My research lies at the intersection between autonomic physiology, electrocardiography, biological psychology and signal analysis. It won’t make much sense unless you know what heart rate variability is and how it works. It’s best organised by topic:
1) Measurement and analysis in heart rate variability (HRV)
Heart rate is cheap and easy to measure, and requires no conscious attention to monitor (you can record it while your attention is absent or elsewhere – during exercise, during stress, during sleep). As a consequence, HRV-based techniques are popular.
This ease of access is a double-edged sword.
While this democratization means that literally everyone has access to HRV, it also means that standards for data collection and analysis get eroded through the sheer volume of researchers and companies using it.
(Social and sports scientists are particularly guilty of this.)
This research attempts to quantify and predict the depth and severity of problems with measurements environments and in individual pieces of research.
Heathers et.al. (2015) – The Elusory Upwards Spiral: A Re-analysis of Kok et.al. (2013). Psychological Science.
Quintana and Heathers (2014) – Considerations in the assessment of heart rate variability in biobehavioral research. Frontiers in Psychology.
Heathers (2014) – Everything Hertz: methodological issues in short-term frequency-domain HRV. Frontiers in Physiology.
Quintana, Heathers, Kemp (2012) – On the validity of using the Polar RS800 heart rate monitor for heart rate variability research. Arbeitphysiologie.
Heathers (2012) – Sympathovagal balance from heart rate variability: an obituary. Experimental Physiology.
2) New techniques in content extraction from ECG
Normal HRV indices are calculated from R-R intervals, which correspond to the ventricular depolarisations of each heart beat. While this obviously works, any normal heartbeat is also full of other pieces of unused information.
Other segment timings drawn from sino-atrial, atrioventricular and ventricular repolarisation indices (roughly corresponding to the P-wave, PR-segment and QT-interval respectively) are almost entirely ignored in the behavioural sciences. So are the different techniques which can be used to approximate the respiratory cycle (like the RS-length, R-wave angle, etc.) Likewise, the morphological components (that is, changes in ECG shape rather than timing… P-waves are surprisingly dynamic, for instance).
Will any of these work in an ‘applied’ or non-invasive context? Can we extract other predictive variables from the heartbeat which will correlate to everyday ‘applied’ contexts – attention, stress, mood, exercise status, etc.?
The answer is “probably”. This question, which is really a ‘physiology’ rather than an ‘engineering’ question, is less aggressively pursued than new techniques which are continually developed to measure RR-intervals.
3) Massive Samples and HRV
I am very concerned that the statistical and explanative power of HRV is somewhere between ‘not ideal’ and ‘totally inadequate’. This is not the case in heavy-duty clinical studies – these have epidemiology-sized samples – but for laboratory experiments.
There is no reason for this to continue. Heart rate monitoring is very, very cheap these days (did you know you can build an ECG from about $7 worth of electronics store parts?)
More importantly, the community at large has never been more engaged in both science as a general interest and personal data collection – elements of Quantified Self are rapidly becoming mainstream, spurred on by continual changes in hardware platforms.
In short, there are very few practical barriers to recruiting MASSIVE samples for investigating HRV phenomena. So, why aren’t researchers doing it?
1) lack of imagination
2) low crossover with commercial motivations
3) not a lot of platforms for secure ‘big data’ sample sizes (although we have ResearchKit now…)
4) you need a very cheap HR monitor
My approach to this problem is based on a stripped-down version of the ithlete system, which we have tested both in the lab and in field samples with hundreds of participants. It’s essentially a photoplethysmograph which runs from an iPhone or iPod, and has no learning curve whatsoever to use.
With incremental improvements over time to the data capture, we are in a position where the measurement precision is off by about 2 to 5%, but the cost for an individual monitor is about a hundred-fold less than lab equipment. Myself and others have recorded some truly massive samples, much bigger than anything previously attempted. This is not only a good research tool, but potentially a great educational tool – we can now teach cardiac physiology or biological psychology by actually doing it, and provide worked examples.
Heathers et.al. (2014) – Smartphone Platform Survey-Scale Heart Rate Collection: a Performance Evaluation in Ethiopia. Wireless Health ’14.
Heathers et.al. (2014) – Blood volume pulse (BVP) derived vagal tone (VT) between 5 and 7 years of age: a methodological investigation of measurement and longitudinal stability. Dev Psychobio
Heathers (2013) – Smartphone-enabled pulse rate variability: an alternative methodology for the collection of heart rate variability in psychophysiological research. IJP.
Future work / stuff in progress / scary questions:
- HR data isn’t HR information… how should we structure, reduce and understand massive datasets of physiological information?
- All the powerful techniques of formal electrophysiology used to diagnose arrhythmia (intracardiac electrograms, signal-averaging, tilt, etc.) aren’t used to inform single-lead ECG techniques. Can we better illuminate normal HRV by comparison to better signals?
- There are now some super-accessible HR recording methods – specifically, from a camera/webcam either by occluding the lens with your finger (making a kind of mock-plethysmograph) or by Gaussian magnification methods. Cameras have quite a slow polling rate, not usually above 60Hz – so how can we improve the accuracy of these highly accessible signals?