2.1.

Pressure Sensing

To enable pressure sensing while minimizing the effects of cable tension, there is a section of bare fiber between the furcation and ferule tubes. This bare fiber section is coiled inside of the probe handle, see Fig. 1. The furcation tube is rigidly attached to the handle and the ferule can move independently from it. The fiber coil shrinks or expands to take up the difference between the ferule and furcation tube separation distance over a nominal 1-cm range of motion. The fibers were selected based on their ability to wind around the relatively tight fiber coil (4-cm diameter).

Fig. 1

An instrument schematic shows the system components. (a) The features include an optical switch (dual shutter) to direct light to either a calibration or tissue-sensing channel for automated calibration measurement. (b) A pressure sensing system was also integrated into the probe handle using a spring-loaded metal ferrule that contacted a force sensor. (c) This was uncoupled from the weight of the cable through use of a coiled segment of bare cable that could flex as the fiber tip was compressed against the sample.

2.2.

Self-Calibration

A system block diagram is shown in Fig. 1(a). The primary innovation is the incorporation of two optical paths: a calibration leg and a measurement leg to facilitate self-calibration with an automatic shutter. The calibration leg is comprised of a short (internally contained) ferule that is terminated with a plastic cap containing a spacer and a Spectralon™ calibration disk. In this way, calibration measurements can be obtained immediately after an experimental measurement with the probe in the same position and nearly the same time, which minimizes the effects of (1) lamp warm up or power drift and (2) effects of fiber bending. Past designs for self-calibration have used two spectrometers: one spectrometer for tissue measurements and another one for calibration.^1,2 However, for both cost and performance reasons this is not the ideal configuration; it is unlikely that any two spectrometers will operate the same, throwing the SC measurements into doubt. Our updated design (Fig. 1) addresses this challenge using only one spectrometer with a $1\times 2$ optical switch to rapidly switch back and forth between tissue measurements and calibration measurements.

2.3.

Analysis

Data were background subtracted by taking a measurement with the lamp shutter closed. Data were also normalized to the calibration standard measurement to obtain the diffuse reflectance relative to the spectrally flat standard. A scalable inverse Monte Carlo-based algorithm was used to extract the underlying tissue optical properties.¹¹ The model utilized measured data from 500 to 600 nm, in 2.5-nm increments to obtain a least squares fit. The model assumed absorption could be characterized by a combination of oxy and deoxy hemoglobin, modified by the pigment packing factor of van Veen et al.,¹² and a nonmelanin skin absorption component.¹³ Scattering is approximated by the relationship, ${\mu }_s^{\prime }=A\times {\lambda }^{-b}$, where ${\mu }_s^{\prime }$ is the reduced scattering coefficient, $A$ is the scattering amplitude, $\lambda$ is the wavelength in nanometers, and $b$ is the scatter power factor, respectively. The model inputs including absorber concentrations and scattering parameters were adjusted to obtain a least squares fit to the measured data using the Python SciPy library.¹⁴

A linear mixed effects model was used to test for significant main effects for pressure (modeled as a factor), time (represented by repeated sample number and modeled as a continuous variable), and site (modeled as a factor for the clinical measurements). Both hemoglobin oxygen saturation and total hemoglobin content were used as response variables to test for effects of these parameters on each. The nlme library in $R$ was used for the analysis using the lme function.¹⁵ The subject was modeled as a random effect, with pressure, time, and site input as fixed effects.

2.4.

