2.1.

MDSFR/SFF Device

For the study, an MDSFR/SFF device was used. This setup has been described in detail elsewhere.¹⁰ Figure 1 shows a schematic of the system. In summary, the system consists of a 19-core coherent fiber bundle wherein each individual fiber is trifurcated at the proximal end to allow selective illumination and detection from the center fiber, the middle ring of six fibers, and the outer ring of 12 fibers, respectively. Via this trifurcation, every fiber is connected to a fiber delivering light from a halogen source (HL-2000-FHSA, Ocean Optics, The Netherlands), a fiber delivering light from a 405-nm LED and a fiber collecting light returning from the specimen and delivering it back to the spectrometer. The fiber diameter is independently regulated by three computer-controlled shutters, with effective diameters of 0.20, 0.60, and 1.08 mm. All measurements of the different fiber diameters are performed consecutively in a single measurement without lifting the probe. Detection was realized by three spectrometers (two S2000s, one USB2000+, Ocean Optics, The Netherlands) with an overlapping spectral range of 350 to 1000 nm and a long-pass filter with a 385-nm cut-off wavelength (GL-385-12, Avantes, The Netherlands) to remove fluorescence excitation light. For the fluorescence measurements, only the largest effective fiber diameter was used. The probe tip was polished at a 15-deg angle to reduce back reflections. Calibration of the system comprises integrating sphere calibration, reference optical phantom calibration, and calibrated lamp calibration. Details on this can be found elsewhere.¹⁰ This procedure takes roughly 3 min to perform and was done prior to every day the system was used. The coordination of illumination and detection, as well as the calibration and recording of the measurements, was performed by a laptop PC with a custom made LabView code.

Fig. 1

Schematic representation of the multidiameter single-fiber reflectance and fluorescence system. Numbers represent individual fibers. Both reflectance and fluorescence spectroscopy are measured in one single measurement. For reflectance spectroscopy, three fiber diameters are used. Panel (a) shows the $200\mu \mathrm{m}$ fiber, panel (b) shows the $600\text{-}\mu \mathrm{m}$ fiber, and panel (c) shows the $1000\text{-}\mu \mathrm{m}$ fiber, all connected both to the halogen lamp and spectrometers. Colors correspond to the reflectance spectra in Fig. 2(a). (d) For fluorescence spectroscopy, only the largest fiber diameter, $1000\mu \mathrm{m}$, is used in conjunction with a 405-nm LED and a spectrometer.

2.2.

Volunteers

For this study, six subjects were included for measurements with the MDSFR/SFF system (Table 1). The group consisted of a young cluster ($\text{age}<45\text{years}$) and a mature cluster. These latter volunteers were included to measure variability in mature skin and to investigate possible differences with younger skin. Inclusion criterion of these subjects was, therefore, aged $>60\text{years}$. None of the volunteers had any apparent skin disease and measurements were done on healthy looking skin. None of the volunteers reported diabetes mellitus, kidney disease, or smoking, factors that are known to influence skin autofluorescence.

Table 1

Summary of volunteers.

The research was done under the Dutch Code of Conduct for the Use of Data in Health Research.

2.3.

Measurement Strategy

The measurement scheme consisted of four locations, being the ventral side of the left forearm, the dorsal side of the left hand, the left cheek and left ala nasi. Each location was marked with a black ink ring, 3 mm in diameter. Around this ring, four marks were made 3 mm from the ring, at 12, 3, 6, and 9 o’clock, totaling five spots. Five measurements were done within the ring and between the ring and the marks surrounding the ring, totaling 25 measurements per location. Each measurement takes roughly 8 s to perform. Care was taken not to put the probe in contact with the ink ring. Between consecutive measurements, users were instructed to lift the probe from the skin and reposition it on the same spot. This was done to remove any effect of prolonged pressure by the probe on the skin, which could possibly lead to lowering of the oxygen saturation due to compression of the superficial microvasculature and potential changes in scattering properties. Users were asked to aim the consecutive measurements as precisely as possible on the same spot. All measurements of all users on one particular subject were done on the same day to correct for possible changes in ambient conditions.

2.4.

