Determination of the distribution of drug concentration and tissue optical properties for ALA-mediated anal photodynamic therapy

Abstract

PDT efficacy depends on the availability and dynamic interactions of photosensitizer, light, and oxygen. Tissue optical properties influence the delivered light dose and impact PDT outcome. In-vivo measurements of tissue optical properties and photosensitizer concentration enable determination of explicit and implicit dose factors affecting PDT and helps to understand the underlying biophysical mechanism of PDT. In this study, we measure tissue optical properties (absorption μ_a (λ) and scattering μ_s’ (λ) coefficients) and PpIX concentration in tissue simulating liquid phantoms with a geometry that resembles anal canal. In-vivo light fluence rate and photosensitizer fluorescence of 405nm excitation light source were acquired using a dual-motor continuous wave transmittance spectroscopy system. We characterized the tissue optical properties correction factor of fluorescence signal using a series of tissue simulating phantoms with known PpIX concentrations and with absorption coefficient between 0.1 – 0.9 cm⁻¹ and reduced scattering coefficient between 5 – 40 cm⁻¹. The results demonstrated that our spectroscopy system could determine the distribution of tissue optical properties and PPIX concentration during anal PDT.

1. INTRODUCTION

PDT is a treatment modality that uses photosensitizers along with light to kill cancer cells (1, 2). The success of PDT depends on the accuracy of a prescribed light dose delivered to the tumor (3). Tissue optical properties are important when it comes to calculation of a needed light dose and determining the penetration depth of light in tissue. Useful information such as treatment depth can also be obtained from changes in tissue optical properties. Photosensitizer concentration is another important parameter to be measured during PDT dosimetry (3). Fluorescence spectroscopy is an important tool for the field of PDT as most photosensitizers are fluorescent dyes. Photosensitizer fluorescence can be used to quantify its concentration and to monitor its photobleaching during PDT. However, quantifying in-vivo fluorescence emission is very challenging because the measured fluorescence intensity can be affected by the spatial and temporal variation in tissue optical properties (4). These variations in the fluorescence intensity may be mistaken as the change in photosensitizer concentration if the effects of tissue optical properties are not accounted for carefully.

Our group has developed a dual-motor continuous wave transmittance spectroscopy system and algorithms for interstitial optical properties measurements (5, 6) and semi-infinite medium conditions (7). For anal PDT, determination of tissue optical properties within the anal canal is challenging as it involves measurements of light fluence rate inside a tissue geometry that resembles a cylindrical hollow cavity. We developed a method to measure light fluence rate within anal canal by incorporating our dual-motor continuous wave transmittance spectroscopy probe onto an anal PDT light applicator (8). An empirical model was then developed to extract tissue optical properties from measured light fluence rate within cylindrical hollow cavity (8). The method and algorithm have been validated against a series of liquid tissue mimicking phantoms and the average uncertainty in recovering μ_a and μ_s’ was 8.5 ±5.7% and 9.9 ±7.5%, respectively (9).

In this study, we modify our system device to acquire fluorescence spectra of 5-aminolevulinic acid (ALA) induced protoporphyrin IX (PpIX) within anal canal using the same anal light applicator that was used for PDT treatment and light fluence rate measurements. We develop an algorithm to separate the spectral contribution of PpIX from other components. We also report an empirical method to eliminate the effects of tissue optical properties on measured fluorescence spectra. The correction function was determined experimentally using a series of tissue simulating phantoms that simulate a wide range of clinically relevant optical properties. Finally, we test the accuracy and validate our method in determining PpIX concentration by using liquid tissue mimicking phantoms with known PpIX concentration.

