2.1

Diffuse optical spectroscopy imaging

Important parameters that influence the trajectories of photons in tissue are the refractive index of the tissue n, the reduced scattering coefficient μs' and the absorption coefficient μa. While n can be considered constant (typically in the order of n=1.4), large variations exist in μs' and μa between different biological tissues. The coefficients μs' and μa are wavelength dependent and provide intrinsic tissue contrast (Figure 2).

DOSI uses intensity modulated laser diodes and NIR broadband illumination to separate the μs' and μa in the complete NIR range, between 650 and 1000 nm [5, 8] (Figure 1). Source and detector fibers are spatially separated, at a distance ρ. Depending on the modulation frequency, μs' and μa and ρ, the detected intensity signal is attenuated and phase-shifted with respect to the illumination source. As the modulation frequency is known, μ_s' and μ_a can be solved for (See details in [8]). In strongly scattering media (μ_s'≫μ_a) such as biological tissue and for large source-detector separations (ρ≫1/(μ_s'μ_a)), the diffusion approximation can be used. In this work four intensity modulated laser diodes (660, 690, 785, 830 nm) are used which perform a frequency sweep between 0 and 500 MHz. The modulated signal is detected using an avalanche photo diode (APD). Wide near-infrared spectral measurements as in Figure 2 are obtained from a halogen lamp, which response is measured using a spectrometer. From this broadband signal, the scatter and absorption coefficients along the complete spectrum between 650 nm and 1000 nm are estimated by calibrating this signal using the scatter and absorption estimates at the 4 modulated wavelengths (See details in [8]).

Figure 1:

Schematic representation of the DOSI measurement setup combining the frequency domain (FD) and broadband steady state (SS) acquisition. The positions of the (red) source fiber and the (green) detector fiber on the tissue are separated by a distance ρ.

Figure 2:

Optical properties for healthy colon and tumor tissue. (a,c) scattering and reduced absorption for healthy colon [10]. (b,d) scattering and reduced absorption for (breast) healthy and tumor tissue [4].

2.2

Expected colon tissue optical properties

Healthy colon tissue consists of a layered structure with varying optical properties. As we are currently unable to measure these properties using DOSI, we have to rely on literature for scattering and absorption estimates. In the scarce literature available with in-vivo colon measurements, Zonios, et al. [9] mainly report measurements of the superficial epithelial layer. Hidović-Rowe and Claridge [10] use in-vitro measurements of absorber concentrations and wavelength independent scattering properties (scatter size, volume fraction) to determine the absorption and scattering within the visible range and validate these in-vivo. Further literature, e.g. [11] report in-vitro measurements without in-vivo validation.

The wavelength independent absorber and scattering properties of the healthy colon layers in [10] allow us to compute the expected absorption and reduced scattering coefficients in the near-infrared region (650-1000 nm). In this work we follow their division of the colon in three tissue layers: mucosa, submucosa and muscularis with average thicknesses of 560, 540 and 800 μm, respectively. The absorption and reduced scattering coefficients in Figure 2a,c have been derived from the hemoglobin, oxygenation, water and lipid contents listed in [10] and the reduced scatter coefficients were estimated using Mie theory from reported scatter diameters and refractive indices.

Colon tumor tissue on the other hand does not have a clear layered structure. It starts growing at the inside of the colon wall and expands both into the colon and penetrates towards the outer layers up to thicknesses of several centimeters. Little information about the optical properties of colon tumor tissue is available from literature. Most work has concentrated on polyp [9] or ex-vivo measurements [11]. Due to lack of in-vivo colon tumor measurements in the near-infrared we will use the optical properties as measured from breast tumor tissue [4], shown in Figure 2b,d.

2.3

Monte Carlo simulations

Frequency domain Monte Carlo simulations [12] were performed to estimate for which source-detector separation distances a sufficiently high phase shift of the modulated signal would be obtained for the separation of scattering and absorption (Section 2.1). To this end the reflectance amplitude and phase were simulated using the tissue properties for both healthy colon and tumor tissue using 10⁵ photons.

To investigate which tissue layers in the healthy colon tissue mainly contribute to measured spectra, photon hitting densities (PHD) were estimated by multiplying the fluences of forward and adjoint Monte Carlo simulations [13]. Below the colon tissue layers a 20 mm inter-organ fat layer was added to prevent artifacts due to a tissue-air boundary. Histograms of PHD per tissue layer were computed to estimate the relative contribution of each tissue to the measured signal.

