Visualization of Physiological Oscillations with Eulerian Video Magnification

Abstract

Researchers at MIT developed Eulerian Video Magnification (EVM) software that uses temporal and spatial filtering to magnify subtle changes in the environment (Wu et.al.). EVM may provide a unique method for the visualization of low frequency human physiological oscillations. IRB approval was obtained to acquire physiological data from healthy graduate students. Tissue Doppler was used to measure the Traube‐Hering Mayer oscillation which is a low amplitude low frequency fluctuation in blood pressure (~0.1Hz). Respiration was controlled at 15 bpm. Videos were processed using MIT's EVM Matlab code, filtered to 0.067‐1.67 Hz (4‐100 cycles/min), gain set to 50%, using a Gaussian level = 4; and high resolution video from the upper right portion of the forehead was used for spectral analysis. To optimize the quantification of human physiological oscillations; red (R), green (G), and blue (B) spectrums were extracted from color video. These spectrums and their average (C) were auto correlated, normalized, and Fast Fourier Transformed to identify frequencies and their power. Heart rate data obtained from ECG was tightly correlated to the C, R, and G (r2=0.986), and B spectrums (r2=0.949). Low frequency changes in blood pressure were also moderately correlated to C, G and B spectrums (r2=0.75). These preliminary data (N=3) demonstrate that EVM can reveal low frequency human physiological oscillations such as the Traube‐Hering Mayer waveform.