Smartphones Could Help People Measure Blood Oxygen Levels at Home in a “Flash”

When we breathe in, oxygen-containing air fills our lungs. This oxygen is then transferred to our red blood cells, which then carry it throughout our bodies. Our bodies require a lot of oxygen to function, and healthy people constantly have an oxygen saturation level of at least 95%.

Ailments like COVID-19 and asthma make it more difficult for humans to absorb oxygen from the lungs. As a result, oxygen saturation levels fall to 90% or lower, signaling the need for medical intervention.

Doctors use pulse oximeters at a clinic to keep an eye on oxygen saturation. The clips you place over your fingertip or ear are pulse oximeters. However, there may be advantages to regularly checking oxygen saturation at home. For instance, it might assist patients in monitoring COVID symptoms.

Researchers from the Universities of Washington (UW) and California San Diego (UCSD) have demonstrated in a proof-of-concept study that cellphones are capable of detecting blood oxygen saturation levels as low as 70%. The U.S. Food and Drug Administration recommends that pulse oximeters be able to measure no lower than this figure (FDA).

Participants in the approach place their finger over the smartphone's camera and flash, which utilizes a deep-learning algorithm to determine the blood oxygen levels. A regulated mixture of nitrogen and oxygen was administered to six test volunteers during the experiment to artificially lower their blood oxygen levels. The smartphone accurately foresaw if the person had low blood oxygen levels 80% of the time.

Mobile devices against pulse oximeters

Utilizing pulse oximeters, those tiny clips that you place over your fingertip, is one method of determining oxygen saturation (some shown here in gray and blue). Researchers from the Universities of Washington and California San Diego have demonstrated in a proof-of-concept study that smartphones can detect blood oxygen saturation levels in a similar range to the standalone devices. Participants in the technique are instructed to place their finger over the smartphone's camera and flash. Credit: University of Washington/Dennis Wise

Other mobile applications that do this function required users to hold their breath during development. But after a minute or so, people start to feel very uneasy and need to breathe, and that's before their blood-oxygen levels have dropped sufficiently to accurately represent the whole range of clinically relevant data, said Jason Hoffman. He is a UW PhD student in the Paul G. Allen School of Computer Science & Engineering and a co-lead author. "We can collect 15 minutes of data from each person using our exam. According to our data, smartphones might function effectively in the critical threshold area.

The fact that practically everyone has a smartphone these days and can use it to measure blood oxygen levels is another advantage.

According to co-author Dr. Matthew Thompson, professor of family medicine at the UW School of Medicine, "this way you might have many measures with your own equipment at either no cost or minimal cost." "In a perfect world, a doctor's office could receive this information without any problems. To swiftly decide whether patients need to go to the emergency room or whether they can stay at home and relax until their next appointment with their primary care provider, this would be particularly helpful for telemedicine appointments or for triage nurses.

Ages of the six volunteers, who were chosen by the researchers, ranged from 20 to 34. Three were classified as male and three as female. While the majority of participants reported as being Caucasian, one individual identified as being African American.

The scientists asked each subject to wear a regular pulse oximeter on one finger, then position another finger on the same hand over a smartphone's camera and flash to collect data for training and testing the algorithm. This configuration was concurrently present on both hands of each participant.

According to senior author Edward Wang, who began this project as a doctoral student at the University of Washington studying electrical and computer engineering and is currently an assistant professor at the Design Lab and the Department of Electrical and Computer Engineering at UC San Diego, "the camera is recording a video: Every time your heart beats, fresh blood flows through the part illuminated by the flash."

According to Wang, who also serves as the director of the UC San Diego DigiHealth Lab, "the camera records how much that blood absorbs the light from the flash in each of the three color channels it measures: red, green, and blue." We can then feed our deep-learning model with those intensity measurements.

To gradually lower oxygen levels, each participant inhaled a regulated blend of oxygen and nitrogen. It took roughly 15 minutes to complete. The team collected more than 10,000 blood oxygen level values between 61% and 100% for all six subjects.

The researchers trained a deep learning algorithm to extract the blood oxygen levels using data from four of the participants. The remaining information was used to validate the methodology, which was subsequently put to the test on brand-new subjects.

Varun Viswanath, a UW alumnus who is currently a doctoral student under Wang's supervision at UC San Diego, commented on the study's co-lead authorship, "Smartphone light can get scattered by all these other components in your finger, which means there's a lot of noise in the data that we're looking at." "Deep learning is a very useful technique here because it can recognize these extremely intricate and nuanced traits and helps you find patterns that you wouldn't otherwise be able to perceive," the researcher said.

By putting the system to the test on more people, the team aims to continue this research.

According to Hoffman, one of his patients had significant calluses on their fingertips, which made it more difficult for the algorithm to correctly estimate their blood oxygen levels. "If we were to include more subjects in our study, we would probably see more calluses and more persons with various skin tones. Then, perhaps, we'd have an algorithm complex enough to better account for all these variations.

However, the researchers noted that this is a positive first step in creating machine-learning-assisted biomedical devices.

The need for such a study, according to Wang, is great. Traditional medical equipment is rigorously tested. We're all still learning, though, and computer science research is only now getting serious about utilizing machine learning for the creation of biomedical devices. We are forcing ourselves to learn how to do things correctly by forcing ourselves to be strict.

By UNIVERSITY OF WASHINGTON