Monte Carlo simulations were performed for wavelengths covering the broadband spectrum: 660, 690, 780, 830, 880, 920, 960, 1000 nm. All layers were assumed to have a limited thickness, but having an infinite extent in the other two dimensions. All simulations were performed using the VirtualPhotonics toolbox (www.virtualphotonics.org).

2.4

Colon phantom

To validate the feasibility of measuring at short source-detector distances in tissues with the properties from Section 2.2, silicon phantoms were created with the properties in Table 1, where μa and μs' are the designed properties at 600 nm. The phantom mimicking healthy colon tissue was made as a homogeneous phantom. Its scatter and absorption coefficients approximated that of a (simulated) layered phantom when measured with a source-detector separation of 10 mm. The latter phantom also contained a 20 mm diameter hole to mimic the inside of the colon. The phantom optical properties were measured using DOSI with ρ∈{5,10,13} mm within the phantom cavity and with a multi-distance DOSI probe with source-detector separations at 15, 20, 28 and 35 mm at the phantom surface. The latter measurement counts as the optical properties ground truth.

Table 1:

Phantom specifications. SG: silicone gel, CA: curing agent, Ni: Nigrosin.

3.1

Monte Carlo: phase shift

Simulated phase shifts at 500 MHz between source signal and detected signal are shown in Figure 3 for healthy colon and tumor tissue. For the target source-detector distance of about 10 mm, phase shifts of about 40-60 degrees can be expected in both tissue types. Given a phase detection precision of about 1 degree with the current fibers, tissue dependent phase shift variations of about 2% can still be detected. For 2.0 mm and 5.0 mm, only phase shifts of 5-40% can be detected, preventing accurate tissue characterization.

Figure 3:

Simulated phase shift as a function of ρ in (a) healthy colon and (b) tumor tissue.

3.2

Monte Carlo: photon hitting densities

Photon hitting densities were computed for 8 wavelengths between 650 nm and 1000 nm for healthy colon and tumor tissue. A typical example at ρ =10 mm for healthy colon at 830 nm is shown in Figure 4a. Although one should note that the visualization uses a logarithmic scale, it is evident that not only the top layers of the colon tissue, but also the auxiliary fat layer contributes significantly to the measured signal at 10 mm source-detector separation. Histograms that show the relative contribution per tissue to the total measured tissue properties are shown in Figure 4b,c,d for a ρ of 5, 10 and 15 mm, respectively. Figure 4c shows that for ρ=10 mm and for most NIR wavelengths almost 50% of the signal comes from the auxiliary fat layer, which is not well defined. Therefore ρ<10 mm is preferable in the rectum. In the thicker tumor tissue (20 mm) practically all signal (>99%) came from the tumor for each simulated ρ and wavelength.

Figure 4:

Simulated (a) photon hitting densities in healthy colon at 830 nm for ρ=10 mm. The logarithmic color scale has base 10 and shows normalized data. Wavelength dependent relative photon hitting densities per tissue for ρ’s of (b) 5 mm, (c) 10 mm and (d) 15 mm.

3.3

Phantom measurements

The optical properties of the colon phantom (Figure 6a) were measured 5 times for ρ∈{5,10,13} mm in the phantom cavity and with the multi-distance probe on the surface. The averaged estimates for μ_a and μ_s' in the healthy colon phantom are shown in Figure 5. Apart from the source-detector separation at 5 mm, all measured spectra show the same trend. The spectra for 10, 13 mm and the multi-distance probe, do differ, but the small standard deviations show that these measurements are reproducible. Especially the difference between the 10 mm’ and multi’ curves show that caution should be taken to compare measurements at different source-detector distances. The deviations at ρ=5 mm are due to small phase shifts and associated almost constant intensity amplitudes over the complete frequency sweep. This prevented accurate model fits and thereby accurate absorption and reduced scattering estimates. At ρ=10 mm absorption and scattering estimates are sufficiently accurate and precise.

Figure 5:

Measurements of (a) absorption and (b) reduced scattering (averaged over 5 measurements) in the healthy colon phantom. Error bars denote standard deviations (shown at coarser sampling).

Figure 6:

(a) Colon phantom and (b) rapid prototyping fiber guide. Diagram of (c) side- and (d) front-view of rectal probe.