Remote Heart Rate Sensors Can Be Biased Against Darker Skin. UCLA Team Offers Solution

Remote Heart Rate Sensors Can Be Biased Against Darker Skin. UCLA Team Offers Solution

Along with the growing popularity of telemedicine, there are devices that allow people to take their vital signs from home and send the results to their doctors via computer. But in many cases, getting accurate remote readings for people of color is an ongoing challenge.

Take, for example, camera-based remote heart rate measurements that detect subtle changes in a patient's facial color due to fluctuations in blood flow under the skin. As part of an emerging class of remote sensing technology, the device has difficulty reading color changes in darker-skinned people, said Achuta Kadambi, an associate professor of electrical and computer engineering at UCLA's Samueli School of Engineering. .

Kadambi and his team have now developed a remote diagnostics technique that addresses this hidden bias toward darker skin by making heart rate readings more accurate for patients of a variety of skin tones. Your secret? It combines light-based measurements from the camera with radio-based radar measurements.

The researchers presented their findings, recently published in the journal ACM Transactions on Graphics, at the SIGGRAPH 2022 conference in Vancouver, British Columbia. Both virtual and in-person conferences are hosted annually by members of the Computing Machinery Association.

The researchers say the advances could lead to a new class of high-performance medical devices and more accurate and fair remote technology, allowing doctors and health care systems to monitor patients remotely, both in clinical settings as well as from the patient's home. .

"To a large extent, this work shows that practical and innovative engineering solutions can overcome persistent biases in medical devices," said Kadambi, a fellow at the California Nanosystems Institute at UCLA. "But first you need to understand that this kind of bias means that today's best technology may not be the best for everyone. Through careful design, we can find the right solutions that work just as well or even better."

The UCLA team's combination of the two techniques shows a promising way to accomplish this, said Kadambi, who is also an associate professor of computer science and the study's principal investigator. As head of the Visual Machines Group at UCLA, he has written about different types of bias in medical devices and how to deal with them.

In developing their new technology, the researchers first showed that the remote sensing device itself was a source of bias, showing in their paper that high levels of melanin, a natural skin pigment, interfered with what is known as photoplethysmography or PPG. , the signal used in today's camera-based remote heart rate measurements.

The PPG signal is also used to measure heart rate by pinching the patient's finger pulse oximeters, as well as applications compatible with some commercial wearable devices and smart watches. This device illuminates the skin and detects changes in the amount of light reflected from the bloodstream just below the surface. The reflected light creates a PPG signal, a measure of the patient's heart rate.

Previous attempts at skin tone distortion in such technologies have generally sought to improve upon the basic standard by adding additional programming or using a wider variety of skin tones. But neither of these approaches addresses the real problem, which is the physics of the device itself, Kadambi said.

Instead, the UCLA researchers turned to another technology that can estimate heart rate: radar. At 77 gigahertz, the radar can detect subtle changes in chest displacement from heart rate. And although this method solves the problem of skin color distortion, it is less reliable than the PPG signal. However, they were successful in combining these two different detection modes, camera and radar, and using machine learning to improve them so that they work together.

In tests with 91 people, the researchers showed that their camera-radar system outperformed remote camera-based PPG in terms of measurement accuracy and fairness across a wide range of skin types.

"Multimodal remote healthcare has the potential to make devices brighter not only for skin tones, but for many characteristics, such as body mass index, gender, and various health conditions," he said. Aleksandr Vilesov, a graduate student in electrical and computer engineering, is from UCLA. and is one of the main authors of this article. "Most of these aspects have not been fully investigated, and some of our future studies will try to understand these biases."

The researchers suggest that such capital-based improvements could be made in other types of technology, such as thermal, acoustic, near-infrared and light polarization sensors.

"The COVID-19 pandemic reveals the need for new technologies for clinicians and care teams to monitor their patients remotely," said study co-author Dr. Laleh Jalilian, an assistant clinical professor of anesthesiology and medicine. perioperative at UCLA Health. "Since the beginning of our collaboration, the primary goal has been to develop medical technology that works fairly and with high precision on patients of all skin colors, because it will give clinicians the confidence to make high-quality medical decisions."

UCLA electrical and computer engineering students Pradyumna Chari and Adnan Armouti are also the paper's lead authors. Other authors of the paper are UCLA electrical and computer engineering graduate students Anirudh Bindiganavale Harish, Kimaya Kulkarni, and Ananya Deoghare, all members of the Visual Machinery Group.

More information: Alexander Vilesov et al., Integrating 77 GHz camera and radar sensing for fair and robust plethysmography, ACM Transactions on Graphics (2022). DOI: 10.1145/3528223.3530161

Quote: Remote heart rate sensor can be hacked on dark skin. UCLA Team Proposes Solution (Aug 24, 2022) Retrieved Aug 24, 2022 from

This document is subject to copyright. No part may be reproduced without written permission, except in any fair trade for research or personal study purposes. The content is provided for informational purposes only.

5 ways to help remote teams succeed!

Post a Comment (0)
Previous Post Next Post