The platform we propose for medical assessment of PPG/ECG signals, is basically composed witha coupled LED-silicon photomultipliers (SiPM) detector. The used silicon photomultiplier SiPM has a total area of 4.0x4.5 mm2 and 4871 square microcells with 60 μm pitch. As light source we’ve used OSRAM LT M673 LEDs in SMD package, emitting at two selected wavelengths. We propose a transmission setup for PPG detector while we usedclassical electrical detectors (at least three as per Einthoven’s triangle) for acquiring ECG signal simultaneously. Clearly, in order to have robust and proper medical measures, the above PPG/ECG coupled signals have to be compliant to the medical standard for that physiological waveforms.
Unfortunately, even though the proposed sensor platform is more efficient with respect to classical ones, it is still affected by such issues corrupting both PPG and ECG signals, such as: electronic noise, body movements, motion artifacts, body tissue issues, breath and heart activity during the measure-session and so on. For this kind of problem, we propose a bio-inspired pipeline for real-time simultaneous adaptive pattern processing of PPG and ECG signal. Preliminary pre-filtering of the PPG/ECG signals is performed by means of IIR Low/High pass filters. A mathematical analysis of the PPG signal is then performed in order to detect some relative extremes i.e. systolic peak(maximum), notch, diastolic peak, minimum of the PPG waveform. We perform adaptive segmentation of the collected pre-filtered PPG timeserie in order to identify each acquired PPG waveform. A Reaction-Diffusion mathematical model is then used to provide a PPG compliant reference signal for robust pattern recognition of the collected pre-filtered signal.
The proposed Reaction-Diffusion model is very innovative as we have supposed to associate the Diastolic phase of the heart to a “Reaction” physical model while we suppose the “Systolic phase” can be mathematically modelled having a “Diffusion” physical proprieties. Several data analysis performed in our laboratory, have pointed out that there are specific cross-correlation between ECG signal and first-derivative of processed PPG waveform, for the same patient.
The compliant first-derivative PPG waveform will be used for analyzing related ECG waveform obtained by automatic segmentation of pre-filtered ECG in the same PPG time onset. Both first-derivative PPG and ECG waveforms are normalized into [0.1]. By means of custom time-rescaling, the proposed pipeline performs ad-hoc sample cross-correlation analysis of that signals. We have noted that compliant ECG waveform shows high cross-correlation value, with respect to corresponding first-derivative PPG waveform. Collected compliant ECG waveforms will also used as reference pattern for subsequent ECG analysis. The proposed pipeline has been tested and validated successfully.