Wearable Near-Infrared Organic Photodiode Arrays for Cardiovascular Health Monitoring and Biomarker Mapping: Signal-to-Noise Ratio Enhancement Using Machine Learning

Photoplethysmography (PPG) is a non-invasive and unobtrusive optical technique that is used to determine and monitor cardiovascular parameters. This technique utilizes light emitting diodes (LEDs) and photodiodes (PDs) placed over the skin to measure blood volumes changes in the subcutaneous arteries. The quality of the PPG signal can be very low due to the weak reflected signal and the low signal to noise ratio (SNR), in particular at larger LED-PD distance. To overcome this, as well as to record larger areas, we have developed ultra-thin flexible near-infrared (NIR) sensitive organic photodiode arrays, integrated with flexible illumination into a wearable. The photodiode arrays consist of 256 organic photodiodes in a
$16 \mathrm{x}16$
matrix, with an area of 4
$\text{cm}^{2}$
. At different body locations, high quality reflective-PPG signals can be measured continuously at a sampling rate of ~ 6 kHz. Using the data recorded from the set PDs, the SNR was improved by applying a machine learning technique, independent component analysis (ICA), to separate the “true” PPG signal from noise and artifacts. The pre-processed PPG signals can then be used to accurately calculate and map certain biomarkers such as peripheral oxygen saturation (SpO2). The high sampling rate in combination with spatial mapping and machine learning will pave the way for future local pulse wave velocity (PWV) measurements.