Heart Rate Variability Logger - app details

Heart Rate Variability Logger is an app I developed to record, plot and export time and frequency domain Heart Rate Variability (HRV) features (as well as heart rate and RR intervals). The app is available for iPhone. This post is a short user guide where I cover in more detail the different features and some implementation choices.


App features:

• Extracts, plots, stores and exports heart rate, rr-intervals, time and frequency domain heart rate variability features (AVNN, SDNN, rMSSD, pNN50, LF, HF, LF/HF) 
• Configurable experience sampling for events annotation 
• Comparison between up to 3 recordings lets you get a better understanding of autonomic regulation and sympathetic/parasympathetic activity under different contexts in a glance 
• Step counter or accelerometer derived motion intensity for user context and activity tracking (step counter only for iPhones 5S) 
• Location tracking (lets you select between either GSM/WiFi to save battery power or GPS for high accuracy)
• Lets you filter data based on custom thresholds
• Shows heart rate in the app badge
• Configurable time window for features computation (choose between 30 seconds, 1, 2 or 5 minutes) 
• RR-intervals correction can be enabled to prevent ectopic beats or artifacts from affecting HRV features 
• Data export using iTunes or Dropbox 
• Tested for recordings longer than 24 hours 

I tried to cover also most of the questions that I got by email in the last few months. Please comment or email me if you have other comments or requests for future versions. More specifically, the post covers:

1. Introduction on HRV and applications
2. Hardware requirements
3. Quick start guide
4. Data acquisition and signal processing pipeline
5. RR-intervals correction
6. Features extraction
7. Data export and format

---

1. Introduction on HRV and applications

The cardiovascular system is mostly controlled by autonomic regulation through the activity of sympathetic and parasympathetic pathways of the autonomic nervous system. HRV analysis attempts to assess cardiac autonomic regulation through quantification of sinus rhythm variability. The sinus rhythm times series is derived from RR intervals (R peaks of the QRS complex derived from the electrocardiogram (ECG)), by extracting only normal (NN) intervals. 

Since HRV aims at quantifying autonomic regulations, it can be used as marker of sympathetic or parasympathetic predominance, and therefore become relevant in many applications. In athletes for example [1] heavy training is responsible for shifting the cardiac autonomic balance toward a predominance of the sympathetic over the parasympathetic drive, and HRV analysis attempts to quantify this shift (see this post for more on HRV for training). 

Other applications span from stress monitoring during public speaking [2] or daily life [3], assessment of pathological conditions [4] and even emotional regulation during financial decision-making (trading) [5].

If you are new to HRV, check out our Ultimate Guide to Heart Rate Variability (HRV).

2. Hardware requirements

Heart Rate Variability Logger requires a Bluetooth Low Energy (also called Bluetooth Smart or BLE or 4.0) heart rate monitor. Personally, the most reliable I tried is Polar's H7 which we also validated in this paper. The first iPhone to support BLE was the 4S.

3. Quick start guide

Open the app, and push the Bluetooth icon to connect to your heart rate monitor. No pin or passcode is required to connect to Bluetooth low energy monitors, the only thing you need to do is to turn on your Bluetooth radio. The app will prompt you in case your Bluetooth radio is off. Once you are connected, the real-time view will be enabled and HRV logger will start processing RR-intervals and computing features. More details in this slideshow (includes info on experience sampling, activity and location tracking):

4. Data acquisition and signal processing pipeline

The following block diagram represents the signal processing pipeline implemented in HRV logger. The single blocks are explained in more detail in sections 5-6.

5. RR-intervals correction

RR-Intervals correction prevents artifacts due to ectopic beats or motion from affecting features computation, as often reported in literature for HRV analysis. 

It is advised to keep RR-interval correction to 20%, meaning that every RR-interval which differs by more than 20% from the previous one, will be discarded. HRV Logger provides three options, the correction can be completely disabled, in case you are interested in looking at PVCs or other events. By default the correction is set to 20%. The third and last setting is 50%, which is a less aggressive correction, since every RR-interval which differs by more than 50% from the previous one, will be discarded. 

6. Features extraction

The time window used to compute features can be configured choosing between 30 seconds, 1, 2 or 5 minutes. Windows of at least 2 minutes should be used when using low frequency features (LF) [2]. Given the non-constant frequency at which RR-intervals are received, a buffer of RR-intervals is used to collect 30 seconds to 5 minutes of data, on which features extraction is executed.

