2.1.

Fluorescence Lifetime Spectroscopy System

The assembled time-resolved fluorescence system uses two lasers: one emitting at 378-nm wavelength and another at 445 nm (BDL-375-SMC and BDL-445-SMC, Becker and Hickl, Berlin, Germany). These wavelengths were primarily chosen to excite NADH and FAD molecules in biological tissues in vivo. The repetition rate can be set at 20, 50, or 80 MHz, and the laser pulse width is around 50 to 100 ps. The excitation light was coupled to a bifurcated fiber (BIF400-UV-VIS, Ocean Optics, Dunedin, Florida) with a common end of two $400\text{-}\mu \mathrm{m}$-diameter fibers, positioned side-by-side. This optical fiber probe delivers the excitation light to the sample, and the collected light passes through bandpass filters before arriving at a hybrid photomultiplier tube (PMT) detector (HPM-100-50, Becker and Hickl, Berlin, Germany) with detection range from 400 to 900 nm and virtually free of after-pulsing effect, which could distort our fluorescence decay by generating delayed photon counts that are not expected in our measurements. These delayed counts may occur due to three processes: (a) when photoelectrons, produced by ionized atoms of the residual gas in photomultipliers, are accelerated toward the photocathode; (b) luminescence of this residual gas; (c) fluorescence effects of dynodes.^46,47 Two different bandpass filters at $440\pm 20\mathrm{nm}$ and $514\pm 15\mathrm{nm}$ were placed in front of the detector to mainly provide detection of NADH and FAD spectral emission regions at 378 and 445 nm, respectively. In addition to these filters, longpass ones (405-nm longpass for 378-nm excitation and 475-nm longpass for 445-nm excitation) were used to measure the signal from a broader range of the emission spectrum. Fluorescence lifetime measurements were performed using TCSPC boards (SPC-150 and DCC-100, Becker and Hickl, Berlin, Germany). The same lasers (378 and 445 nm) are used for fluorescence spectroscopy measurements. The fluorescence spectrum is collected using a portable spectrometer (USB2000-FLG, Dunedin, Florida). Figure 1 shows a scheme of the assembled experimental setup.

Fig. 1

Schematic drawing of the components for acquisition of fluorescence spectra and fluorescence lifetimes. The excitation light of one of the diode lasers (378 or 445 nm) is delivered to the sample through a bifurcated fiber optic probe. The fluorescence and backscattered light of this sample are collected using the same probe, which will send the light to the filter holder. A combination of four possible filters can be used to remove the backscattered light and acquire the spectral region of interest: bandpass filters at $440\pm 20\mathrm{nm}$ or at $514\pm 15\mathrm{nm}$ and 405- or 475-nm longpass filters. The fluorescence goes to a spectrometer or a hybrid PMT, and, then, the fluorescence spectrum and fluorescence decay curve are measured. The PMT is connected to two TCSPC boards: the detector control module (DCC-100) and the time-correlated single photon counting module (SPC-150).

2.2.

Mechanical Encasement of the Clinical System

A $48.5\mathrm{cm}\times 38.5\mathrm{cm}\times 19\mathrm{cm}$ metallic suitcase encases all the components, and the equipment can be plugged into an outlet using an IEC 60320 C13 computer power cord. The mechanical structure of the system was planned so that it could be carried without relevant optical misalignments, unusual exposure to intense light sources, dust, and chemical substances. The front panel (consisting of the external controls of the lasers, filter exchange module, power button, and USB connections for mouse and keyboard) of the portable system becomes accessible to the user by support of a shelf in upper part inside the suitcase (Fig. 3), which also optimizes the arrangement for minimum volume and appropriate weight distribution. In addition, orientation and intercomponent distances were set to allow their fixation and suitable refrigeration.

