1. Introduction

Optical spectroscopy of turbid media is a fundamental tool for the characterization of a variety of highly diffusive materials, that enables analysis of biological tissues, food and powders, among others. In biomedicine, clinical instruments exploit optical techniques to provide useful insights for both diagnostic and therapeutic purposes. In the last years, a particular effort has been devoted to the development of instruments that - simultaneously and non-invasively -provide the absorption μa and the reduced scattering μ_s′ coefficients over the 600–1000 nm spectral bandwidth: this range is interesting for tissue diagnosis because the low absorption of light enables the in-depth investigation of a sample. Spectroscopic systems are advantageous compared to instruments working at a single or at few wavelengths because they yield more robust and accurate measurements, particularly in the presence of noise. In some cases, the use of a limited number of wavelengths can give rise to an incomplete representation of the tissue under study due to great variations in optical properties within the spectra. In general four major absorbers can be individuated in tissues: deoxy-hemoglobin, oxy-hemoglobin (Hb and HbO₂, respectively), lipid and water, but spectroscopic methods can also cater to other possible chromophores in the tissue [1]. Furthermore, broadband optical characterization of biological tissue provides fundamental information for the design of a system for pre-clinical or clinical applications like optical mammography and tissue oximetry.

Only a few systems, based on time-resolved (TR) [2, 3], steady state (SS) [4] or combined frequency domain and SS techniques [5] have been developed. In particular, time-resolved spectroscopy (TRS) offers great sensitivity with the capability of uncoupling the contribution of absorption from that of scattering and to probe the sample at different depths. Yet these instruments commonly have the drawback of being cumbersome and difficult to use, therefore not being suitable for portable monitoring. The recent development of new photonic and detector technologies [6, 7] has opened the way to the implementation of clinically compatible broadband systems. Recently, TRS instruments have been developed, taking advantage of the large spectral width of the supercontinuum generated by photonic crystal fibers (PCF) pumped by a titanium-sapphire laser [8, 9, 10]. However these systems do not yet fulfill the requirement of portability because of the use of solid state laser sources. Furthermore, currently available photomultiplier (PMTs) have the limitation of a restricted spectral bandwidth, poor temporal resolution or the need to provide water cooling and nitrogen pumping [3].

In this work we present the first (to our knowledge) portable, clinically compatible TRS system covering the entire 600-1000 nm bandwidth. By employing a PCF pumped by a fiber laser the dimension of the system has been significantly reduced. Furthermore, a single photon avalanche diode (SPAD) has been used as the detector, to combine fast timing resolution and large spectral efficiency, without the need of water cooling.

2. Material and methods

2.1. Instrument layout

A schematic overview of the system is illustrated in Fig. 1. The light source consists of a fiber laser (SC450, Fianium, UK) providing white-light pulses shorter than 10 ps at a repetition rate of 20 MHz in the 465–1750 nm bandwidth with a total power of 2.6 W. The supercontinuum is spectrally dispersed by a F2-glass prism (PS854, Thorlabs, Germany) and then focused by an achromatic doublet with focal length of 75 mm (AC508-075-B, Thorlabs, Germany) on a 50 μm core graded-index fiber. The fiber - FC connectorized - is mounted with a 2-axis tilt on a 3-dimensional translational stage for precise alignment. Light entering the lens at different angles focuses in slightly different positions in the focal plane, therefore only a small spectral bandwidth couples into the fiber. The prism is mounted on a motorized rotational stage in order to sequentially tune the light source: it is rotated in 120 steps of 1.1 mrad to cover the 600–1000 nm bandwidth. A 1 mm pinhole is placed after the prism to limit the incident power at the fiber head and to avoid dangerous back reflections to the laser. The power can be adjusted by means of a motorized circularly variable neutral-density filter placed in front of the fiber. A small part of the light (3%) is split by using a fused splitter (Phoenix, UK) and delivered to a spectrometer (USB2000, Ocean Optics,USA), for monitoring porpouse. The remaining light (97%) is guided to the sample. The power at the probe -determined by the source spectrum, the wavelength dependent optical losses and the fiber coupling efficiency- is shown in Fig. 2. Ripples are due to the spectral nonuniformity of the supercontinuum light. The full width half maximum (FWHM) bandwidth around the tuning wavelength is also presented, increasing from less than 5 nm at 600 nm to about 20 nm at 1000 nm, because of the non-linear dispersion of the prism.

 

Fig. 1. System layout. Shown in Fig.: prism (P), lens (L), pinhole(PH), neutral density variable filter (ND), fiber connector (FC), beam splitter (BS), optical spectrum analyzer(OMA), single photon avalanche diode (SPAD), time correlated single photon counting board (TC-SPC).

Download Full Size | PDF

Light diffusely transmitted through the sample is detected using a single-photon avalanche diode with 100 μm diameter (PDM-100, MPD, Italy), placed directly in contact with the sample. The signal from the detector is driven to a time-correlated single-photon counting board (SPC-130, Becker & Hickl, Germany); synchronization is provided directly from the laser. The temporal resolution of the overall system is about 70 ps, limited by the chromatic dispersion in the optical fibers and the SPAD temporal resolution. The acquisition of the time-resolved curves and the synchronous movement of the prism and the variable filter are automatically controlled by a PC. We calculate the optical properties by fitting the experimental data with the solution of the diffusion equation - convolved with the instrument response function - with extrapolated boundary conditions for a homogeneous slab [11]. Since the fit of time-resolved measurements is based only on the shape and not on the amplitude of the curves, the estimated optical properties do not depend on the presence of superficial absorbers (e.g. melanin).

