo. Resonant Frequency Biofeedback Train- ing to Increase Cardiac Variability: Rationale and Manual for Training. Ap- plied Psychophysiology and Biofeedback, Vol. 25, No. 3. (2000)

[10] P. M. Lehrer, R. Gevirtz. Heart rete variability biofeedback: how and why does it work?. Frontiers in psychology volume 5 article 756. (2014)

(50)

BIBLIOGRAPHY

[11] E. H. Hon, S. T. Lee. Electronic Evaluations of the Fetal Heart Rate Patterns Preceding Fetal Death, Further Observations. Am J Obstet Gynec; 87: 814–26.

(1965)

[12] Markad V. Kamath, Mari A. Watanabe, Adrian R.M. Upton. Heart Rate Vari- ability (HRV) Signal Analysis. Taylor & Francis. (2013)

[13] M. T. La Rovere, MD; G. D. Pinna, MS; R. Maestri, MS; A. Mortara, MD; S.

Capomolla, MD; O. Febo, MD; R. Ferrari, MD; M. Franchini, MD; M. Gnemmi, MD; C. Opasich, MD; P. G. Riccardi, MD; E. Traversi, MD; F. Cobelli, MD.

Short-Term Heart Rate Variability Strongly Predicts Sudden Cardiac Death in Chronic Heart Failure Patients Circulation. 2003;107:565-570. (2003)

[14] L. Mourot, M. Bouhaddi, S. Perrey et al. Decrease in heart rate variability with overtraining: assessment by the Poincaré plot analysis. Clinincal Physiology and Functional Imaging 24. (2004)

[15] Amanda L. Wheat, K. T. Larkin. Biofeedback of Heart Rate Variability and Related Physiology: A Critical Review. Clinincal Physiology and Functional Imaging 35:229-242. (2010)

[16] D. Moss, F. Shaffer. Heart Rate Variability Training. Biofeedback foundation of Europe. (2009)

[17] Reviewed by D. Kiefer. Overview of Biofeedback.

http://www.webmd.com/a-to-z-guides/biofeedback-therapy-uses-benefits.

Found 2014-08-27. WebMD. (2014)

[18] P. M. Lehrer, E. Vashcillo, B. Vaschillo et al. Biofeedback treatment for asthma.

Chest journal vol 126 no.2. (2004)

[19] X. G. Lin, Q. Xiang, X. Fu et al. Heart Rate Variability Biofeedback Decreases Blood Pressure in Prehypertensive Subjects by Improving Autonomic Function and Baroreflex. The Journal of Alternative and Complementary Medicine volume: 18 Issue 2. (2012)

[20] Intercure, Inc. http://www.resperate.com/. Found 2014-10-24

[21] A. K. Gupta. Respiration Rate Measurement Based on Impedance Pneumogra- phy. Texas Instruments. (2011)

[22] D. S. Goldstein, O. Bentho, M. Y. Park, Y. Sharabi. Low-frequency power of heart rate variability is not a measure of cardiac sympathetic tone but may be a measure of modulation of cardiac autonomic outflows by baroreflexes. Exper- imental Physiology. (2011)

[23] G. D. Clifford. Signal Processing Methods for Heart Rate Variability. Ph D thesis. Department of Engineering Science, University of Oxford. (2002)

34

(51)

BIBLIOGRAPHY

[24] J. Pan and W. J. Tompkins A real-Time QRS Detection Algorithm. IEEE Transactions on Biomedical Engineering. Vol. BME-32, NO. 3. (1985)

[25] L. Pradhan and M. S. Islam. Replacing outliers with existing data in Inter Beat Interval Signal for Heart Rate Variability analysis. Mälardalen University, Västerås, Sweden. (2010)

[26] N. Lippman , K. M. Stein , B. B. Lerman Comparison of methods for removal of ectopy in measurement of heart rate variability. Am J Physiol, 267(1):H411- H418, (1994)

[27] H. J. Seltman. Experimental Design and Analysis. Carnegie Mellon Uni- versity. http://www.stat.cmu.edu/~hseltman/309/Book/PrefTOC.pdf. Found 2015-02-04. (2009)

[28] Computing Education Research Group, Monash University

http://www.csse.monash.edu.au/~smarkham/resources/anova.htm. Found 2014-12-30.

[29] S. A. Julious. Sample size of 12 per group rule of thumb for a pilot study.

Pharmaceutical Statistics., vol. 4, no. 4, pp. 287–291. (2005)

[30] H. Sedghamiz. An online algorithm for R,S and T wave detection. Linköping University.

http://www.mathworks.de/matlabcentral/fileexchange/45404-ecg-q-r-s-wave- online-detector. Found 2014-11-24. (2013)

[31] J. Ramshur. HRVAS. Matlab program. http://sourceforge.net/projects/hrvas/.

