Evid Health- leveraging ready-at-hand technology to improve access to health

Chronic obstructive pulmonary disease (COPD) remains a leading cause of morbidity and mortality.

Exacerbations result in an acute drop in oxygen saturation (SpO2) levels and early detection is pivotal to ensure timely intervention which, in turn, lowers the risk of hospitalisation and improves outcomes (Blondeel, 2021; Zhang et al., 2023; Patel et al., 2003).

We are building an AI-based “smartphone as pulse oximeter” solution that allows users with COPD to obtain their SpO2 by placing their finger over the rear camera of their smartphone.

We aim to integrate this technology into the primary care remote consulting workflow developed by our partners Fibricheck. We have developed a CNN-based AI model for estimating SpO2 from the "finger-tip videos".

When testing our AI model on a publicly available dataset we achieved a mean average error (MAE) of 1.86% (ISO standard requirement MAE<2%) to the ground truth label (pulse oximetry) for saturations 70-100% on 6 participants. This includes an MAE of 1.76% in a subject with Dark skin. Our model is more accurate than the model used by the authors (MAE= 5%) of this dataset (Hoffman et al., 2022).

Using support from the NIHR Fast Grant, with our partners at the High-Altitude Centre (London), we aim to perform a data collection experiment. We have developed a data collection app for this experiment already.

After this experiment, we aim to have sufficient pre-clinical evidence base to pursue ISO standard professional testing at the Birmingham MD-TEC lab. With the Asthma + Lung UK Charity, we will undertake a support group study with their Respiratory Insights Network.

This will involve 10 x COPD persons with lived disorder (PWLDs) including persons with limited technological literacy and older age groups. This study will help us identify areas of bias in our UX and incorporate best practices in digital inclusion.