Measurement Outcomes

Blood and tissue optical properties were determined from the MDSFR spectra. A custom made MATLAB script was used to extract values for blood oxygen saturation (${\mathrm{StO}}_2$), blood volume fraction (BVF), vessel diameter (VD), average gamma (${\gamma }_{\mathrm{ave}}$) where $\gamma =(1-g_2)/(1-g_1)$ and $g_1$ and $g_2$ are the first and second moment of the phase function, respectively, reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }$ at 800 nm, and integrated intrinsic fluorescence excited using 405 nm ($Q{\mu }_{\mathrm{a}}^f$). A detailed description of the MDSFR analysis have been described previously.⁸

The script automatically excluded multidiameter spectra where spectra from the different fiber diameters overlapped (minimal wavelength 625 nm), if the maximum residual of reflectance is higher than 10% or if the maximum residual fluorescence is higher than $3\times {10}^{-5}/\mathrm{mm}$. To assess variability between users, all measurements of one location were pooled to extract group averages.

2.5.

Statistical Analysis

User variability was assessed by grouping measurements of individual users within the particular subset (i.e., volunteer and location). Subsequently, the user average and standard deviation were calculated. For this study, we defined that if the group average was within the standard deviation of the user, that user was within the system’s limit and variability was acceptable. Spots with less than three measurements were excluded from the analysis. A one-way analysis of variance (ANOVA) was used to assess differences in the fitted parameters between users.

2.6.

Validation

Optical measurements may be influenced by handling of the probe in two ways. First, the probe might not be properly placed on the surface of the tissue. The probe tip is polished at a 15 deg angle, which requires a slight angle at which the probe is to be placed on the tissue. If this is not the case, or if the probe is lifted during the measurement, room light might enter the fiber, which is detrimental to the correct recovery of reflectance and fluorescence spectra. However, this ambient light is easily recognized in the acquired fluorescence spectrum, and the measurement can immediately be retaken.

The second cause of error during measurement, possibly affecting the optical properties, is the pressure at which the probe tip is held against the tissue. For the MDSFR/SFF system, it is preferable to exert as minimal pressure on the probe tip as possible. If the application pressure is too high, this could lead to dimensional and physiological changes in the tissue that would ultimately alter optical properties. If the pressure is too high, BVF will drop as the blood is forced out. Furthermore, ${\mu }_{\mathrm{s}}^{\prime }$ will increase if tissue is compressed, since this property describes the density of the particles in the interrogated volume. Ultimately, the ${\mathrm{StO}}_2$ of the tissue will also decrease, since cells consume oxygen faster than it gets replenished. Contrary to disturbances in fluorescence due to room light, changes in these oxygen parameters will not be immediately apparent. A careful examination of variability between users is, therefore, required.

Another potential source of error is movement of the probe during acquisition of the individual MDSFR spectra and subsequent SFF spectrum. Spectra taken with the different fiber diameters are taken consecutively, and it is possible that the probe moves between spectra. This will affect the recovery of $\gamma$, ${\mu }_{\mathrm{a}}$, and ${\mu }_{\mathrm{s}}^{\prime }$ compared to the intrinsic fluorescence.

3.1.

Study Population

The current study focused on variability within as well as between different users of the MDSFR and SFF measurements. Five untrained users were asked to perform multiple consecutive MDSFR/SFF measurements on three test subjects. Furthermore, one test user performed measurements on three older volunteers. It is imaginable that the skin’s architecture changes with age, leading to fundamentally different optical properties which may lead to a larger variability between users. By including three older individuals, the effect of this could be investigated. All subjects were measured on the arm and face. Additionally, subjects were also measured on the hand and nose. These locations were added, since (pre-)malignant lesions are often found in these regions. It is important to be aware of any location-specific variability.

Table 1 summarizes the characteristics of the volunteers included in this study. The first group of volunteers (A, B, and C) with an average age of 36 years was measured by multiple users. The second group also consisted of three volunteers (D, E, and F), with an average age of 69 years. These were only measured by one user. Median skin type according to the Fitzpatrick scale was 2.

Figure 2(a) shows typical MDSFR spectra from the face. All data in Fig. 2 are from the same measurement on patient E. The double dip between 500 and 600 nm, particularly in the largest effective diameter, is a feature of oxyhemoglobin. Figure 2(b) shows a typical $\gamma$ curve. The wavelength dependence of $\gamma$, especially below 500 nm, is apparent from the figure.

Fig. 2

Example of (a) typical single-fiber reflectance and (b) phase function $\gamma$. The red (upper) line represents a spectrum measured with a diameter of $1000\mu \mathrm{m}$, the blue (middle) line represents $600\mu \mathrm{m}$, and the green (bottom) line represents a diameter of $200\mu \mathrm{m}$. Error bars denote one standard deviation. (c) Corrected intrinsic fluorescence $Q{\mu }_{\mathrm{a}}^f$ (d) with the residual. Light gray line represents measurement and black line represents fitted fluorescence.