Patient-Derived Xenograft Tumor Study

Animal studies were performed in accordance with a protocol approved by the Duke Institutional Animal Care and Use Committee (IACUC). About 6-week-old female athymic nude mice were obtained from Duke breeding core facility. Human glioma patient-derived xenograft (PDX) tumors (kindly provided by Dr. Stephen Keir, Duke University) were implanted subcutaneously in the flank of the nude mice. Six nude mice were transplanted with glioma PDX tumors in the flank. When tumors were $\sim 300{\mathrm{mm}}^3$, the Zenascope was used to acquire spectra at a series of different pressures as assessed by the force sensor to produce calibrated pressure measurements (24, 48, 72, 120, 192 mmHg). This was done by the technician, who placed the probe and attempted to match the pressure as closely as possible to the target value prior to collecting the measurement ($\pm 20\%$). The median exposure time was 1570 ms. The PDX model was chosen as this has been demonstrated to more faithfully replicate the structural and functional characteristics of human tumors, as compared to cell line xenograft models.¹⁶ After each measurement, the probe was removed from the tissue, and replaced for five repeated measurements at each pressure level. In addition, to replicate the typical sensing method, the spring-mounted probe was fixed in place, and the probe was placed similarly, which the technician (who was not experienced in optical spectroscopy techniques and blinded to the study intent) was instructed to place the probe using minimal pressure while ensuring complete contact with the tissue. This is referred to as the “fixed” probe measurement.

Figure 2 shows box and scatter plots of the full dataset. It can be seen that the hemoglobin oxygen saturation (HbSat) is significantly affected by pressure, with higher pressures resulting in lower HbSat levels ($p=0.02$ by linear mixed effects model). Other parameters (total hemoglobin concentration and mean reduced scattering coefficient) did not show a significant difference. The fixed probe measurement appears similar to the higher pressure level for all parameters.

Fig. 2

The optical outputs as a function of the nominal applied pressure. Each point corresponds to a single measurement ($n=6$ animals with five repeated measurements each). It can be seen that the (a) Hb saturation exhibits a systematic decrease with increasing applied pressure. (b) The total Hb and (c) mean reduced scattering coefficient can be seen to be relatively more consistent across pressure. The case of a fixed probe is also shown, which appears similar to the higher pressure levels.

Figure 3 shows a similar plot looking at the intra-animal change in these parameters, focusing on those related to absorption as scattering is not significantly affected. The mean endpoint value obtained at lowest pressure measurement (24 mmHg) was taken as the baseline for each animal, and was subtracted from the averaged endpoint obtained for each subsequent pressure level. It can be seen that in general there is a suppression in all of the parameters as a function of pressure, but that HbSat and oxygenated Hb concentration change more consistently in a negative direction (all subjects appear at or below zero for the highest pressure level as well as the fixed probe case). This may explain why HbSat showed a significant effect of pressure, whereas total hemoglobin concentration did not (Table 1).

Fig. 3

The relative in change for each optical endpoint is shown for each animal. The average of five repeated measurements was taken for each animal for each pressure level, and the lowest pressure level measurement was subtracted from this to obtain the change in signal. The most consistent changes are seen for (a) Hb saturation and (c) oxy Hb concentration, both shows a median decrease with increasing pressure. (b) The total hemoglobin, and (d) deoxy hemoglobin also show decreasing trends with increasing pressure.

Table 1

Statistical significance for repetition number and pressure level using a linear mixed effects model for both hemoglobin saturation and total hemoglobin content for PDX tumor model.

2.5.

Volunteer Study

To study the influence of pressure in a realistic clinical setting, a healthy volunteer study was initiated at Duke University Medical Center, which was approved by the Duke University Institutional Review Board. To minimize the time required for measurement and simulate a more realistic clinical measurement mode, the software was updated to enable automated pressure sensing and repeated measurements using a single probe placement. This was achieved through continuous sampling of the force sensor upon activation of the probe along with continuous sampling of the spectrometer for the optical measurement. For each spectra obtained, it was checked if the force sensor (calibrated to units of pressure) remained within the specified valid range during the entire measurement, as well as the sensor readings immediately before and after the spectra were obtained. Three pressure levels were tested. This process was repeated 10 times per measurement site and pressure setting. Four sites were measured in each subject, cheek, tongue, sublingual, and gums. Each measurement set (site and pressure combination) took a median time of three seconds to acquire 10 replicates plus one calibration measurement. The median exposure time per acquisition was 182 ms.

