1.

Introduction

Clinical research trials using near-IR spectroscopy (NIRS) to measure cerebral oxygenation status, blood flow, and blood volume at the bedside have been ongoing^{1 2 3 4} since the 1980s. The spectrometers emit and receive photons at multiple wavelengths via fiber optic bundles approximately 3 m in length containing 1000 or more fibers. Often these bundles become fractured, either partially or completely, due to the frequent handling they undergo at the bedside and during transport between studies. An early report of flexible fiber optic bronchoscopy cites 1 of usages failing from malfunction of the bending apparatus and broken fibers.⁵ In clinical NIRS trials, it is possible that the fractures can go undetected and subsequently compromise data interpretation.⁶ Unfortunately, an algorithm for warning of excessive fracture levels has not yet been devised because of the paradox of distinguishing between impaired light transmission and biologically weak signals.⁷

Economically priced liquid light guides using quartz windows to seal in the nontoxic transmission medium are now available with terminal fittings suitable for clinical NIRS devices. Besides the obvious advantage in overcoming breakage, liquid light guides do not suffer the packing fraction transmission losses arising from the combined dead spaces between optical fibers.

We investigated whether a liquid light guide would be a suitable receiver for clinical NIRS data collections.

2.

Method

A Hamamatsu NIRO-500 (Hamamatsu Photonics KK, Hamamatsu, Japan) spectrophotometer was used to emit and receive photons, measure background darkness, and compute optical densities for wavelengths of 777, 828, 849, and 910 nm. The spectrometer uses four sequentially pulsed laser diodes with each diode having a pulse frequency of 1.9 kHz at a duration of 100 ns and a mean power output of 2 mW.

In all trials, light was transmitted to the test object via the same standard NIRO-500 Type C flexible fiber optic bundle having a 90-deg prism housed in a 10-mm-diam terminal. Both the transmitting and receiving cables were laid out linearly to minimize transmission losses caused by bending of the light guide.

The receiving optode cable was either a single terminal fiber optic bundle, a dual-branch fiber optic bundle, a linear array panel fiber optic bundle, or a liquid light guide. The single-terminal fiber optic was a standard NIRO-500 Type C (Hamamatsu Photonics, London) flexible bundle containing 1000 fibers in a 3-m length with a 90-deg prism housed in a 10-mm-diam terminal. The dual-branch fiber optic was a K54-203 (Edmund Scientific Co., Barrington, New Jersey), 2 m long with 50-μm fibers in a 6.4-mm-diam sheath having an 82 packing fraction. The linear array panel fiber optic was a K54-228 (Edmund Scientific Co., Barrington, New Jersey), 1.25 m long with 50-μm fibers in a 10-mm-diam sheath with an 83 packing fraction and having a 47×41-mm reflection/capture chamber abutting the fanned fiber terminus. The liquid light guide was a 77635 visual–near IR (VIS-NIR) (Oriel Instruments, Stratford, Connecticut), 2 m in length and 3 mm in diameter with a 0.52 numerical aperture and 72-deg cone of acceptance.

The NIRO-500 terminal interfaces and the dual-branch’s “Y” junction were wrapped with conventional self-melding, rubber, electrical insulation tape to ensure extraneous light occlusion.

There were three test objects: a stable liquid phantom, an unstable suspension phantom, and a forearm. Each phantom was contained in a 90×90×70 mm flat-sided, clear, transparent flask. The stable phantom was 0.5 g of fluoresceine dye in 565 ml of water tinted with 1 ml of India ink. The unstable phantom was a suspension of 1.25 g copper oxychloride in 565 ml of water. The forearm was the widest portion of the left forearm of a healthy human adult male. (Approval by the University of British Columbia Human Ethics Review Committee was obtained.) The phantom concentrations were chosen to achieve a maximum photomultiplier tube (PMT)-applied voltage and a typical midrange PMT-applied voltage.

Throughout the trials, the emitting optode remained fixed at the same central site on the test object. The receiving optode was affixed directly opposite the emitting optode on the phantom (transverse mode), and adjacent the emitting optode on the forearm (reflection mode). The phantoms had a 90-mm interoptode space, while the forearm had a 30-mm interoptode space. A path length scattering factor of 1.0 was used for the phantoms and 4.16 was used for the forearm.⁸ Once the optodes were attached, the test object was completely wrapped in a light-occluding blanket, folded and sealed on all edges. The unstable phantom was stirred rapidly for 3 s prior to wrapping and allowed to settle thereafter during each trial.

The dark signal level (INC DARK) of the photodiode monitoring the output of the laser diodes and the dark signal level between laser firings (PC DARK) were recorded whenever the data collection was zeroed to establish the optical density relative starting values. At the same time, the PMT operating temperature and applied high voltage were also recorded.

The single terminal, NIRO-500 Type C, fiber optic receiver was deemed to be the comparison standard (control) and was retested between trials of the other receivers. Each trial consisted of approximately 3 min of data collection at 1-s intervals to record optical density values for each wavelength. There were no interventions. Each test object had four standard receiver trials and one trial each for the dual-branched, linear array, and liquid receivers. Thus, there were 21 trials for the entire study.

Each trial’s signal noise was analyzed by deriving a least-squares linear best-fit line for each wavelength’s optical density data and then calculating the squared error between each datum point and its best-fit equivalent. The square root of the mean squared errors (RMSE) was taken as the measure of signal noise.

