Remotely estimating Vital Signs from Ambient-light Photoplethysmography

Abstract

This thesis looks at different aspects of the relatively new and non-invasive technique of remotely estimating vital signs by the use of ambient light photoplethysmography (PPG). Improving signal quality and potentially expanding vital sign detection options can prove a useful tool in several applications.First, an assessment of the signal quality by comparing signals from the visual and theinfrared parts of the light spectrum will be done. Possibilities to estimate differential blood pressure by looking at changes in phase difference between a signal from one extremity (forehead) to another extremity (hand) is investigated. Finally, an experiment is done to find information about microcirculations from a colonoscopy and laparoscopy video by analyzing frequency spectrum and performing a pulsality mapping and a power mapping.From recordings of eight subjects by using webcam and thermal video, the webcam provessuperior in PPG signal quality. This is likely due the low spatial and temperature resolution on the thermal camera. Results from two videos recorded with prototype camera using a 4-channel CMOS sensor indicates this sensor type as being able to yield good results with correct parameters for the recording. In terms for signal quality, the green channel of the RGB color space outperforms the HSV color space. Using independent component analysis (ICA) also generally gives better results, particularly on the more challenging videos.The phase difference between forehead and hand PPG signal (related to pulse transit time, this is called DPTT) is not possible with the thermal video used, due to poor signal quality and noise. With the webcam video, the estimations are possible, but the accuracy is limited and questionable.The results from the experiment to estimate change in blood pressure (BP) based on achange in DPTT from resting condition to a condition of elevated pulse after short interval workout cannot be deemed reliable. However, with more accurate DPTT measurements through improved compensation of movement and light fluctuations, and possibly higher frame rate, this is a viable next step in remote PPG vital sign detection.The trial with the laparoscopy and colonoscopy videos to see whether any informationabout the microcirculation in the tissues can be revealed, yields no conclusive results. The spectra show more than one possible PPG peak, so better algorithms to stabilize movement and compensate for light fluctuations are required for reliable results. The theory is still not ruled out, and is encouraged for further works.