Our group used this fluorescence lifetime spectroscopy system for successful in-vivo measurements.^15,16 Previously, it was assembled over an optical bench and should be disassembled to be transported. However, after its assembly inside a suitcase, it can be carried without setting up cables and TCSPC boards connections in the place where the system will be used, transporting of separated parts, and assembling its structure. After transporting the suitcase by hand and/or in a vehicle, there were no significant differences in our calibration measurements (data not shown). When assembled in the suitcase, its volume was reduced to 38% (final volume: $\mathrm{38,166}{\mathrm{cm}}^3$) of its previous one, as the original volume included the computer fixed in a separate larger structure. In addition to the decrease in system volume, it has mechanical, electrical, and optical stability and allows fast changing optical filters during clinical measurements. Furthermore, a metallic fiber tip (Fig. 2) was designed to provide easier handling by the system operator and, consequently, fewer variations of pressure, illumination, and angulation during clinical measurements. Since the optical system is optical fiber based, it is possible to reach any external part of the body for measurements, making this system available for direct or endoscopic measurements, and further coupling to fluorescence lifetime imaging microscopy measurements.

Fig. 2

Fiber tip used for more stable measurements in different environments. (a) Dimensions of fiber tip in its top view and cross-section. (b) Fiber tip assembled with the support structure (excluding screws) to hold the fiber optic probe inside the tip. A comparison of sizes of the fiber probe with and without the tip can be made in the photograph displaying each part of the assembled fiber tip.

The system components are all enclosed in the suitcase, which includes a cover that eliminates contact between the system operator and electrical/optical components. This cover also provides a better-designed front panel, safety in clinical use, and protection in case of liquid or fluid spillage, as well as allows for surface disinfection. The control panel (Fig. 3) was customized to allow direct and immediate access to the most useful features for the system operator.

Fig. 3

(a) System assembly after its encasement provides protection for its components and a user-friendly design by exhibiting a control panel with access to external controls and the filter holder. The sample can be placed in any region around the system within the range of the optical fiber (2 m of length). (b) Top view schematic showing the position of each of the system components inside the suitcase. Parts in the upper part of the suitcase, such as 445-nm laser and external controls, are supported by a shelf, which holds the some of the components of the front panel.

2.3.

System Characterization

We performed measurements of the fluorescence lifetime of the fluorescent molecule Rhodamine 6G and the instrument response function (IRF) to evaluate the calibration of the system and determine the resolution of measured lifetimes. The temporal broadening of the laser pulse due to its propagation in multimode fibers, i.e., its temporal width, was measured after it passes through the excitation and collecting fibers and after it strikes a scattering medium [solution 0.5% Lipofundin^® (Laboratories B. Braun S.A., Rio de Janeiro, Brazil) in distilled water]. A $2.5\mu \mathrm{g}/\mathrm{mL}$ Rhodamine 6G (Sigma-Aldrich, St. Louis, Missouri) in distilled water solution, used for verifying the system calibration, was characterized by its absorption and fluorescence spectra using a 475-nm longpass filter, and 378 and 445 nm as excitation wavelengths. The average fluorescence lifetime of this solution was compared to the values reported in the literature for the assessment of the calibration.

To estimate the functionality of the customized fiber tip in the reproducibility of our in-vivo measurements, a preliminary test was performed by evaluating their standard deviations when using or not using the tip at different conditions: in a dark environment or with illumination from fluorescent lamps, with the fiber at 45-deg surface inclination, and with the fiber optic probe gently touching the tissue surface. Ten measurements were performed for each condition, at the volar forearm of one volunteer, always at the same point and by the same operator.

2.4.

Clinical Measurements

The subjects were evaluated at the Dermatology Ambulatory of Amaral Carvalho Hospital (Jahu, Sao Paulo, Brazil). The investigated lesions were delimited by the dermatologist and cleaned with a colorless detergent solution (an alcoholic solution of 0.5% chlorhexidine) without any fluorescence emission under the excitation wavelengths used in this study. Skin lesions measurements were performed at five sites inside each lesion area and five sites at clinically normal tissue in the surrounding area of each lesion (at the same anatomical site), delimited by the dermatologist. The study protocol was approved by the Research Ethics Committee of Amaral Carvalho (38719514.8.0000.5434), and the six enrolled patients signed an informed consent form. To evaluate the clinical feasibility of our system, three hypochromic nevus (HN), three BCCs, and adjacent normal tissue for each of the six abnormal regions were interrogated.

