A method to detect cardiac arrhythmias with a webcam

We describe a novel method to detect the presence of atrial fibrillation (A-Fib) by video recording of the face of a patient. The signal processing algorithm extracts A-Fib episodes by processing time-series signals generated by video images in a remote sensing environment. The process removes the non-stationary components of the green component of the RGB signal. Beat-to-beat pulse points are detected using an adaptive threshold technique; the successive thresholds are based on variations detected in previous magnitudes of the pulse peaks. The existence of cardiac arrhythmia is determined by quantifying the variability of the peak-to-peak intervals. We report preliminary results of our on-going study on six subjects undergoing electro-cardioversion for persistent A-Fib.