Photoplethysmography (PPG) is an optical technique used to detect changes in blood volume. This data is useful for monitoring a range of cardiovascular health measures — like blood pressure, heart rate and arterial oxygen saturation — that can predict serious conditions, like heart disease.
The use of PPG sensors is uncomplicated and inexpensive, making it a popular option for doctors wanting to track patient cardiovascular health. However, the tech does have some limitations and room for improvement. Recent innovations are using the pattern-finding abilities of AI to overcome some of these issues and make PPG sensors even more valuable.
Video-Based Heart Rate Monitoring With AI
A new study from researchers at the Chinese Academy of Sciences recently explored the use of AI, PPG and video to estimate patient heart rate.
As your heart beats, changes in blood volume create extraordinarily subtle changes in your skin color. These color shifts aren't visible to the human eye, but a sufficiently trained AI algorithm can pick up on them and estimate patient heart rate.
The project achieved an error rate of less than five beats per minute when tested against a database containing more than 2,000 visible light videos and 700 near-infrared videos of 107 test subjects. To check how well their model could handle sub-optimal video footage, the team curated this database so it included high-quality video of test subjects. It also scrutinized video with poor lighting conditions and significant head movement.
The technique provides a lighter weight and a more convenient approach to heart rate monitoring, which is typically achieved with patient monitors that require contact to work. A video-based PPG could also be effective in situations where doctors can't be near patients, but still need essential patient health information.
The tech can only be used to estimate patient heart rate, but the team behind the project hopes to branch out shortly. Possible future experiments may apply the model to other physiological measurements, like breath rate.
Measuring Blood Pressure With Neural Networks and PPG Waveforms
High blood pressure is the leading cause of death throughout the world and can have serious health implications if left uncontrolled. Because of this, it's standard for doctors to take blood pressure, especially if that patient is at risk of hypertension.
Typically, blood pressure is measured using catheters, or non-evasively with cuffs. Both of these methods can be uncomfortable for patients and impractical in some cases. They only provide a snapshot of blood volume that may not accurately reflect fluctuations in blood pressure over the course of a day.
With PPG-based blood pressure estimation, doctors can take these measurements noninvasive. Depending on how the PPG signal is recorded, it may also be possible to update these measurements throughout the day.
To monitor blood pressure with PPG, however, you need to use a hybrid approach that requires both a PPG and ECG sensor. This has some significant limitations. It requires two different sensors that need to be placed at fixed locations on the body and held there while the readings are taken, which can be uncomfortable or difficult to maintain for some patients. PPG and ECG sensors are also highly sensitive to motion artifacts, and data from both sensors will need to be processed before it can be analyzed.
A new approach to PPG-based blood pressure estimation uses an artificial neural network to pull important waveform features from a PPG signal and estimate blood pressure without an ECG.
This use of a neural network provides some advantages over traditional methods of waveform feature extraction, which can fail to account for things that vary from patient to patient. One study found that blood pressure models generated by neural networks were much more accurate than conventional ones and could provide doctors with better blood pressure readings.
Got a computer vision project?
Hire us for your computer vision or machine learning projects. Learn more about computer vision and machine learning and read our computer vision case studies.
The use of this approach could make taking blood pressure both simpler and more inexpensive for doctors while maintaining accuracy. The recent proliferation of consumer-grade PPG sensors — both the Apple Watch and newer Fitbits come equipped with PPGs — may also make it easier for doctors to get these measurements. In the near future, it may be possible for patients to remotely provide doctors with blood pressure information, with no visit to the doctor's office necessary.
Researchers Are Improving PPG With Artificial Intelligence
PPG sensors are essential equipment for doctors wanting to quickly and non-evasively take critical cardiovascular health measurements like blood pressure and heart rate. New applications of AI tech to these sensors may make them even more valuable over the next few years, allowing doctors to measure patient heart rate with video or measure blood pressure without an ECG.
Author Bio:
Jenna Tsui is a technology journalist with writing experience in future & disruptive technologies, AI, medtech, and scientific development. To see more of Jenna's work, visit The Byte Beat , follow her on Twitter or check her out on LinkedIn.
Image credit: Gerd Altmann from Pixabay
![](https://www.facebook.com/sharer/sharer.php?u=http://uniwebb.com/blog/post/photoplethysmography-is-improving-thanks-to-ai/1799/&p[images][0]=http://uniwebb.com/in/images/photoplethysmography-is-improving-thanks-to-ai/1799.png>
<i class="fa fa-facebook"></i>
</a>
</div>
<div class="rrssb rrssb-twitter">
<a class="social-icon si-borderless si-twitter popup" href=){kind=link}
Leave a Comment
comments powered by Disqus