2.5.

Spectral Data Processing

Each acquired decay profile was processed using the SPCImage software (Becker and Hickl, Berlin, Germany). For the determination of Rhodamine 6G lifetime, the fluorescence decay curve was fitted by a monoexponential decay for both 378- and 445-nm excitations, 20- and 50-MHz repetition rates. These repetition rates were evaluated to check the reproducibility of our fluorescence lifetime values when considering the trade-off between the time range and temporal resolution of the acquired fluorescence decay, since we can record only 1024 points in each measurement. For skin lesions measurements, the decay curve was fitted by bi- and triexponential decays. We obtained relative coefficients and lifetimes for these multiexponential decays by performing a fitting of the experimental fluorescence decay $F(t)$ [Eq. (1)] to the convolution between the IRF (about 300 ps of full-width at half-maximum) with a multiexponential decay function

The average intensity lifetime is the average amount of time a fluorophore spends in its excited state before returning to the ground state and is calculated by averaging $t$ over the intensity decay, which is proportional to $a\tau$.^10,48 On the other hand, the amplitude average lifetime is the lifetime a fluorophore would have if its steady-state fluorescence was the same as a fluorophore or multiple fluorophores with several lifetimes.^48–50 In biological tissues, even though both average lifetimes provide information about the microenvironment of targeted fluorophores in tissue, the first one is independent of the final steady-state intensity, whereas the second one is associated to the optical processes that will result in changes in this intensity. In our case, we expect to extract features from the local microenvironment of the fluorophores with the intensity ${\tau }_{\mathrm{avg}}$ and information from all the interrogated volume with the amplitude ${\tau }_{\mathrm{avg}}$.

The average lifetimes of each investigated biological tissue are individually shown on a boxplot graph (Figs. 5 and 6). The statistical differences between the groups were evaluated using a Wilcoxon rank sum test ($p<0.001$) when comparing HN with BCC, and a Wilcoxon signed-rank test ($p<0.001$) for comparisons between normal tissue and each skin lesion considering the Bonferroni correction for multiple comparisons.

3.1.

Instrument Response Function and System Calibration

The IRF was obtained by measurements of the temporal pulse width after the laser pulse passes through excitation and collecting fibers and a little distance (typically 100 to $300\mu \mathrm{m}$ for the wavelengths used in this study) in a medium with high scatterer concentration, which induces an immediate scattering of the laser pulse and a minimal detection of deeper ballistic photons. The IRF measurements showed that the laser pulse temporal width is 280 ps for the 378-nm laser and 310 ps for the 445-nm laser, making it possible to obtain accurate measurements for most of the average lifetimes of molecules present in biological tissues (200 to 5000 ps). In addition, this temporal width allows a satisfactory fluorescence lifetime resolution to distinguish the shortest lifetime among the fluorophores to be monitored in future biological tissues (NADH and FAD), which can reach lifetime values between 300 and 400 ps.

After determining the IRF, we evaluated the Rhodamine 6G solution by its absorption and fluorescence spectra, which were similar to the ones presented by its manufacturer.⁸ The Rhodamine 6G fluorescence decay was measured using 20- and 50-MHz laser repetition rates and both measurements resulted in the average lifetime obtained for Rhodamine. The average value was $(4.1\pm 0.2)\mathrm{ns}$, which is the same as what is reported in the literature: 4.1 ns (in water solution).^51–54 The agreement between these values indicates that our fluorescence lifetime spectroscopy system was properly calibrated.

3.2.

