2.1.1.

Motility standard

The motility standard was an aqueous solution containing two components selected to simulate the optical properties of scattering and absorbance that characterize the interaction of blood with light in the near-infrared (NIR) range. The motility standard was comprised of 0.2% 100-nm polystyrene beads (Sigma-Aldrich, St. Louis, Missouri) and 0.1% India ink (Higgins Inks, Leeds, Massachusetts) manually mixed in deionized water.¹⁴ India ink was used to simulate NIR absorbance because it interacts with the incident NIR light in a manner similar to hemoglobin.

2.1.2.

Bare-vessel phantoms

Bare-vessel phantoms were generated by positioning four segments of 1.58-mm inner-diameter silastic tubing (Dow Corning, Midland, Michigan) such that they lay parallel at distances of 1, 5, and 10 mm apart. Silastic tubing was chosen because it was readily available, optically clear, and demonstrated elasticity similar to that of human arterial vessels, with a Young modulus of 2.1 MPa at 200% elongation.¹⁵ The motility standard was placed inside the tubing and pumped with a two-roller peristaltic pump (Watson Marlow, Wilmington, Massachusetts; Model# sciQ32) at varying frequencies to generate pulsatile fluid flow. The pump was capable of producing pulse frequencies over the physiologic range of the human heartbeat [40 to 200 beats per minute (bpm)].^16,17 The pressure of the fluid was measured directly with the calibrated digital pressure transducer (calibrated against a Dwyer Series 490 digital manometer) and recorded using LabVIEW software (Signal Express, version 2.5).

In order to determine the pump pressure that provided a 2% luminal volume change consistent with physiologic tissue-volume changes during the cardiac cycle,^15,18,19 a section of the silastic tubing was filled with a 2% India ink solution and examined under a microscope to measure the diameter change that occurred with various pump pressures. The change in the cross sectional area of the inner lumen of the tube was linearly correlated with pump pressure (Table 1; $R^2=0.9826$). This result demonstrated that under the range of pressures tested, the silastic tubing remained in the elastic region of its stress–strain relationship. A pressure change of 150 mmHg provided the diameter change that resulted in the desired 2% lumen volume change, which is essential for a PPG imaging system.

Table 1

Stress–strain relationship for the pressure settings used in the bench test.

2.1.3.

Tissue phantoms

Tissue phantoms were comprised of a gelatin–lipid solution molded around the silastic tubing. The tissue phantoms were designed to model the characteristic absorbance and scattering observed during light’s interaction with skin. In skin tissue, light absorption is primarily dependent upon hemoglobin and is therefore proportional to the volume fraction of whole blood. In humans, the average volume fraction of whole blood in the skin is $\sim 0.2\%$.²⁰ By incorporating 0.2% volume of our motility standard (i.e., phantom blood substitute) into the gelatin mixture, we simulated the volume fraction of whole blood in the tissue phantom. Dermal scattering simulation was achieved using a lipid emulsion solution that is biologically similar to the bilipid membrane of cells and organelles, which are the primary contributors to light scattering in skin tissue. We chose to neglect the effects of melanin in these experiments because melanin has a low overall concentration in the skin that varies widely based on skin tone.

Three different tissue phantoms were constructed according to the methods outlined by De Grand et al.⁸ Briefly, 10% w/v type B gelatin (JT Baker, Center Valley, Pennsylvania) in Tris-buffered saline (pH 7.4; Alfa Aesar, Haverhill, Massachusetts) was mixed with a sterile 20% w/v intralipid fat emulsion (Baxter International, Deerfield, Illinois). Tissue phantoms with intralipid concentrations of 1%, 4%, and 6% were mixed. A concentration of 6% resulted in a tissue phantom that was opaque to the human eye. To the final mixture, the motility standard (0.2% of total volume) was added to approximate the volume fraction of whole blood in humans. The mixture was poured to a depth of 8 mm in a Petri dish with a single vessel submerged in the phantom. Once the tissue phantom had set, the phantom was connected to the pulsatile flow pump. The regions of tubing not immersed in the tissue phantom were covered with opaque cloth so that PPG measurements were collected only from tubing fully encased within the tissue phantom.

2.1.4.

Optical characterization

To ensure the acquisition of any signal by the noncontact PPG imager was indeed true signal, PPG images were acquired of two 4% intralipid tissue phantoms under two different cases. In the first case, the motility standard was pumped through the tissue phantom at 60 bpm with 150-mmHg pressure. For the second case, 4% intralipid tissue phantom fluid was prepared and pumped through the tissue phantom at 60 bpm and 150 mmHg. PPG outputs were collected at a distance of 15 cm and then assessed. The outputs were used to determine if any resultant signal was due to the presence of motility standard flowing through the vessel and the accompanying volume change, or specular reflection and refractive index mismatch of the tissue phantom and associated tubing.

