The technique documented here is very simple, it involves the patient placing their finger on the camera lens, and the smartphone flash through a deep learning (AI) algorithm deciphers the oxygen levels in the blood from blood flow models.
Hypoxemia occurs when the blood does not carry enough oxygen to adequately supply the tissues, leading to hypoxia, characterized by an abnormally low level of oxygen in the tissues. Today, the measurement of oxygen levels in the blood can be done by:
- an invasive method, which requires taking blood from an artery and measuring arterial oxygen saturation (SaO2);
- a non-invasive method, using a small device placed at the fingertip called a hypoxemia pulse oximeter or pulse oximeter, which determines blood oxygen saturation via the blood’s ability to absorb red and infrared light , through the skin.
If pulse oximeters provide accurate blood oxygen saturation readings sufficient for the diagnosis of hypoxemia, integrating this capability into an unmodified smartphone via a simple app could allow more people to access this marker. health. The smartphone’s camera records the amount of blood that absorbs the light from the flash in each of the 3 color channels it measures: red, green and blue. These intensity measurements are then integrated and analyzed by the deep learning model.
Smartphone can detect up to 70% SaO2
The Washington team brings a first validation of such a detection system via the camera of the smartphone, from contact data only provided by studies which came to feed a specific algorithm.
What mechanism? When we breathe in, our lungs fill with oxygen and this oxygen is distributed to our red blood cells to be transported throughout our body. Our bodies need a lot of oxygen to function, and healthy people always have an oxygen saturation of at least 95%. Conditions like asthma or COVID-19 make it harder for the body to absorb oxygen from the lungs. This leads to oxygen saturation levels that drop to 90% or less. This threshold indicates the need for health care.
The study proof of concept confirms that with such an app, it becomes possible to replace the pulse oximeter with a simple smartphone. The detection is very simple, just place your finger on the camera and the flash of a smartphone, which uses the deep learning algorithm to decipher the oxygen levels in the blood.
- When the team artificially reduced the oxygen level in the blood of 6 participants, the smartphone predicted in 80% of cases, insufficient oxygen in the blood;
- compared to a conventional pulse oximeter, the smartphone app appears just as sensitive.
Jason Hoffman, co-author and research fellow at UW’s Paul G. Allen School of Computer Science & Engineering, comments on these findings: With our test, we are able to collect 15 minutes of data on each subject. Our data shows that smartphones operate accurately in the critical threshold range.”
The argument is the same as for all smartphone health applications: everyone has one and on the same smartphone, it is becoming increasingly possible to bring together all of their health apps. In addition and in an ideal world, this information could be transmitted seamlessly to the doctor’s office.
The team hopes to continue this research by testing the algorithm on a larger number of participants.
Traditional medical devices are subject to rigorous testing. Computer science research is just beginning to tackle the use of machine learning for the development of biomedical devices, an extremely promising line of development that is already finding many applications in diagnostics.