Customized Fiber Tip Improves the Stability of In-Vivo Measurements

To improve the reproducibility of the clinical measurements in biological tissues, a fiber tip was designed to make the measurements less influenced by distinct clinical environments, especially concerning ambient light and operator handling. The tip operation was evaluated by simulating the conditions where future measurements could be performed in clinics. Measurements of volar forearm tissue with or without the fiber tip showed no difference among steady-state fluorescence spectra normalized by the intensity maximum (data not shown).

For time-resolved measurements, since the increase of the number of exponential components may induce unwanted overfitting of the fluorescence decay, we would like to use as less exponential components as possible and extract the highest amount of biochemical information we can. Then, we calculated average fluorescence lifetimes for bi- and triexponential decays in order to evaluate potential fluctuations due to changes in the composition of the interrogated tissue volume for each environmental condition (angled at 45 deg, illuminated, and pressured) and due to the solution obtained from the fitting process. In our measurements, both bi- and triexponential fittings showed the same value for the average lifetime for all the groups, which could indicate that the biexponential fitting was sufficient to describe fluorescence decays (Figs. 4 and 5).

Fig. 4

Amplitude fluorescence lifetimes and their respective CVs when using bandpass filters for (a and b) 378-nm and (c and d) 445-nm excitations. Measurements using the probe were more stable and reproducible.

Fig. 5

Intensity fluorescence lifetimes and their respective CVs when using bandpass filters for (a and b) 378-nm and (c and d) 445-nm excitations. Measurements using the probe were more stable and reproducible.

Measurements using the fiber tip showed a higher reproducibility and stability. This stability can be confirmed by lower variations of average fluorescence lifetimes on measurements in a dark environment (Figs. 4 and 5). Furthermore, when the fiber tip was used, we observed fewer changes in both mean values of intensity and amplitude average lifetimes among the different conditions presented in this study (Figs. 4 and 5). The coefficient of variation (CV) was also smaller for most of the cases when using the tip. When the probe is pressed in tissue, almost no difference in the CV was obtained using this tip. Regarding measurements using bandpass filters, the use of the tip allowed more stability compared to the longpass ones, whose use led to very similar average lifetimes in the presence or absence of the tip (data not shown).

3.3.

System Validation: In-Vivo Measurements of Skin Lesions

The system validation was performed by comparing the fluorescence spectra and lifetimes for normal tissue and two types of skin lesions (HN and BCC). The fluorescence spectra for 378 nm were very similar for the three tissue types, but, when using the 445-nm excitation, HN and normal tissue had slightly lower intensities compared to BCC in the 520- to 700-nm and 580- to 700-nm spectral regions, respectively (Fig. 6).

Fig. 6

Average fluorescence spectra of normal tissue, HNs, and BCC for (a) 378-nm and (b, c) 445-nm excitations.

To show the data distribution for each interrogated tissue (HN, BCC, and normal tissue), we chose to present them in boxplot graphs of average fluorescence lifetimes (Figs. 7 and 8). In these boxplots, the upper and lower lines indicate the 25th and 75th percentiles, whereas upper and lower whiskers indicate the 5th and 95th percentiles, respectively. The line in the middle of the box represents the median, and the point in its center is the mean of the data distribution. Based on these boxplots, the fluorescence lifetime values showed a tendency to be lower for BCC when using the 378-nm excitation [Figs. 7(a), 7(b), 8(a), and 8(b)] and for HN when using the 445-nm excitation [Figs. 7(c), 7(d), 8(c), and 8(d)].

Fig. 7

Boxplot graphs of amplitude average fluorescence lifetimes when using bandpass filters for (a and b) 378-nm and (c and d) 445-nm excitations. For 378 nm, lifetimes were lower for BCC, and for 445 nm, for HN, when compared to the other groups. The amplitude values showed an improved discrimination when compared to the intensity average lifetimes. The (*) symbol shows each pair with significant statistical difference ($p<0.001$, Wilcoxon rank sum test when comparing HN with BCC, and $p<0.001$, Wilcoxon signed-rank test for differences between normal tissue and each skin lesion).

