About

Who We Are

CoVital is one of the highlighted projects of Helpful Engineering, our parent organization.

We are an international group of volunteers from all backgrounds working over a slack channel of 200+ people to bring this project together.

Background

COVID-19’s major symptom is a lung infection (pneumonia) that results in lower blood oxygenation. However, patients may not physically feel short of breath, so it is crucial to closely monitor their blood oxygenation levels. This is traditionally measured with medical-grade pulse oximeters, but not everyone has access to these devices, especially those who live in remote regions.

Core Objectives

  • Provide patients with a way to monitor their own blood oxygenation. Currently, people who have or suspect they have COVID-19 do not have a reliable way to track their blood oxygenation outside a medical facility, which makes self-isolation difficult and ultimately more risky.

  • Help discharged patients monitor their vitals so that they can act fast and call a medical emergency before they’re in critical danger.

What Currently Exists

With the advent of smartphones, there have been a handful of mobile applications created to measure blood oxygenation using the phone’s camera.

However, it is clear that this technology is still underdeveloped. As of now, there is no reliable evidence to show that the apps currently on the market are able to accurately and consistently measure blood oxygen at a level comparable to medical-grade devices.

Our Solution

This is where we can make a difference. CoVital collects fingertip video data using different smartphones from patients with a wide range of blood oxygen saturation levels. We evaluate the current methodologies, and then build new artificial intelligence-based models to improve on the current results and develop an algorithm that works for any patient with a smartphone.


Steps

  1. Build a large open-source dataset of fingertip videos from a variety of sources (different patients, smartphones, countries, etc.).

  2. Use this dataset to evaluate the latest methods on SpO2 estimation, and see if they are sufficient (2% error margin). If current methods are not adequate, we aim to further develop and refine these methods to improve accuracy and precision.

  3. Design and develop an open-source application that provides a health diary and stores data about blood oxygenation, for anyone across the globe to monitor their vitals. This would also provide medical teams with crucial patient data.


Relevant Links

  • There are existing APKs that align with our objectives but do not currently achieve adequate results. Find a working version of healthwatcher here.

  • For a detailed introduction to the science behind why measuring blood oxygen saturation with a phone camera works, and which machine learning methods we are looking to investigate, learn more here.