Who, what & why?

ithlete finger sensor validation in practiceIn an earlier blog post Measurement of HRV we described how heart rate variability (HRV) measurements were originally performed using the Electro Cardiograph (ECG or EKG) and then later by Polar chest straps once those had been validated for beat to beat accuracy in the 1990s.

The ithlete Finger Sensor uses the principle of photoplethysmography (PPG), which is a fancy term for measuring how blood pulses through the body’s arteries by shining light (usually infrared) through it. The Finger Sensor is closely modeled on the medical monitoring version used in all hospitals, but with the processing performed in your phone instead of a bedside monitor.

The Finger Sensor has been used in several research studies, and was validated originally by Dr James Heathers for use in a community health study in East Africa. Now, sports scientists Dr Mike Esco and Andrew Flatt have taken an in-depth look at the accuracy of the Finger Sensor in an athletic population.

What did they do?

They recruited 30 college age men and women all of whom were active, and some US national standard collegiate athletes.

All participants wore a medical ECG (Biopac MP100) as well as the ithlete Finger Sensor. At least one minute was allowed for heart rate stabilization before the ithlete app measurement was performed. The test was repeated in the lying (supine), seated, and standing positions, since all of these are popular with HRV users for different reasons. The comparison was made using the ithlete (20x LnRMSSD) scale with time-synchronized data from the Biopac ECG.

What did they find?

They found very good agreement between the mean values of HRV in all three positions:


All positions had correlations of 0.98 or higher to the gold standard ECG.

Bland-Altman plots are a standard way of revealing random and systematic differences (such as proportional biases) between two different methods, and these are shown below for the three positions:

Charts collated

What does it mean?

What we are looking for is low bias and tight limits of agreement, as this means we can rely on the Finger Sensor measure being a reliable substitute for the gold standard ECG, and ideally one that can be used interchangeably.

The Standard Error of Estimate averages <1.7 over all three positions, so in practical terms if you take a large number of simultaneous HRV readings you can expect there to be up to a 1.7 difference between the Finger Sensor and chest strap two thirds of the time.

When compared to the Fitbit HR accuracy study, this is significantly better, since the Fitbit devices could only manage a Standard Error of Estimate of 17 bpm for heart rate (HR), a much less demanding measure than HRV http://www.myithlete.com/blog-how-accurate-are-consumer-pulse-sensors-part-2-the-fitbit-case/

Why measure in 3 positions?

Its worth also mentioning why the Authors chose three different positions for the comparisons. In their words

While the supine position is well tolerated in athletes, it does not promote the most appropriate position for morning measures since the subject may be at-risk of falling back to sleep. In addition, extremely low heart rates with concurrent low HRV has been noted as parasympathetic saturation of the sinoatrial node may be present when an elite aerobically trained athlete assumes a supine position. On the other hand, seated and standing positions may be more sensitive for tracking changes in autonomic activity in response to training in elite athletes.

Overall, we recommend the seated position when using the finger sensor, as it is comfortable, and you can rest your hand on a firm surface to ensure the minimum of motion disturbances (to which PPG technology is sensitive).

Full Paper