Fig. 8

Boxplot graphs of intensity average fluorescence lifetimes when using bandpass filters for (a and b) 378-nm and (c and d) 445-nm excitations. The * symbol shows each pair with significant statistical difference ($p<0.001$, Wilcoxon rank sum test when comparing HN with BCC, and $p<0.001$, Wilcoxon signed-rank test for differences between normal tissue and each skin lesion).

We could discriminate between each combination of clinically similar pair-tissues (HN and BCC, HN and normal tissue, BCC, and normal tissue) showed in this study. The significant statistical difference [Wilcoxon rank sum test ($p<0.001$) for comparisons of HN between BCC and Wilcoxon signed-rank test ($p<0.001$) for normal tissue and each lesion] is indicated by the star symbol in Figs. 7 and 8. A better differentiation was observed in the average amplitude lifetimes. By combining data from 378- and 445-nm excitations, the discrimination between each pair of tissues in an automated system may be possible. Moreover, this combination can improve the differentiation among tissues when using multivariate analysis techniques.

We believe the main reasons for this differentiation are the concentration of chromophores in tissue, such as melanin and hemoglobin, and changes in the fluorescence of keratin, NADH, FAD, elastin, and collagen. The first fluorophore is mostly present in the outermost stratum of the epidermis (keratinocytes). A higher contribution of the second and third ones arises from cells in part of the dermis and in the lower sections of the epidermis.³³ Depending on the thickness of the epidermis, we may be able to detect differences of collagen and elastin in the dermis. The dermis has a dense collagen network, which, in BCC lesions, is degraded by tumor-associated matrix metalloproteinases. Then, this degraded collagen is replaced by cell-rich BCC tumor nests.³⁴ The tumor cells have a different metabolic activity due to the Warburg effect,⁵⁵ thus leading to changes in free and bound states of NADH and FAD, and the fluorescence lifetimes we measure.¹⁶ Also, differences in collagen content and elastin microenvironment may be causes of the changes in the fluorescence lifetime values we obtained. In regards to the contribution of the chromophores, since melanin is a strong absorber for short wavelengths,⁵⁶ it decreases the light penetration of the excitation sources we are using (378 and 445 nm). Then, we can collect fluorescence from deeper layers of skin in HN, which has less melanin content than normal tissue and BCC.⁵⁷ Also, due to hemoglobin absorption, increased blood content in BCCs could also attenuate both excitation and fluorescence light.³⁴ When using the average amplitude lifetime for 378-nm excitation measurements, it was possible to differentiate all combinations among lesions and normal tissue. This suggests that contributions of fluorescence of collagen, elastin, NADH, and absorption of melanin and hemoglobin are crucial factors for this differentiation. For all cases, a shift to lower values of both amplitude and intensity was observed when comparing normal tissue with BCC. This trend was also observed in other results reported in the literature,^30,33,34 in agreement with their description. Using the same statistical test as by Thompson et al.,³⁰ we achieved a $p$-value of $6\times {10}^{-5}$ for analysis of the pair HN–normal tissue, which is similar to their result of a $5.6\times {10}^{-6}$ $p$-value for the comparison between measurements in situ of normal skin tissue and a set of lesions (BCC, SCC, benign naevus, dysplastic naevus, malignant melanoma, and AK). In addition, we could discriminate among similar lesions, as described by Galletly et al., who observed significant statistical differences ($p<0.05$, Mann–Whitney U-test) in the average fluorescence lifetime when comparing measurements ex situ of nodular and superficial BCCs with infiltrative, micronodular, and mixed subtypes of BCC in an ex-vivo study.³⁴ Although these are preliminary data to validate the system, they show the potential of a portable fluorescence lifetimes spectroscopy system for clinical measurements and investigation of biological tissues in vivo.