2. METHODS

2.1. Light fluence rate and PpIX fluorescence measurements

The light detection device consisted of two parallel, 2 mm (outer diameter) light transmitting catheters (Flexi-Needle, Best Medical International, Springfield, VA) placed at a distance h apart. One of two catheters held a point light source while the other held a calibrated isotropic detector. The isotropic detector is scanned along the length of the light source-containing catheter using a stepper motor. The detectors used in this study are optical fiber-based isotropic detectors (Medlight, Switzerland) of the scattering-tip type (10). The light collected by the detector was digitized using a photodiode-based in-vivo light dosimetry system (11). The pair of catheters are placed 5 mm apart in a groove on the periphery of a cylindrical light applicator. The radius and length of the light applicator is 1.27 cm and 9 cm, respectively. Light fluence rate distribution inside the anal cavity is simulated by submerging the light applicator in liquid tissue-mimicking phantom.

For PpIX fluorescence measurements, a 45° side-firing fiber is inserted into one of the two parallel catheters. The fiber is connected to a 405nm laser used as the excitation light source for PpIX. The same fiber is used to collect fluorescence emission and the fluorescence signal is diverted by a dichroic mirror for detection using a CCD- based spectrograph (Princeton Instruments, Princeton, NJ), as shown in Figure 1. The fiber is guided by a stepper motor to perform a 13-step scan along the length of the containing catheter with 5mm separation between each measurement. This allows for measurements of fluorescence spectra for 6 cm along the anal canal. The acquisition time of each measurements is set to be 200ms and the total time needed to complete a linear scan is no longer than 3 seconds.

Figure 1.

Schematic drawing of fluorescence measurements using a 45° side-firing fiber inserted into one of the two parallel catheters.

2.2. Phantoms preparations

Liquid phantoms with known optical properties are used in this study. Black India ink (Higgins black India ink #4418, Bellwood, IL) is added as light absorbers to simulate tissue absorption; while Intralipid (stock concentration 20% Intralipid, Fresenius Kabi, Uppsala, Sweden) is added as elastic scatterers to simulate tissue scattering properties. PpIX is dissolved in dimethyl sulfoxide before added to the liquid phantoms to get different final concentrations. During measurements, the light applicator was covered by transparent plastic foil to avoid liquid filling the space in between the two catheters. The dependence of the absorption coefficient on the black India ink concentration was determined by measuring the transmission of various concentrations of pure ink in water and found to be:

$${\mathrm{\mu }}_a(665nm)=6.721\times [ink]c{m}^{-1}$$ (1)

where [ink] is microliters of ink per milliliter of water. The scattering coefficient is related to Intralipids by:

$${\mathrm{\mu }}_s\prime =13.282\times (\%IL)c{m}^{-1}$$ (2)

where (%IL) is the Intralipid concentration used (%IL = 20%*x/(x+y), where x is the volume (in l) of 20% stock intralipid, y is the water volume (in l)).

2.3. Optical properties determination

Scatter light fluence rate (ϕ_s) is determined by subtracting primary light fluence rate (ϕ_p), which is measured with the anal light applicator in air, from total light fluence rate (ϕ_t), which is measured with the anal light applicator inserted into the liquid tissue simulating phantom. The relationship between ϕ_t, ϕ_p and ϕ_s is given in equation (3).

$${\varphi }_t={\varphi }_p+{\varphi }_s$$ (3)

The scatter light fluence rate is then fitted using the following equations for parameter b₁, b₂ and b₃:

$${\varphi }_s/S=\frac{b_1}{r}{e}^{-b_2\cdot r}+b_3$$ (4)

where S is the source power; r is the source-detector separation $\left(\sqrt{x^2+h^2}\right)$; x is the displacement of the detector along the catheter; and h is the perpendicular distance between the two catheters.