3.2.

Variability within Users

After exclusion of the spectra that matched the exclusion criteria mentioned above, user variability was assessed in three volunteers (A, B, and C). Figures 3(a)–3(c) show the measured group averages of the blood parameters for each user on all subjects. Figures 3(d)–3(g) show the extracted optical properties of all measurements. For clarity, only the measurements done on the center spot of the arm and on the face at the central spot are shown ($n=5$ per user). Variability on the hand and nose and the other spots is, however, comparable to the trend seen in the arm and face as shown in the figures. One user had only two valid measurements on volunteer C. This user was excluded from the analysis.

Fig. 3

Variability of parameters (a) blood oxygen saturation (${\mathrm{StO}}_2$, %), (b) blood volume fraction (BVF), (c) vessel diameter (VD, mm), (d) average gamma (${\gamma }_{\mathrm{ave}}$), (e) reduced scattering coefficient at 800 nm (${\mathrm{mm}}^{-1}$), and (f) corrected intrinsic autofluorescence (${\mathrm{mm}}^{-1}$) of volunteers A, B, and C, measured by multiple users (one to eight). $\text{Average}\pm 1\text{standard deviation}$. Solid lines represent data collected from the arm and dashed lines represent data collected from the face.

Overall, vascular parameters showed more intrauser variability than the optical properties [Figs. 3(a)–3(c) versus 3(d)–3(f)]. Standard deviations were larger compared to the optical properties. With respect to ${\mathrm{StO}}_2$ [Fig. 3(a)], some users tended to display very narrow results, whereas others showed a relatively large variation in saturation. The variability of the BVF was uniform and individual users showed very little variability, especially with respect to measurements on the arm. Users 4 and 5 showed relative higher variability in their measurement on the face of volunteer B [Fig. 3(b), second panel]. Variability of the VD showed a similar trend [Fig. 3(c)], although users 6 and 7 showed far higher variability on the face of volunteer C compared to any other combination of user and volunteer [Figs. 3(c), third panel].

The scattering phase function parameter ${\gamma }_{\mathrm{ave}}$ [Fig. 3(d)] showed less intrauser variability compared to the blood properties. Intrauser variability of ${\mu }_{\mathrm{s}}^{\prime }$ (800) was uniform but quite large [Fig. 3(e)]. Variability of the intrinsic autofluorescence excited at 405 nm showed very low variability within users.

3.3.

Variability between Users

To examine variability between users, the spread of the individual users was compared with the measured group mean of the different variables. A maximum deviation of one standard deviation from the group mean was set as the cut-off point. Furthermore, ANOVAs were performed to assess this statistically. Measurements on the different locations were separated.

For ${\mathrm{StO}}_2$, most of the measurement on the arm fell within one deviation of the group mean. As described above, some users showed considerable variation compared to others. Measurements on the face showed a similar trend, although for volunteer B, users 3 and 5 measured higher oxygen saturations than users 2, 4, and 8. Indeed, ANOVA showed significant differences on all locations on all volunteers, except for the face of volunteer A and the arm of volunteer C (Table 2, $p>0.05$). BVF shows interuser variability that is overall within one standard deviation of the group mean. Variability is especially small on the arm. ANOVA only showed a nonsignificant difference on the face of volunteer A. Variability of the VD is also primarily within one standard deviation of the group mean. This is true for both the arm and the face of all volunteers. ANOVA confirmed the similarity of measurement on the face of volunteer A and the arm and face of volunteer C (Table 2, $p>0.05$).

Table 2

One-way analysis of variance (ANOVA) of differences of user variability on different measurement locations and volunteers, where the users comprise the difference groups of the ANOVA. p-values higher than 0.05 were considered nonsignificant.

The optical properties ${\gamma }_{\mathrm{ave}}$, ${\mu }_{\mathrm{s}}^{\prime }$ (800), and $Q{\mu }_{\mathrm{a}}^f$ showed interuser variability that was much more uniform and highly concordant with the respective nonsignificant ($p>0.05$) ANOVA calculations. For the average gamma (${\gamma }_{\mathrm{ave}}$), all users measured values within one standard deviation of the group mean in volunteers A and B on both locations. For volunteer C, user 1 on the arm and user 6 on the face reported standard deviations that did not overlap the group mean. All calculated ANOVA $p$-values are higher than 0.05, suggesting similarity between the users. This was also true for the reduced scattering coefficient at 800 nm. All users reported measurements that were within one standard deviation of the group mean, except measurements of user 1 on the arm of volunteer C. ANOVA results show highly nonsignificant $p$-values for all locations on all volunteers. Finally, for the corrected integrated intrinsic autofluorescence excited at 405 nm, all measurements, on the arm as well as the face, fell within one standard deviation. This translated to $p$-values above 0.05 for all locations on all three volunteers.

