Intestinal perfusion monitoring using photoplethysmography

Abstract

In abdominal trauma patients, monitoring intestinal perfusion and oxygen consumption is essential during the resuscitation period. Photoplethysmography is an optical technique potentially capable of monitoring these changes in real time to provide the medical staff with a timely and quantitative measure of the adequacy of resuscitation. The challenges for using optical techniques in monitoring hemodynamics in intestinal tissue are discussed, and the solutions to these challenges are presented using a combination of Monte Carlo modeling and theoretical analysis of light propagation in tissue. In particular, it is shown that by using visible wavelengths (i.e., 470 and 525 nm), the perfusion signal is enhanced and the background contribution is decreased compared with using traditional near-infrared wavelengths leading to an order of magnitude enhancement in the signal-to-background ratio. It was further shown that, using the visible wavelengths, similar sensitivity to oxygenation changes could be obtained (over 50% compared with that of near-infrared wavelengths). This is mainly due to the increased contrast between tissue and blood in that spectral region and the confinement of the photons to the thickness of the small intestine. Moreover, the modeling results show that the source to detector separation should be limited to roughly 6 mm while using traditional near-infrared light, with a few centimeters source to detector separation leads to poor signal-to-background ratio. Finally, a visible wavelength system is tested in an in vivo porcine study, and the possibility of monitoring intestinal perfusion changes is showed.

Fig. 1

Extinction coefficient of reduced hemoglobin (Hb, in blue) and oxygenated hemoglobin (${\mathrm{HbO}}_2$, in red). Multiple isobestic points can be seen in the visible and NIR ranges. The NIR isobestic point around 800 nm is the most commonly used for near-infrared spectroscopy (NIRS) and photoplethysmography (PPG) sensors (plot based on data from Ref. 18).

Fig. 2

The benchtop PPG sensor (on left) including the laptop, data acquisition card, and sensor interface electronics (from left to right). The right panel shows the visible (upper) and NIR LEDs (lower) used in this study.

Fig. 3

Perfusion signal as a function of wavelength when the intestine is perfused with oxygenated blood. The solid lines show the SBR calculated from the change in the optical properties between perfused and ischemic tissues [Eq. (2)]. The dots are the Monte Carlo (MC) simulation results at the wavelengths of interest. The data show over an order of magnitude enhancement when moving from the NIR to VIS wavelengths.

Fig. 4

Oxygenation signal as calculated by Eq. (1) (solid line) and through MC simulations (red dots). Note the multiple isobestic points in the visible range and the bands of oxygenation-sensitive wavelengths.

Fig. 5

Ratio of the average absorption coefficient of oxy- to deoxy-hemoglobin over the band of commercially available LEDs in the visible and NIR ranges. The data show that the change in the hemoglobin absorption coefficient with oxygenation changes is similar for NIR and VIS wavelengths.

Fig. 6

Mean penetration depth for different source to detector separations as a function of wavelength. This caption shows the corresponding source to detector separation for each line (in case colors are not available, note that the caption is in the same order as the lines appear on the graph).

Fig. 7

Probability density function of the photon penetration depth for the visible and NIR isobestic wavelengths (525 and 805 nm) for multiple source to detector separations.

Fig. 8

Transmittance (squares) and reflectance (dots) as a function of wavelength as calculated by MC simulations. Note that the $y$-data is shown on a log scale.

Fig. 9

The time domain (right column) PPG signal collected in vivo and the corresponding fast Fourier transform (FFT) spectrum (left column) for VIS and NIR wavelengths. In Fig. 14 in the Appendix, the 1 to 5-Hz region is zoomed in showing the cardiac cycle and high-frequency motion peaks.

Fig. 10

The signal-to-background ratio from the in vivo porcine data (black squares) compared with the MC simulations results (red dots).

Fig. 11

In vivo porcine occlusion study data. (a) The change in time of the FFT peak on the two wavelengths of interest. The gray areas correspond to the occlusion periods. (b) The average of the FFT peaks during baseline and occlusion periods.

Fig. 12

Mean penetration depth as a function of source to detector separation for multiple wavelengths of interest.

Fig. 13

Heart rate (a) and modulation ratio (b) measured from the visible wavelengths data. The gray segments indicate the occlusion periods during which the monitored section of the intestine was clamped.

Fig. 14

Zoom in for the 1 to 5-Hz range in the FFT of the AC signal for all six wavelengths. The cardiac cycle peak is four to five times higher for the 470- and 525-nm wavelengths compared with the NIR wavelengths.