The following features are extracted from RR-intervals:

• Mean of beat-to-beat intervals (AVNN),
  
• Standard deviation of beat-to-beat intervals (SDNN), 
  
• Square root of the mean squared difference of successive RRs (rMSSD),
  
• Number of pairs of successive RRs that differ by more than 50 ms (pNN50),
  
• Low frequency power (LF, 0.04-0.15 Hz),
  
• High frequency power (HF, 0.15-0.40 Hz),
  
• Low to High frequency power ratio (LFHF).

Additionally, Motion Intensity is computed from accelerometer data.

The term NN (e.g. AVNN) is used instead of RR to emphasize the fact that the beats used to extract HRV features are sinus beats (this is also why it is important to keep the RR-interval correction enabled). Assuming we have $n$ beats in our buffer:

RR-intervals
Time-domain features
AVNN, mean of NN intervals:
\[AVNN = \frac{1}{n} \sum_{k=1}^n NN_k\]
SDNN, standard deviation of NN intervals:
\[SDNN =\sqrt{\frac{1}{n} \sum_{k=1}^{n} (NN_k - AVNN)^2}\]
where AVNN is computed as above.
rMSSD, square root of the mean squared difference of successive NN intervals:
\[rMSSD = \sqrt{\frac{1}{n-1} \sum_{k=1}^{n-1} (NN_{k+1}-NN_{k})^2}\]
pNN50, number of pairs of successive RRs that differ by more than 50 ms
The difference between beats is calculated as above:
\[NN_{k+1} - NN_{k}\]
Then, if $n50$ is the number of beats for which we have a difference greater than 50 ms ($(NN_{k+1}-NN_{k}) > 50$), pNN50 is computed as:
\[pNN50 = \frac{n50}{n} 100\]

Frequency-domain features
Since RR-intervals come at non-constant frequency, they are linearly interpolated before frequency analysis. After interpolation, a hamming window is applied before performing the FFT. 

Accelerometer
Motion Intensity
Motion Intensity (MI) is computed after filtering the accelerometer data to remove the static effect due to gravity (basically isolating dynamic movement only). Accelerometer is sampled at 2Hz, to prevent battery drain. Assuming $n$ filtered accelerometer samples, MI is computed as follows:
\[MI = \frac{1}{n} \sum_{k=1}^n ACC_{k}  \]
where $ACC_{k}$ is the sum of the absolute signal over the three axis:
\[ACC_{k} = abs(ACCx) + abs(ACCy) + abs(ACCz) \]
MI should be use simply to get an indication of the user movement when carrying the phone. Especially in situations where HR and HRV can be affected by both movement and stress, it can help in acquiring more context.  


For iPhones 5S motion intensity is replaced by the actual steps taken, computed using the M7 co-processor. 

Location Tracking
Location tracking is available since version 4.0. By enabling location tracking you can add even more context to your HRV recordings. If coarse location information is sufficient, do not enable the high accuracy location tracking, since it will drain more battery (uses GPS).

7. Data export and format

Data can be exported either by connecting your iPhone to iTunes, or by connecting HRV Logger to your Dropbox account. Four files are exported for each recording:

• date_HR_recording_name.csv. download example.
  
• date_RR_recording_name.csv. download example.
  
• date_Features_recording_name.csv. download example.
  
• date_Events_recording_name.csv. download example.
  

All files are in csv format, making it easy to import them in other tools. The example files include approximately 2 hours, split in half sedentary and half running.

References

[1] Aubert, André E., Bert Seps, and Frank Beckers. "Heart rate variability in athletes." Sports Medicine 33.12 (2003): 889-919.
[2] Kusserow, Martin, Oliver Amft, and Gerhard Tröster. "Analysis of heart stress response for a public talk assistant system." Ambient Intelligence. Springer Berlin Heidelberg, 2008. 326-342.
[3] Vrijkotte, Tanja GM, Lorenz JP van Doornen, and Eco JC de Geus. "Effects of work stress on ambulatory blood pressure, heart rate, and heart rate variability." Hypertension 35.4 (2000): 880-886.
[4] Nolan, James, et al. "Prospective study of heart rate variability and mortality in chronic heart failure results of the United Kingdom heart failure evaluation and assessment of risk trial (UK-Heart)." Circulation 98.15 (1998): 1510-1516.
[5] Fenton-O'Creevy, Mark, et al. "Emotion regulation and trader expertise: Heart rate variability on the trading floor." Journal of Neuroscience, Psychology, and Economics 5.4 (2012): 227.
[6] Berntson, Gary G. "Heart rate variability: Origins, methods and interpretive caveats." Psychophysiology 34 (1997): 623-648.