Spectral distortion in diffuse molecular luminescence tomography in turbid media

Abstract

The influence of tissue optical properties on the shape of near-infrared (NIR) fluorescence emission spectra propagating through multiple centimeters of tissue-like media was investigated. Fluorescence emission spectra measured from 6 cm homogeneous tissue-simulating phantoms show dramatic spectral distortion which results in emission peak shifts of up to 60 nm in wavelength. Measured spectral shapes are highly dependent on the photon path length and the scattered photon field in the NIR amplifies the wavelength-dependent absorption of the fluorescence spectra. Simulations of the peak propagation using diffusion modeling describe the experimental observations and confirm the path length dependence of fluorescence emission spectra. Spectral changes are largest for long path length measurements and thus will be most important in human tomography studies in the NIR. Spectrally resolved detection strategies are required to detect and interpret these effects which may otherwise produce erroneous intensity measurements. This observed phenomenon is analogous to beam hardening in x-ray tomography, which can lead to image artifacts without appropriate compensation. The peak shift toward longer wavelengths, and therefore lower energy photons, observed for NIR luminescent signals propagating through tissue may readily be described as a beam softening phenomenon.

Figure 1

Photographs of phantoms used in this study. A liquid intralipid-based phantom (a) readily allows changing of the fluorophore concentration, while gelatin phantoms containing TiOB_2B scatterer shown in (b) eliminate the need for an external container. Gelatin phantoms without scattering material were also used to assess the emission spectra without scatter (c).

Figure 2

Normalized LuTex fluorescence emission as measured in dilute concentration (2 μg∕ml) in DI water.

Figure 3

A diagram of the experimental system with 16 spectrometers (Princeton Instruments Inc., Acton MA) is provided in (a) with a photograph shown in (b). Each spectrograph is coupled to a fiber optic, which surrounds a tissue phantom in a circular geometry (c). A homogeneous Teflon calibration phantom is pictured here with the 16 fibers coupled to the exterior in a circle.

Figure 4

Fluorescence emission of 5 μM LuTex measured in the multispectrometer system in phantoms composed of (a) gelatin and (b) DI water. These spectra represent the nonturbid baselines since neither phantom contained significant scattering media.

Figure 5

LuTex fluorescence emission experimentally measured at different source-detector distances in homogeneous scattering gelatin and intralipid based phantoms. The circular domain shown represents a cross section of the cylindrical phantom, illustrating the input (red inward-pointing arrow) and output measurement sites (blue outward-pointing arrows). For illustration purposes, the intensity (shown as logarithm of intensity) of the diffuse excitation field is plotted in the circular region.

Figure 6

Fluorescence emission spectra through a liquid phantom of DI water and varying concentrations of intralipid are shown. Only spectra measured through intralipid concentrations of 0.1% and above are shown for the two detectors farthest from the source.

Figure 7

Absorption coefficients of 100% water and 300 nM LuTex are plotted for discrete wavelengths (green line with data points) with LuTex fluorescence emission (red dotted line) in dilute solution, indicating the overlap in absorption and emission at the shorter wavelengths from 700 to 750 nm.

Figure 8

Diffusion based modeling of the fluorescence peak through a 60 mm diameter phantom, similar to Fig. 5. The dilute emission spectrum of LuTex is shown as a dotted line, and the calculated emission spectrum from diffuse emission through the region is shown at discrete wavelengths (black circles). Experimental measurements are shown in blue for intralipid liquid phantoms. The circular domain shown represents a cross section of the cylindrical phantom upon which the intensity (logarithm) of the excitation field calculated from the diffusion equation is plotted for illustration purposes.

Figure 9

Extinction coefficients of chromophores used in the simulation are shown in (a). These are used to calculate tissue absorption coefficients at any wavelength within the NIR. The simulated domain contains contrasts in the various absorbing constituents, the spatial distributions of which are shown in (b).

Figure 10

Fluorescence yield of LuTex is calculated from the drug concentration and the fluorescence quantum yield. Based on the concentration of LuTex assumed in Fig. 9, the target fluorescence yield is shown in (a). Reconstructing images from data collected with long-pass filtering can be an intractable problem as shown in (b), while resolving the emitted spectrum allows accurate recovery of the true fluorescence activity, as shown in (c). All images shown are in units of mm⁻¹.