3.4.

Variability Differences within Different Locations

It is known that the skin of different parts of the body is structurally different. To investigate if this influenced variability, measurements of the arm and face were compared.

No conclusive differences were found concerning ${\mathrm{StO}}_2$ [Fig. 3(a)]. No real trend in variability differences between the arm and the face could be discerned. Standard deviations of BVF and VD of the arm were smaller than the face [Figs. 3(b) and 3(c), respectively]. This was true for all user measurements on all volunteers. ${\gamma }_{\mathrm{ave}}$ showed a slightly broader standard deviation on the face compared to the arm [Fig. 3(d)]. This was true for almost all users on all volunteers. For ${\mu }_{\mathrm{s}}^{\prime }$ (800), no clear trend can be extracted from the results, although it seemed variability is either comparable between locations or higher in the face [Fig. 3(e)]. $Q{\mu }_{\mathrm{a}}^f$ did not show any clear differences in variability in the arm or face [Fig. 3(f)]. Some large individual differences were observed, but no overall conclusion could be drawn from these results.

3.5.

Variability in Mature Skin

To determine the intrauser variability of vascular and optical properties in older individuals, one user repeated the measurements in a second group of volunteers with mature skin (volunteers D, E, and F). Measurements were only done on one spot on the arm, hand, face, and nose. To be consistent with the first group of younger volunteers, every location was measured five times. Results of the mature skin are shown in Fig. 4. For reference, scales of the graphs are the same as the corresponding previous graphs in Fig. 3. It is clear that the ranges of intraperson variability with respect to ${\mathrm{StO}}_2$ in the mature skin were similar compared to that of the young volunteers. Intraperson variability of BVF was also similar in young and mature skin. Overall, VD was comparably variable in mature skin compared to young skin. The VD on the arm of volunteer E is very variable due to two very high data points. Furthermore, volunteers D and F showed less variability on the arm compared to the face, which is in accordance with what was seen in younger volunteers A to C.

Fig. 4

Variability of parameters (a) ${\mathrm{StO}}_2$ (%), (b) BVF, (c) VD, (d) ${\gamma }_{\mathrm{ave}}$, (e) ${\mu }_{\mathrm{s}}^{\prime }$ (800) (reduced scattering coefficient at 800 nm, ${\mathrm{mm}}^{-1}$) and (f) $Q{\mu }_{\mathrm{a}}^f$ (corrected intrinsic autofluorescence, ${\mathrm{mm}}^{-1}$) in mature skin of volunteers D, E, and F, measured by one user.

${\gamma }_{\mathrm{ave}}$ showed similar intraperson variability in mature skin compared to young skin. Variability of ${\mu }_{\mathrm{s}}^{\prime }$ (800) was notably smaller in mature skin. This was true for both the measurements on the arm as well as the face. Last, $Q{\mu }_{\mathrm{a}}^f$ showed a slighter higher variability in mature skin.

3.6.

Variability of Scattering Due to Different Fiber Diameters

Since the skin is morphologically subdivided in different layers, all of the layers have their specific cell types and structures. The depth of interrogation of the probe depends on the diameter. The smallest diameter interrogates the shallowest depth, the largest diameter the deepest part. To examine whether the fiber diameter would explain variability in scattering, the reduced scattering at 800 nm was examined for all three fiber diameters. For this analysis, the data of volunteer B were used and separated by fiber diameter (Fig. 5). Shown are the mean value ($\pm \text{standard deviation}$) for ${\mu }_{\mathrm{s}}^{\prime }$ (800) for users 2, 3, 4, 5, and 8. Pooled by user, the size of the error bars is roughly equal, except for the ${\mu }_{\mathrm{s}}^{\prime }$ (800) measured with the 600-nm diameter by user 2. Mean scattering coefficient shows a trend of being higher when measured with the 600-nm fiber but no trend is seen with regard to the size of the variability of ${\mu }_{\mathrm{s}}^{\prime }$ (800).

Fig. 5

Variability of ${\mu }_{\mathrm{s}}^{\prime }$ (800) (reduced scattering coefficient at 800 nm, ${\mathrm{mm}}^{-1}$) on volunteer B, subdivided by fiber diameter ($x$-axis) and user (different panels).