We wanted to test the ability to compute vital signs from video on various platforms and in various environments to ensure robust adoption across a plethora of applications. Drone technology has seen significant advancements in the last two decades, enabling high quality imaging to be accessible as off-the-shelf consumer devices.
At Presage Technologies, we believe this has widespread utility in assisting any emergency medical need in remote or dangerous scenarios, including but not limited to battlefield medicine, remote triaging in inaccessible terrain or during a natural disaster, or victim assessment during an active shooter situation.
Testing
To explore the robustness of Vitals by Video to camera motion-induced with outdoor drone imaging, we tested our software with a standard, affordable device. The DJI Mavic Air 2S supports several frame rates (30, 60fps) and resolutions (5.4k, 4k, 2.7k). We acquired outdoor drone footage, hovering approximately 5ft. above a stationary subject adorned with a pulse oximeter, and transferred and processed the video on a standard mac laptop.
Our Vitals by Video calculated heart rate was within 3bpm of the pulse oximeter used to measure the same value throughout the duration of the video. The respiration rate fell within 2 bpm of the manually counted value. We repeated this acquisition at approximately 15ft. above the subject and found the same bounds on accuracy. Both videos were acquired at 4k resolution and 30fps.
Implementation
Currently, in stable scenarios where we have control of the surrounding environment, our Vitals by Video is as accurate as a wearable medical device such as a pulse oximeter.
Additionally, we are making great strides in handling challenging scenarios like high winds or moving shadows. Strong winds cause large degrees of motion in the drone, and subsequent blur and low-quality signal in the video. We anticipate acquiring video at higher frame rates and lower exposure time will resolve this challenge. Another case involves moving shadows from obscuring objects such as trees or the drone itself. We plan to address this with algorithm-based solutions that are robust to shadows, as well as with custom hardware, such as near-infrared (NIR) wavelength illumination and collection, in which the solar spectrum is minimal.
Conclusion
We envision incorporating our Vitals by Video software into an embedded device that acquires and processes on the drone. In doing so, we can enable real time auto adjustment of acquisition parameters, improving accuracy, and reducing the required acquisition time.
Fully embedded systems have the advantage of custom hardware, including NIR wavelength illumination with corresponding sensors to handle night-time imaging and inconsistent or inadequate solar illumination. With the right partnership, we envision our technology will make a major contribution to the advancements of remote emergency medicine.
For more information, contact Presage Technologies. We would love to hear about your intended use case.