Figure 4 shows boxplots of each endpoint, broken down by site and pressure levels. It can be seen that for any given site, there is a consistent decrease in the median hemoglobin saturation as a function of applied pressure. In contrast, the total hemoglobin concentration decreases in some cases but not for all. The linear mixed effects model indicates however, that for both Hb saturation, as well as Total Hb, the site, and applied pressure are independently significant factors, whereas measurement repeat number is also significant for Hb saturation (Table 2). When broken down into oxy and deoxy Hb concentrations, it can be seen that the oxy Hb shows a consistent decrease, whereas deoxy Hb does not, and in some cases even increases with increasing pressure. Thus the consistent change in Hb Saturation appears to be driven by a consistent drop in the concentration of oxy Hb. Finally, mean reduced scattering does not show any observable trend with pressure applied [Fig. 4(e)].

Fig. 4

Optical parameters from a set of nine healthy volunteers. A series of 10 measurements were acquired sequentially from each of the four sites from each subject, for each of the three target pressure ranges. Deviations as a function of pressure can be seen in particular for (a) Hb saturation, and (c) oxy Hb concentration, with other parameters, namely (b) total Hb, (d) deoxy Hb, and (e) mean reduced scattering coefficient, showing more variable responses.

Table 2

Statistical significance for the monitored parameters using a linear mixed effects model for both hemoglobin saturation and total hemoglobin content for the volunteer study.

In further examining the intra-subject change in these parameters as a function of pressure, the lowest pressure measurements were averaged and subtracted from the averaged signal from each of the other pressure levels for each subject (Fig. 5). In this case, it can be seen that there are consistent drops in the median hemoglobin saturation, total Hb, and oxy Hb at the higher pressure levels, which appear, in general, to increase in magnitude with increasing pressure.

Fig. 5

Relative change in optical parameters relative to the lowest pressure is shown with respect to tissue site and pressure level. It can be seen that the largest and most consistent deviations appear for (a) Hb saturation and (c) oxy-Hb concentration, with more consistent values seen for (b) total hemoglobin, and (d) deoxy Hb.

Finally, because a series of 10 repeated measurements were acquired from each site with the probe held in contact, it is possible to evaluate the dynamic changes in each endpoint as a function of time after the probe achieved the desired pressure level. The Hb saturation is shown as a function of measurement number (Fig. 6) for each of the four sites [(a) cheek, (b) gum, (c) tongue, and (d) under tongue]. The local mean and 95% confidence intervals (CI) are plotted using the geom_smooth function from the ggplot2 package in R (indicating significant deviations from zero where the CI does not overlap the abscissa).¹⁵ Notably, the hemoglobin saturation shows a significant decrease over time for the highest pressure range for all tissue sites ($p<0.05$ for nonoverlapping CI). There is some overlap in the CI among the different pressure levels, but the highest and lowest pressure levels are significantly separated for all sites except for the cheek at the later measurement times, with the higher pressure resulting in a larger reduction in hemoglobin saturation relative to the starting point. In contrast, total hemoglobin content does not appear to show consistent significant trends (Fig. 7).

Fig. 6

For volunteer subject measurements, the Hb saturation measurement as a function of repeated measurement number is shown for each of the nine subjects at each of the three pressure levels. The moving average mean (loess) is also shown along with the 95% confidence intervals. This is shown for (a) cheek, (b) gum, (c) tongue, and (d) under the tongue. It can be seen that the mean Hb saturation significantly drops for the majority of the site-pressure combinations with increasing reduction over time. In addition, the higher pressure levels in general can be seen to have a larger drop over time.

Fig. 7

For volunteer subject measurements, the total Hb measurement as a function of repeated measurement number is shown for each of the nine subjects at each of the three pressure levels. The moving average mean (loess) is also shown along with the 95% confidence intervals. This is shown for (a) cheek, (b) gum, (c) tongue, and (d) under the tongue. It can be seen that the total Hb largely does not show consistent or significant changes as a function of time for the tested conditions.