MORE

This app can detect dangerously low blood oxygen levels

One day, patients struggling to breathe in bed with a respiratory condition like asthma or COVID-19 may have a diagnostic solution at hand…literally. A team of scientists has developed a camera-based blood oxygen sensor that only requires a smartphone and a finger.

It remains to be seen whether the results will hold up in larger studies or whether the new method will be able to avoid the well-known skin tone biases of commercially available pulse oximeters. However, the researchers see their method as a promising and accessible alternative to tools that warn patients of dangerously low blood oxygen levels – a condition called hypoxemia.

“Our results in this pilot study of [six] subjects, provide a positive indication that a smartphone could be used in the future to assess risk of hypoxemia without additional hardware,” the authors write in the study, published Monday in the journal npj Digital medicine. They added that further research could lead to a cost-effective way to treat chronic respiratory conditions like asthma and COPD, as well as acute conditions like COVID-19.

Existing smartphone-based oximetry — a term used to describe the process of measuring one’s oxygen levels — has been deemed unreliable and inaccurate compared to traditional pulse oximeters, which shine light through a person’s finger and measure blood oxygen levels based on it calculate how much of it light goes through. And many smartphone methods require the user to hold their breath for long periods of time, which can be uncomfortable or impossible. The scientists set out to build a system that relied on smartphone video captured with the flash on of a person’s finger while they were breathing normally. A deep learning model would then calculate blood oxygen levels based on the video.

The study’s six participants strapped on masks and breathed a mixture of oxygen and nitrogen for about 15 minutes while the oxygen levels were slowly lowered. They placed one of their fingers in a conventional pulse oximeter and another on a smartphone camera. Data from four of these participants was used to train the model, which then, based on the videos, predicted blood oxygen levels for the remaining two participants. These results were compared to the pulse oximeter readings.

When the smartphone camera method classified the readings as being below 92 percent blood oxygen saturation (a common yardstick used to advise patients to go to the hospital for possible hypoxemia), it was wrong 22 percent of the time for all six participants. When it classified readings as over 92 percent, it was wrong 14 percent of the time compared to the pulse oximetry data.

While these results mean that this method is not yet ready for the clinic, the researchers hope that future work will build on this technique. Training the model on a large and diverse data set can improve its accuracy, especially for people with thick fingertip skin and people of color who are currently not well served with pulse oximeters due to the two wavelengths of light used by the devices. Follow-up studies may also consider comparing the model’s predictions to arterial blood gas readings, which, unlike pulse oximetry data, have not been shown to be racially biased.

“One of our subjects had thick calluses on their fingers, which made it difficult for our algorithm to accurately determine their blood oxygen levels,” lead author and University of Washington computer scientist Jason Hoffman said in a press release. “If we expanded this study to include more subjects, we would likely see more people with calluses and more people with different skin tones. Then we could potentially have an algorithm with enough complexity to better model all these differences.”

https://www.thedailybeast.com/this-app-may-detect-if-you-have-dangerously-low-blood-oxygen?source=articles&via=rss This app can detect dangerously low blood oxygen levels

Source link

Related Articles

Leave a Reply

Your email address will not be published.

Back to top button

Adblock Detected

Please disable ad blocking software and refresh the page.