Finally, the optical properties of the phantoms can be determined from b₁, b₂ and b₃ values using Eqs. (5) – (7) as shown below.

$$b_1=0.204{\mathrm{\mu }}_s^{\prime 0.46}$$ (5)

$$b_2=1.53{\mathrm{\mu }}_a^{\prime 0.5}{\mathrm{\mu }}_s^{\prime 0.2}+0.6$$ (6)

$$b_3=0.1Rd$$ (7)

where $R_d=0.4843a^{\prime }\left(1+e^{-4.428\sqrt{1-a^{\prime }}}\right)\left(e^{-2.65\sqrt{1-a^{\prime }}}\right);a^{\prime }=\frac{{\mathrm{\mu }}_s^{\prime }}{{\mathrm{\mu }}_s^{\prime }+{\mathrm{\mu }}_a}$. (12)

2.4. Fluorescence data analysis

The raw fluorescence spectra collected are analyzed using a SVD fitting algorithm (13, 14) to separate the contribution of PpIX from other basis components. This algorithm requires the basis spectra of the known components that comprise the measured fluorescence emission spectrum. The first basis component is the autofluorescence spectrum, created by recording the spectrum of 405nm laser from a non-fluorescing scattering solution Intralipid at 1% concentration that passes through the optical filter. The second basis component is the fluorescence spectrum of PpIX, measured at a concentration of 9 mgkg⁻¹. A 40-term Fourier series is included in the SVD algorithm as the third basis component to account for any unknown spectroscopic components, e.g. ambient room light, that contaminates the measured spectra. The Fourier components are given a much lower weight in the fitting routine than that of the excitation source and PpIX components to restrict their application to the unknown components of the spectrum that cannot be fit by combinations of these known components. In the cases presented here, the basis spectra of the first two components adequately account for the measured fluorescence, and the Fourier components constitute only a minor contribution to the fit. Spectra of the all basis components and an example of the SVD fit to one fluorescence spectrum measured from liquid phantom are shown in Figure 2.

Figure 2.

Example of a raw fluorescence spectrum measured from liquid phantom and its SVD fit using the autofluorescence, PpIX and Fourier basis spectral components.

3. RESULTS AND DISCUSSION

The SVD fitting algorithm reduces the measured PpIX fluorescence spectrum to a set of unitless SVD amplitudes. Figure 3 shows the relation between different absolute PpIX concentrations and SVD amplitudes obtained from the fittings. This result was established using a series of tissue simulating phantoms with increasing PpIX concentration (0.1 mg kg⁻¹ to 60 mg kg⁻¹) and optical properties μ_a = 0.3 cm⁻¹ and μ_s’ = 10 cm⁻¹. Fluorescence spectra measured are shown in Figure 4. PpIX concentration is found to be 0.95×SVD +1.64 mg kg⁻¹ and the minimum detectable level of PpIX concentration of the instrument is 0.5 mg kg⁻¹.

Figure 3.

PpIX concentration calibration curve.

Figure 4.

Fluorescence spectra of Photofrin at concentrations ranging from 0.1 mg kg⁻¹ to 60 mg kg⁻¹ in tissue simulating phantom with μ_a = 0.3 cm⁻¹ and μ_s’ = 10 cm⁻¹.

Besides photosensitizer concentration, changes in background tissue optical properties can alter the measured fluorescence intensity and affect the SVD amplitudes from the fittings. Figure 5 shows the variations in SVD amplitudes of fluorescence spectra measured from 16 liquid phantoms with constant PpIX concentrations of 9 mgkg⁻¹ but different optical properties, μ_a between 0.1–0.9 cm⁻¹ and μ_s’ between 5–40 cm⁻¹. Generally, fluorescence intensity and hence SVD amplitude increases with reduced scattering coefficient but decreases with increasing absorption coefficient. These variations in SVD amplitudes may be mistaken for changes in photosensitizer concentration if the effect of background optical properties is overlooked.

Figure 5.

Fluorescence SVD amplitude for PpIX in tissue-simulating phantoms with different optical properties (μ_a = 0.1–0.9 cm⁻¹ and μ_s’ = 5–40 cm⁻¹) but constant PpIX concentration of 9 mg kg⁻¹.

