Validation of multispectral singlet oxygen luminescence dosimetry (MSOLD) for photofrin-mediated photodynamic therapy

Abstract

Accurate dosimetry is crucial for the ongoing development and clinical study of photodynamic therapy (PDT). Current dosimetry standards range from less accurate methods involving measurement of only light fluence and photosensitizer concentration during treatment, to significantly improved methods such as singlet oxygen explicit dosimetry (SOED), a macroscopic model that includes an additional important parameter in its dosimetric calculations: ground-state oxygen concentration ([³O₂]). However, neither of these models is a method of direct dosimetry. Multispectral singlet oxygen luminescence dosimetry (MSOLD) shows promise in this regard but requires significant improvement in signal quality and remains to be validated in a clinical setting. In this study, we validate a linearly increasing MSOLD signal with an InGaAs photodiode detector for increasing concentration (0 mg/kg to 200 mg/kg) in tissue-simulating phantoms containing photofrin, calculating a calibration curve based on 1270 nm peak-intensity signal and area under the curve for background-subtracted singlet oxygen emission. Additionally, we validate MSOLD against the current clinical dosimetry standard, SOED, through simultaneous measurement of SOED parameters and MSOLD signal for varying concentrations (50 μM – 300 μM). Finally, we investigate the effects of using very high gain amplification on InGaAs photodiode detectors to amplify the MSOLD signal for use in clinical models. We show that a calibration curve relating photosensitizer concentration (PS) and MSOLD signal can be established. Additionally, we demonstrate good correlation between MSOLD signal and SOED-calculated [¹O₂]_rx. However, we show that when using high amplification on InGaAs photodiodes for long illumination times, the inherent instability in these detectors becomes apparent.

1. INTRODUCTION

Multispectral singlet oxygen luminescence dosimetry has shown promise in recent studies for use as a possible in-vivo, direct method of photodynamic therapy (PDT) dosimetry [1]. MSOLD involves direct measurement of fluorescence emission during treatment for a containing the singlet oxygen reactive species (¹O₂) emission peak (Figure 1). This spectrum is typically measured from approximately 1200 nm to 1300 nm to capture both the ¹O₂ emission peak (1270 nm) and enough of the background photosensitizer (PS) and contamination signals to perform spectral discrimination. Spectral analysis is then used to directly quantify singlet oxygen concentration from the weak, short-lived, characteristic 1270nm fluorescence emission over the expected PS spectrum (figure 1). Recent studies have employed the use of InGaAs detectors over traditional silicon-based detectors to more effectively capture this NIR spectral signal [2–4] due the weak nature of the ¹O₂ emission. However, the use of this detector can present challenges at higher gain settings (which may be pursued to offset the weak ¹O₂ emission) due to the inherently higher leakage currents compared to their silicon-based counterparts

Figure 1.

Photofrin spectrometer signal obtained for the spectral signal from 1000 nm to 1600 nm using an InGaAs spectrometer (Avante, AvaSpec NIR256, Apeldoorn, NL).

In this study, we validate the possibility of calculating a calibration curve that relates photofrin concentration to MSOLD signal through both 1270nm peak-intensity signal as well as area under the curve signal. Additionally, we validated MSOLD in photofrin-mediated PDT in liquid phantoms against the current clinical explicit dosimetry standard: SOED [5–7]. SOED-calculated reactive oxygen species and MSOLD emission signal were analyzed for correlation. We also investigated the effects that an InGaAs detector can have on this relationship when used at very gains.

2. METHODS

2.1. MSOLD spectral measurement setup

Tissue-simulating liquid phantoms in disposable cuvettes were prepared in water at photofrin concentrations of 50 mg/kg, 100 mg/kg, 200 mg/kg, as well as a 0 mg/kg control. A fiber with micro-lens tip (Pioneer Optics Company, Bloomfield, CT, USA) was attached via SMA connector to a 633 nm CW laser array (8 W max output), with which illumination was performed with a 1 cm spot-size. A high-NA SMA-SMA patch cable (Thorlabs, Newton, NJ, USA) was used to collect liquid phantom emission during laser illumination at a fluence rate of 300 mW∙cm^-2. The collected light was passed through coupled collimators to create free-space in which a filter wheel (Thorlabs, Newton, NJ, USA) was located, containing a range of bandpass filters (1200 nm, 1240 nm, 1250 nm, 1270 nm, 1300 nm). Finally, the filtered signal was captured with a mounted InGaAs photodiode (Thorlabs, Newton, NJ, USA) and a custom photodiode amplifier which has increased gain (200 x). We used such a high gain to test increased signal for mouse and clinical models, which have shown resistance to strong signal acquisition to the inherent weakness of the [¹O₂] signal in such mediums. In-house made software was used to quantify signal strength from the detector.

2.2. MSOLD spectral analysis and calibration curve

An exponential fit was applied to the 1200 nm, 1240 nm, and 1300 nm points of each spectrum for quantification of the PS background. Subsequently, a Gaussian fit was established on the PS background-subtracted spectrum to obtain the singlet oxygen emission spectrum. An element-by-element sum of these two fits results in a full, PS + ¹O₂ spectrum fit (Figure 2). The peak intensity of the 1270 nm peak, as well as the area under the curve of the Gaussian fit were used in establishing a calibration curve the relates the known PS concentration to the normalized MSOLD signal.

Figure 2.

(a) MSOLD spectral measurements for varying concentrations and (b )the respective calibration curve established from background-subtracted peak intensity and area under the curve.