The mean and standard deviations of the RMSEs at each wavelength from the four trials of the NIRO-500’s Type C receiving cable on each test object were derived as the comparison standards for each test object. Thus, there were 12 comparison standards. At each wavelength for each test object, the difference between the dual-branch, linear array, or liquid guide RMSE and the NIRO-500 Type C’s mean RMSE was divided by the NIRO-500 Type C’s standard deviation of RMSE to determine whether they were more than 2 standard deviations of the comparison standard and thus were significantly different. The dual-branch, linear array or liquid guide RMSE was then expressed as a percentage of the NIRO-500 Type C’s mean RMSE to indicate the magnitude of difference.

3.

Results

The mean PMT temperature was 28.2 °C ±2 for all trials. It did not change during a trial and was within ±0.15 °C of the starting value between trials on the same phantom.

The average dark signal during laser firing (INC DARK) was −454 ±3 for all trials. It did not change during a trial and was within 4 counts (0.9) between trials on the same phantom.

The dark signal when the lasers were not firing (PC DARK) and the PMT applied high voltage both changed between optode types and between test objects. A further description of these results is given in Figures 1 and 2.

Figure 1

Our 849-nm optical density data collections with shifted baselines for ease of comparison between light guides type during trials of the stable phantom (A), unstable phantom (B), and human forearm (C).

Figure 2

Comparison of the light occlusion between laser firings (PC Dark Count) for all trials, where “Stable” indicates the stable suspension particle liquid phantom setup, “Unstable” indicates the unstable precipitate liquid phantom setup, and “Forearm” indicates the human forearm setup.

The RMSE value, standard deviation comparison, and mean value percentage difference for each trial are given in Table 1.

Table 1

4.

Discussion

Efficient light transmission depends in part on the sample/sensor interface, interface stability, minimal bending of the light guide, and extraneous light occlusion. The trials were conducted with the optodes strongly affixed to the sample vessel/forearm. Each light guide’s instrument fitting had the same alignment and clearance gap to the PMT. Guide bending was not noticeable and the apparatus was left undisturbed to avoid sensor movement. Opaque shielding totally surrounded the sample, light guide, and instrument interface. Therefore we feel that efficiency was maintained between trials and was superior to that normally achieved in a clinical setting.

The PMT could not directly be tested to ensure it was equally sensitive to all photons exiting the light guide regardless of exit angles. Nor could the PMT/light guide alignment be shifted to test for maximum/minimum detection. The gap between the PMT light guide apertures could not be altered without causing a mechanical misfit that caused light leakage and misalignment.

A different choice of sample media could yield different results. We based our choice on having a comparison to typical clinical NIRS data collections such that the phantoms would have superior SNRs to that of the forearm. Since the forearm has mixed chromophore concentrations, i.e., myoglobin, hemoglobin, deoxyhemoglobin, and cytochrome c oxidase copper redox, that are not constant, both a stable and an unstable single chromophore phantom were required for comparison.

The stable PMT temperature, INC DARK, and PC DARK signals and their magnitudes indicate a very high level of extraneous light occlusion, that remained intact throughout any given trial, and was consistent between trials. Therefore, the observed changes in the PMT applied high voltage during and between trials is likely almost all due to the transmission efficiency of the receiver being tested.

The greater variation in PC Dark Count during the unstable phantom trials may be due to residual internal reflection from the uncontrolled settling precipitate being counted at the start of the cutoff cycle rather than at the end of the laser firing cycle.

The PMT applied high-voltage values, in Figure 1, imply that some receivers are equally efficient at light transmission (e.g., single bundle at 2249 V versus liquid at 2250 V) since the applied voltage is in proportion to the illumination present. However, here there is no indication of the rate of change in illumination occurring every second, and consequently, no indication of signal quality.

Table 1 indicates that there is no significant difference between the linear array and the single bundle NIRO-500 Type C standard fiber optic on any of the test objects. There is also no significant difference between the dual-branched and the single-bundle fiber optic receivers on the human forearm and the unstable phantom. However, there is a modest improvement in signal noise using the dual-branched fiber optic on the stable phantom, and in that case, it is superior to either the linear array or liquid light guide as well. The liquid light guide is significantly different from the single-bundle standard and had from 51 to 149 greater signal noise than the single-bundle standard. In all cases, the liquid light guide had greater signal noise than any of the fiber optic receivers.

Contrary to our observations, liquid light guides should have superior performance characteristics to that of fiber optics because they do not suffer packing fraction transmission losses and typically have a two to three times wider acceptance cone. However, our sample concentrations are graded as stable, slightly unstable, and greatly unstable, and cause different patterns of change in photon scattering (Figure 3). The liquid light guide’s wide acceptance cone may enable it to acquire a greater photon density containing more diversely scattered photons, thus making it relatively more sensitive to scattering than fiber optic bundles. Scattering can easily be minimized in lab bench experiments so that the superior attributes of liquid light guides can be realized. However, our interest is in clinical applications that sample living tissue in pathological states. Consequently, performance must be evaluated from trials replicating less than ideal settings.

Figure 3

Comparison of the excitation voltage applied to the PMT (high voltage) to enable the photoelectric effect, for all trials as described in Figure 1.

In our study, the liquid light guide had one half to three times more signal noise than the fiber optic bundle configurations. This would preclude its use for noise-sensitive studies such as monitoring of cytochrome a, a₃ redox status, or measuring of cerebral blood flow via inspired oxygen manipulation. However, these are not excessive noise levels for many other clinical monitoring protocols where investigations would benefit from using liquid light guides to overcome the difficulties associated with fiber breakage.