To account for the effect of background optical properties on the measured fluorescence, we determined a set of correction factors (CF) which adjusts the SVD amplitudes for all optical properties to be the same with the SVD amplitude at the reference optical properties. The reference optical properties used in this study were μ_a = 0.3 cm⁻¹ and μ_s’ = 10 cm⁻¹, for which the PpIX calibration curve was performed. The relationship between CF and optical properties is shown in Figure 6. CF shows two-dimensional optical properties dependence which can be described using an analytical expression,$CF=8.64{\mu }_a^{1.75}{\mu }_s^{\prime -0.38}+0.41$. Figure 7 shows the corrected SVD amplitudes after multiplying CF to the uncorrected SVD amplitudes in Figure 5.

Figure 6.

Correction factors determined experimentally to account for the effect of optical properties on fluorescence SVD amplitudes.

Figure 7.

Corrected PpIX fluorescence SVD amplitudes after optical properties effect was eliminated by multiplying CF to the uncorrected SVD amplitudes.

To test the accuracy of our photosensitizer quantification method, we performed a series of test phantoms with constant PpIX concentration of 9mgkg⁻¹ but with different optical properties. We calculated correction factors based on the measured optical properties and applied these values on the fluorescence SVD amplitudes. The recovered PpIX concentrations, and their errors based on the expected values are shown in Table 1. The maximum error from individual measurements is around 27% in phantom 4. The average measured PpIX concentration is 9.3±1 mgkg⁻¹ and the average error is 9.9±5.8%. This result demonstrated that our method could recover PpIX concentrations from liquid phantoms with different background optical properties with great accuracy, given that the errors from absorption coefficient and reduced scattering coefficient measurements are 8.5% and 9.9%.

Table 1.

Absorption and reduced scattering coefficients, correction factors for optical properties, expected and recovered PpIX concentration and the percentage error of recovered PpIX concentrations for 20 test phantoms.

Phantom	μa (cm−1)	μs’ (cm−1)	Correction factor	[PpIX]expected (mg/kg)	[PpIX]recovered (mg/kg)	Percentage error (%)
1	0.09	5.8	0.470	9	7.1	21.1
2	0.08	9.6	0.456	9	8.4	6.5
3	0.09	17.3	0.452	9	9.9	10.2
4	0.09	33.6	0.443	9	11.4	27.1
5	0.27	5.6	0.860	9	9.9	10.1
6	0.28	10.3	0.772	9	8.5	5.4
7	0.29	18.5	0.735	9	8.8	2.5
8	0.31	35.6	0.692	9	10.	11.4
9	0.44	5.2	1.477	9	9.7	7.3
10	0.42	9.8	1.194	9	9.9	9.8
11	0.46	18.3	1.125	9	9.6	6.9
12	0.48	36.2	0.993	9	9.4	4.2
13	0.63	4.9	2.473	9	9.8	8.4
14	0.65	9.9	2.075	9	9.8	9.2
15	0.68	18.7	1.816	9	10.2	13.6
16	0.65	35.8	1.433	9	10.4	15.7
17	0.81	4.4	3.704	9	8.3	7.7
18	0.81	8.8	2.925	9	8.2	9.2
19	0.85	16.9	2.577	9	8.8	2.4
20	0.83	33.5	2.014	9	8.1	9.9
Average					9.3	9.9
Standard deviation					1.0	5.8

4. CONCLUSION

Accurate determinations of the optical properties of tissue and photosensitizer concentration are important in ensuring successful dosimetry of anal PDT. In this study, we developed a method to measure tissue optical properties and PpIX concentration from a phantom geometry that resembles anal canal. A SVD algorithm is developed to separate PpIX fluorescence from the contribution of other spectral components. An analytical correction factor is determined experimentally to eliminate the effect of background optical properties on the PpIX SVD amplitudes. The method of quantifying in-vivo PpIX concentration for anal PDT was tested using a series of liquid tissue-mimicking phantoms with known PpIX concentration and the error was found to be 9.9±5.8%.