Found 2014-11-24. last updated 2014

[32] K. Shafqat, S. K. Pal and P. A. Kyriacou. Evaluation of two detrending tech- niques for application in Heart Rate Variability. Proceedings 29th Annual Inter- national Conference of the IEEE Engineering in Medicine and Biology Society EMBS, pp. 267–270 (2007)

[33] A. Boardman, F. S. Schlindwein, A. P. Rocha and A. Leite. A study on the optimum order of autoregressive models for heart rate variabilityPhysiol. Meas.

23 325. (2002)

(52)

(53)

Appendix A

Collected Data

Table A.1. SDNN data for all subjects [ms]

Subject 1a 1b 1c 1d 1e 3a 3b 3c 3d 3e

1 117.8 109.1 130.3 158.4 140.8 67.9 82.4 108 98.4 107.9

2 49.8 60.2 72.9 68.5 66.8 56.1 71.4 71.6 73.3 105

3 72.5 99.3 115.6 128.9 146 74.1 128.9 73.1 85.4 128.9 4 67.7 93.1 103.7 101.9 85.9 71.6 115.3 91 78.1 104.7

5 47 49.3 67.9 70.5 82.2 71.4 64.4 77 96.2 82.7

6 74.3 96.4 91 91.1 101.2 49.3 54 103.8 78.9 80.2

7 36.4 67.2 74.5 69 65.9 57.6 61.5 55.1 56.1 68

8 44.7 53.8 56.3 58.7 54.2 44.7 48.7 53.3 58.3 58

9 58.1 65.9 100.8 92.5 125.5 65.6 84.4 82.4 65.6 81.6

10 62.9 51.4 54.4 60 47 55.9 68.6 66.4 86.6 70.7

11 74.7 78.6 63.5 69.7 80.5 64 64.4 78.6 62.6 86.9

12 44.8 39.8 78.1 76.1 65.6 63.2 59.7 47.5 51.4 65.3 Mean 62.6 72 84.1 87.1 88.5 61.8 75.3 75.7 74.2 86.7

Std 21.8 22.8 24.2 30.2 33.1 9.2 24.3 19 15.7 21

(54)

APPENDIX A. COLLECTED DATA

Table A.2. Mean HR data for all subjects [bpm]

1 79 80.6 76.6 69.7 74.6 88.4 84.7 90.3 79.6 78.6 2 75.4 73.2 66 63.5 60.3 67.9 67.9 65.5 63.1 65 3 89.5 78.5 73 77.7 66.7 81.5 72.9 73 72.6 72.1 4 62.5 65.9 60.7 59.7 59 70.1 67.2 66.8 66.1 62.7 5 85.7 79.9 75.4 77.3 73.9 79.6 78.1 78.9 76.1 78.3 6 73.8 74.6 67 72.7 69.4 73.1 72.6 72 70.6 70.6 7 77.4 74.3 67.5 68.3 62.3 72 64.9 68.5 69 67 8 71.6 72.3 67.5 68.4 66.3 81.1 77.3 71.7 68.9 66.8 9 87 79.6 84.6 78.6 73.3 70.5 72.6 74.4 64.6 68.8 10 69.2 63.5 59.1 60.2 61.9 60.8 57.7 57.7 62.5 57.9 11 83.1 77.8 85.3 78.6 76.9 78.4 75.3 78.8 72.1 70.6 12 97.8 81.1 79.7 78.5 71.3 72.7 83.5 81.8 77 79.7 Mean 79.3 75.1 71.9 71.1 68 74.7 72.9 73.3 70.2 69.8

Std 9.8 5.7 8.7 7.2 6.1 7.4 7.7 8.5 5.6 6.7

Table A.3. LF power data for all subjects [ms²]

1 2114.1 1913.7 2270.4 3705.2 3143.8 670.3 1317.1 1138.1 1779.8 2067.6

2 343.8 213 170.4 267.6 345.2 201.7 277.5 181.2 472.9 901.2

3 954.4 1776 2621.6 1853.1 5143.5 1080 2395.8 1479.9 960.4 3597.7 4 533.9 1389.9 1123.1 989.2 1024.8 961.9 1151.7 748.3 851 883.6 5 584.3 503.9 743.6 1257.5 1261.3 565.7 772.5 558 622.9 738.7 6 403.5 1288.2 1287.7 1526.3 1791.1 254 255.9 1024.2 891.4 863.1

7 166.1 458.1 277 352.6 410.1 494 319.1 214 168 347.5

8 121.2 90.1 150.9 306.8 238.9 148.9 151.7 175.8 196.9 116.8 9 383.1 482.4 721.3 1354.9 1497.3 529 1701.2 1268.4 386.8 576.6 10 446.2 226.6 245.1 331.2 326.4 392.9 888.9 458.8 937.9 447.7 11 1304.6 432.1 572.7 297 344.9 354.2 970.5 1827.9 449.3 894.1 12 71.8 223.3 158.8 314.2 325.6 346.3 268.5 138.7 179.3 439.5 Mean 618.9 749.8 861.9 1046.3 1321.1 499.9 872.5 767.8 658.1 989.5 Std 586.7 653.6 833.8 1015.2 1482.1 288.1 689.4 573.8 462.6 954

