Pulse rate prediction using facial vision

Hypertension is one of the leading causes of cardiovascular diseases. On average, 1 in every 4 humans is a victim of hypertension. This symptom-less chronic metabolism disease usually goes unnoticed and at times untreated owing to the busy lifestyle of today’s population that puts its health on the backburner.

One thing us humans are always surrounded by are smartphones. Days begin with our smartphones and end with our smartphones. It has become the most important part of our lives because everything is stored on that device.

The prevalent procedure of measuring one’s blood pressure accurately is predominantly cuff-based. The physical bulk and the costly nature of this equipment make it difficult to use it on a regular basis without a doctor’s supervision. This causes the simple process of measuring blood pressure come off as a tedious task in such a digital era.

To make matters convenient, researchers have been on a constant pursuit to look for ways to make the task of taking blood pressure more accessible and easier, which could potentially save many lives.

The Department of Developmental Neuroscience at the University of Toronto, with Kang Lee as the lead researcher has developed a technology which can accurately measure blood pressure using a smartphone within 30 seconds.

This technology, more commonly referred to as transdermal optical imaging, makes use of Photoplethysmogram (PPG) signal extracted from a face video for BP and Heart Rate estimation. The entire process of estimation of BP takes the form of a step by step process.

Natural light can penetrate just below the skin’s surface which in invisible to the naked eye but can easily be viewed in smartphones. By keeping the camera fixed at roughly half a metre distance from the person's face, a face video of the subject is taken (under ambient lighting settings).

The individual is requested to sit still with their eyes closed for one minute while the video is being filmed. With the help of algorithms, the PPG signal is extracted in the form of pulsation information based on the intensity variation in the blood concentration underneath the skin. This pulsation information or pattern is then used to predict the blood pressure.

Lee and his colleagues used an iPhone with transdermal optical imaging software to capture two-minute videos of the blood flow of 1,328 Canadian and Chinese people. The researchers compared blood pressure readings utilising a standard cuff-based continuous blood pressure measurement system to systolic, diastolic, and pulse pressure measures taken from smartphone recordings.

The data was utilised by the researchers to train the technology to reliably calculate blood pressure and pulse from face blood flow patterns. Transdermal optical imaging predicted systolic blood pressure with approximately 95% accuracy and diastolic blood pressure with nearly 96 percent accuracy, according to the researchers.

There are still a few kinks to be worked out before this technology can be practically applied by the general masses. If future research can confirm this promising result in hypertensive individuals and with video camera readings taken in everyday life, getting blood pressure information with a click of a camera could become a reality.