The optical properties of the tissue phantoms were also evaluated. Vis–NIR spectrophotometry was carried out on 1%, 4%, and 6% intralipid concentrations of the tissue phantom and the motility standard to assess how the attenuation coefficients compared. Given the limitations of the spectrophotometer, the observed attenuation coefficient is a combination of the scattering and absorptive properties of the media. Appropriate concentrations of the intralipid solutions were prepared and placed inside $1.0\mathrm{cm}\times 1.0\mathrm{cm}$ plastic cuvettes (BrandTech Scientific Inc, Essex, Connecticut). The optical data were collected using a lamp, fiber, and spectrometer and processed using OceanView software (Ocean Optics, Dunedin, Florida). This data aimed to provide additional information on the optical qualities of the tissue phantom.

2.2.1.

Experimental photoplethysmography sensor

The PPG sensor in this investigation was the DeepView DVS 1000p (Spectral MD, Inc.). This system was comprised of a 12-bit $1025\times 512\text{pixel}$ CMOS monochromatic imager (Photonis Lynx CMOS). The sensor contained square pixels of $5.6\mu \mathrm{m}$ length. The camera was connected via cable to a frame-grabber in a computer using a camera link in base mode. The camera was equipped with a $1/3$ in. M12-mount lens with focal length 2.8 mm ($f2.0$). The camera collected frames at 30 frames per second using a rolling shutter.

The illumination source consisted of four 850-nm infrared light-emitting diodes (LEDs). Red and NIR light have historically been used for PPG systems since they maximize penetration for optical imaging techniques.²¹ The LEDs were encased in glass lenses that produced a beam angle of 18 deg, and no cross-polarization was used. Each LED was driven by a DC current source at 150 mA, 100% duty cycle. The LEDs were mounted to the camera using a custom ballast. The ballast positioned the LEDs in a plane $\sim 6\mathrm{cm}$ in front of the camera lens. In this plane, the LEDs were positioned perpendicular to the imaging target with their center of irradiance in each corner of the camera’s field of view.

For the bench setup, the camera and illumination system were mounted on a rail system orthogonal to the imaging target on an optical bench. The working range for this instrument was 10 to 20 cm from the imaging target’s surface. The data capture time was 27 s (800 total frames). A diagram of the experimental setup is shown in Fig. 1.

Fig. 1

(a) Tissue phantom in a Petri dish with an elastic tube embedded in the homogeneous tissue matrix, which mimics the blood flow under the skin. (b) The phantom apparatus is designed to simulate human pulsatile blood flow in a laboratory setting. The peristaltic pump drives the motility fluid through an elastic phantom vessel, 8.0 mm below the gelatin–intralipid tissue-like phantom matrix. Due to the elasticity of the tubing, an $\sim 2\%$ volume expansion in the phantom vessel, similar to that of human arteries, occurs with each cycle of the peristaltic pump.

2.3.1.

Pulse pressure detection

Using the bare-vessel phantom, pulse pressure was varied from 0 to 1000 mmHg while pumping at a constant frequency of 60 bpm. We measured the response of the systems to the changing pulse pressures by measuring their power spectral density (PSD). The maximum value of the PSD at the frequency of the pump was determined for each pressure amplitude setting. The PSD was then calculated by taking the power spectrum of the time varying signal. The zero-pulse pressure (i.e., “no-flow” condition) was used as a negative control.

2.3.2.

Pulse frequency detection

The ability of the noncontact PPG imager to detect pulsatile flow was assessed using both the bare-vessel and tissue phantoms. In the bare-vessel phantom, the frequency of the PPG waveform was varied by changing the frequency of the pump. The frequency was set to three values encompassing a physiologic range from bradycardia to tachycardia (40, 120, and 200 bpm), and the ability of each system to measure pulse frequency was then tested. To generate pulse frequency data from the noncontact PPG imager, a Fourier transform was applied to the time-domain waveform of the device. The result from the noncontact PPG imaging system was compared with that of the algorithm derived pulse oximeter.

The ability of the noncontact PPG imager to detect the pulse frequency in phantoms of varying intralipid concentrations (1%, 4%, and 6%) was also evaluated. Intralipid 1% solution has been described to accurately mimic skin tissue optical properties, including scattering.⁷ However, the 1% concentration matrix did not visually conceal the embedded silastic tubing when filled with the motility standard. Therefore, we increased the intralipid concentration to 4% and 6% in order to better obscure the tube and motility standard during testing. The noncontact PPG imager was placed 15 cm from the target tissue phantom to detect the fluid flow signal in both single point mode and image mode. Fluid was pumped such that during each pump cycle a 2% volume change occurred in the embedded vessel phantom. Two-dimensional colormap images of the tissue phantom setup were acquired with the noncontact PPG imager: one in a no-flow condition to serve as a baseline control and another with the pump oscillating at a frequency of 60 bpm.

2.3.3.

Spatial resolution

We assessed the spatial resolution of the noncontact PPG imager using the bare-vessel phantom set to a pulse pressure of 150 mmHg and frequency of 60 bpm. The distance from the imager to the bare-vessel phantom was incrementally increased from 10 to 20 cm while the signal strength and location were evaluated. Colormap outputs from the device were used to assess the ability of the imaging system to resolve the individual vessels.