38

(55)

Table A.4. HF power data for all subjects [ms²]

1 408.7 466.1 542.7 1268.6 591.2 174.6 193.9 568.7 639.7 564.1

2 136.8 300 619.1 720.7 852.1 494.5 420.1 460.1 500.3 404.2

3 278.9 849 1142.4 1110 2022.9 329.8 1163.3 612.9 296.4 1166.4

4 185.8 455.6 309.5 349.7 298 124 241.6 202.8 148.1 454.6

5 129.7 199.4 254.3 372.8 506.4 328.3 239.2 245.7 306.1 193

6 608.2 449.7 1148.2 580.3 655.4 369.8 220.7 459.9 637.4 766.7 7 106.2 150.8 202.1 215.2 356.5 175.8 300.9 179.9 230.2 264.2

8 300 382.3 618.2 374.2 440 103.4 208.2 173.2 563.3 641.3

9 150.9 259.2 334.7 686.2 1176.2 805.5 523.1 308 515.3 1223.2 10 153.1 214.6 239 182.4 172.3 162.3 262.7 349.5 242.1 353.2 11 373 785.4 255.1 603.5 576.9 142.1 263.5 275.3 333.7 570.6

12 46 138.2 140.1 282.8 268.8 309.1 99.2 128.6 131.9 129.7

Mean 239.77 387.53 483.78 562.17 659.71 293.26 344.7 330.37 378.69 560.92 Std 160.56 232.3 347.74 343.86 509.77 200.94 279.86 161.14 183.76 349.34

Table A.5. Total power data for all subjects [ms²]

1 2657.2 2541.3 3000.4 5236.9 3993 926.5 1639.2 1794.3 2577.1 2889.7 2 517.3 545.6 823.3 1026.9 1239.5 737.9 818.6 686.4 1048.5 1418.5 3 1281.5 2723.6 3875.3 3184.7 7259.3 1472.8 3646.5 2147.8 1312 4880.6 4 752.3 2043 1541.4 1447.9 1445.1 1186.4 1549.4 1031.7 1055.5 1442.6 5 740.9 731.7 1039.2 1696.3 1912.9 905.5 1066.1 888.8 992.7 973.8 6 1057.7 1888.3 2577.6 2273.7 2580.6 643.6 522.7 1592.4 1633.8 1722.3 7 288.1 667.2 522.1 598.6 800.4 704.8 641.5 413.5 431.3 645.6

8 447.5 491.2 813 721.5 703.1 262.7 378.5 366.2 782.6 773.6

9 576.5 799.1 1096.5 2148.9 2793.6 1368.1 2343.5 1634.9 946.3 1836 10 670.9 458.9 513.3 533.5 543.4 567.4 1224.4 847.6 1228.2 855.9 11 1721.9 1246.3 862.7 921.6 947.4 529.7 1286.8 2327.8 821.9 1592.9 12 125.4 387.6 330.6 630.4 627.2 725.7 409.6 293.8 325.7 605.4 Mean 903.1 1210.3 1416.3 1701.7 2070.5 835.9 1293.9 1168.8 1096.3 1636.4 Std 704.8 858.6 1127.4 1383.1 1946.3 356.1 941.2 707.3 587.2 1208.6

(56)

APPENDIX A. COLLECTED DATA

Table A.6. Correlation p-values for resonant breathing frequency comparisons

Comparison Correlation matrices Weight - resonant freq. 1 0.505

0.505 1

Height - resonant freq. 1 0.582

0.582 1

Table A.7. Correlation p-values for comparison between BF effect and coffee drink- ing, repectively BF effect and exercise habits.

Comparison Correlation matrices

Absolute change in SDNN Relative change in SDNN

Coffee - BF effect 1.000 0.188 1.000 0.731

0.188 1.000 0.731 1.000

Exercise - BF effect 1.000 0.695 1.000 0.258

0.695 1.000 0.258 1.000

40

(57)

Appendix B

Participation Questionnaire

Below are the questions from the form the participants filled in before the trial.

1. Name and surname 2. Birth date (YY-mm-dd) 3. Gender

• Male

• Female 4. Weight in kg 5. Height in cm

6. How often do you drink coffee?

• Daily

• Sometimes

• Never

7. Do you regularily use any type of tobacco?

• Smoking

• Snus

• Other

• None

8. How often do you exercise?

• Less than once a week

• Once a week

• Twice a week

• More than twice a week

9. Do you have any of the following conditions?

(58)

APPENDIX B. PARTICIPATION QUESTIONNAIRE

• Diabetes

• High blood pressure

• Any cardiovascular disease

• Nickel allergy

• Pregnant or lactating

• Active medical implant

• No, none of them

10. Are you fully functionally capable?

• Yes

• No

(59)

(60)

TRITA 2015:014

www.kth.se

A Pilot Study on the Effectiveness ofHeart Rate Variability Biofeedback onHealthy Subjects