 

Fig. 2. Measured power at the sample (red circles connected by lines) and spectral width (blue circles) of the 120 wavelength bands.

Download Full Size | PDF

2.2. Calibration and in-vivo measurements

The ability of the system to determine μ_a and μ_s′ spectra was tested first on tissue phantoms and then in-vivo. Measurements were carried out on epoxy phantoms with titanium oxide (TiO₂) particles as a scatterer, and black toner as an absorber [12]. Eight phantoms (4.5 cm thick and with typical optical properties of biological tissues) are presented: four phantoms with varying concentration of toner from 0 to 36 μg/g in steps of 12 μg/g and the same concentration of TiO₂ (1.4 mg/g); other four with the same concentration of toner (12 μg/g) and TiO₂ of: 0.7, 1.4, 2.1 and 2.9 mg/g. Additionally, the transmittance measurement was performed on the breast of a healthy female volunteer, 28 years old in age, with body mass index (BMI) of 17.3 kg/m². The tissue was compressed in the lateral direction to a thickness of 3.8 cm. The acquisition time was set to 1 s per wavelength, for a total of 120 s measurement time. The optical power at the sample was adjusted to have a count rate of about 0.3 MHz in order not to exceed the statistic limits of single photon counting.

3. Results and discussion

3.1. Instrument characterization

Figures 3 and 4 show the recovered spectra for μ_a and μ_s′ of the phantoms. The four additional lines represent the conventional values for the phantoms as previously calibrated using a laboratory setup for TR diffuse spectroscopy [3]. Comparing to the calibration values, an overestimation of μ_a at low absorption and underestimation at high absorption is observed. The deviation from the conventional values of μ_s′ increases for greater scattering. In general, the recovered coefficients are close to the calibration ones: the average deviation is approximately 9% for μ_a and 7% for μ_s′. Furthermore, a good linearity in absorption (4%) was estimated from Fig. 3, and for reduced scattering (2%) from Fig. 4. It is worth saying that the true optical properties of the phantom are not accurately defined and the problem of the assessment of absolute values of absorption and scattering is still significant, with a discrepancy among different system of more than 10% [12].

A unique quality of the present TRS system is its ease of use and maintenance: no daily optical re-alignment is required. The reproducibility of the measurements was evaluated by calculating the optical properties of the same phantom (concentration of toner and TiO₂ respectively of 12 μg/g and 1.4 mg/g) among measurements carried out on 10 different days. We found values of μ_a and μ_s′ that were reproducible within 2%. The stability of the system is worth noting. Approximately 30 minutes are required for the laser source to warm-up and, during operation, oscillations of the optical power are appreciable, particularly at the shorter wavelengths (<650 nm). This causes an uncertainty in the measurement that we evaluated by repeatedly acquiring the optical properties of one phantom for 2 hours and every 3 minutes. By averaging over the entire spectral range we found a coefficient of variation (CV) of 0.6% for μ_a and 0.5% for μ_s′, but at the most critical wavelength (600 nm), non-negligible CVs of 1.6% for μ_a and 1.7% for μ_s′ were calculated.

 

Fig. 3. Absorption spectra of the phantoms. The absorption increases with toner concentration (indicated in the label) and the scattering is kept constant. The solid lines represent the expected values.

Download Full Size | PDF

 

Fig. 4. Reduced scattering spectra of the phantoms. The scattering is increasing with titanium oxyde concentration (values are shown in the label) and absorption is kept constant. The solid lines represent the expected values.

Download Full Size | PDF

 

Fig. 5. Absorption (blue) and reduced scattering spectra (red) of the volunteer breast. The solid lines are the fitted spectra.

Download Full Size | PDF

3.2. In-vivo measurements

The measured absorption and reduced scattering spectra of the breast are shown in Fig. 5. The absorption spectrum was fitted for water, fat, Hb and HbO₂. The scattering spectrum was fitted to a power law of the type μ_s′=a(λ/λ ₀)^-b, with λ ₀=600 nm. The best fit to the data is shown with lines in the Fig. and provides the following concentrations: 53% for water, 16% for lipids, 35 μM for total hemoglobin content (THC=[Hb]+[HbO₂]), 71% for the oxygen saturation (St0₂=[HbO₂]/[Hb+HbO₂]). The resulting scattering a and b were respectively 17.3 cm^-1 and 1.13. The measurements at the shortest wavelengths (<650 nm) was challenging, because of high scattering, high absorption of hemoglobin and low available power. Therefore this spectral range has not been considered for the estimation of the constituents concentration and of the scattering parameters. Future work will be dedicated to the optimization of the optical setup to increase the power in the low part of the bandwidth, to improve the spectral resolution in the near-infrared as well as to increase the bandwidth of the measurement at both shorter and longer wavelengths. The rapid progresses in the field of fiber lasers and micro-structured fibers could lead, in the next years, to the construction of systems with spectral power density which currently can be achieved only with laboratory systems based on solid state tunable lasers [3].

4. Conclusion

In summary, we developed a TRS system for the optical characterization of highly diffusive media that implements a supercontinuum fiber laser source and SPAD detection covering the entire 600–1000 nm bandwidth. The system proved valuable for in-vivo measurement of the absorption and scattering spectra of the female breast and to determine the composition and the structural properties of the tissue. As the instrument is compact it can be easily transported to perform spectroscopy in a variety of applications of turbid media analysis.