2.3. Simultaneous MSOLD/SOED measurement setup

Additional tissue-simulating liquid phantoms were prepared in water at photofrin concentrations of 50 μM, 100 μM, 300 μM, with 0.2% by volume of a lipoprotein colloidal suspension, Intralipid (Fresenius Kabi, Uppsala, Sweden), added. An identical experimental setup to the previously mentioned was used, focusing on the 1270 nm spectral measurement during illumination time. We also must also consider increased photosensitizer autophosphorescence (PSAF) at increasing concentrations causing nonlinear signal behavior from the PS background. Rather than using a multichannel system to monitor the full MSOLD signal for this, full MSOLD “instantaneous” spectra and background signals were obtained immediately prior to pre-, during, and post-PDT conditions. These were defined as the first, middle-most, and last 10 seconds of treatment. This PSAF signal was then fit over time to calculate cumulative MSOLD – PSAF for each concentration. PS and ground-state oxygen ([³O₂]) concentration were monitored for SOED [¹O₂]_rx calculations using a custom contact probe [8] and dissolved oxygen monitor (Oxford Optronix, Abingdon, OX, UK), respectively. PS concentration was measured every 100 seconds during brief interruptions of the laser light, and oxygen partial pressure was measured every 10 seconds. Oxygen partial pressure in mmHg was converted to μM by multiplication of a factor of 1.3 [4]. Fluence rate (mW) was measured with a laser power meter (Coherent, Santa Clara, CA, US). Single-fraction illumination was performed at 120 mW∙cm⁻² (1.1cm spot-size) for 700 s to a total fluence of 84 J∙cm².

2.4. SOED calculations

Ground-state oxygen partial pressure measurements were taken every 10 seconds during laser illumination. Phantom PS concentration measurements were taken with the contact probe before and after PDT, as well as every 100 seconds during illumination. A best-fit exponential was calculated for both oxygen and PS concentrations, with which the SOED-calculated reacted singlet oxygen was calculated as [4–6]:

$${\left[1O_2\right]}_{rx}=\xi {\displaystyle {\int }_t^{t+dt}\frac{\left[3O_2\right]\left(\tau \right)}{\left[3O_2\right]\left(\tau \right)+\beta }}C\left(\tau \right)\phi d\tau$$ (1)

Here, [3O₂](τ) is the ground-state oxygen concentration and C(τ) is the PS concentration at time τ. The ξ and β constants are the specific oxygen consumption rate and oxygen-quenching threshold concentration, with values of 10.3 × 10⁻³ cm²∙mW⁻¹∙s⁻¹ and 11.9 μM, respectively [4]. [¹O₂]_rx was calculated for every second of treatment for all treatment conditions.

2.5. Correlation analysis

A linear fit was calculated for cumulative MSOLD counts and SOED-calculated [¹O₂]_rx. Additionally, linear fits were established for instantaneous pre-, during, and post-PDT conditions. These were defined as the first, middle-most, and last 10 seconds of treatment. The time-constrained analysis was performed to measure the effect that detector instability over long treatment times at very high gain settings as well as the PS concentration-dependent autophosphorescence effect on correlation between both methods of [¹O₂]_rx quantification. Pearson correlation coefficients were calculated for all conditions.

3. RESULTS

3.1. MSOLD spectral measurements and calibration curve

The MSOLD spectral signal was measured for a series of tissue-simulating phantoms in water with PS concentrations ranging between 0 mg/kg to 200 mg/kg. A best fit for the background was established using the 1200 nm, 1240nm and 1300 nm spectral points. A gaussian curve was fit to the background-subtracted ¹O₂ emission data for the 1250 to 1300nm spectral points. The resulting MSOLD spectrum and background fit are shown in Figure 2 (a), where a clear increase in both peak 1270 nm and background signal is observed. Figure 2 (b) shows a calibration curve that was established using the background-subtracted 1270 nm peak intensity, as well as the area under the curve of the ¹O₂ gaussian-fit emission feature over the background. A linear increased in normalized MSOLD signal is seen for increasing PS concentration. The spectral integration curve has a slightly higher slope than the background-subtracted 1270 nm peak intensity curve.

3.2. SOED measurements and calculations

Tissue simulating phantoms of PS in water were prepared at varying concentrations with 0.2% by volume of Intralipid scatterer to simultaneously measure 1270nm peak-intensity, [³O₂] and PS concentration [C] while illuminated with uniform 633 nm light from a CW laser. PS concentration was measured every 100 seconds using the previously mentioned contact probe [8] during brief interruptions of the laser light. The mean of three PS concentration measurements were used for the SOED model calculations, while [³O₂] was measured every 10 seconds with the oxygen monitor. Equation (1) was used to calculate [1O₂]_rx, with ξ and β being the specific oxygen consumption rate and oxygen-quenching threshold concentration parameters mentioned previously. Figure 3 (a) shows normalized measured oxygen concentrations during light on for all PS concentrations. The solid line is a best-fit exponential function. Figure 3 (b) shows mean PS concentration as hollow symbols. The solid line is a best-fit exponential, while the dashed line is the SOED-calculated concentration based on fluence rate and [³O₂] consumption.

Figure 3.

(a) Normalized diffuse oxygen monitor measurements (hollow symbols) for all PS concentrations in tissue-simulating phantoms with best-fit exponential curves (solid lines). Initial [³O₂] were 187.5 μM, 175.75 μM, and 187.6 μM for the 50 μM, 100 μM, and 300 μM tissue-simulating phantoms, respectively. (b) Normalized mean PS concentrations measured with custom contact probe [8] (hollow symbols), with best-fit exponential curves (solid lines) and expected SOED-calculated PS concentrations from fluence rate and ground-state oxygen concentration. Initial [PS] were 51.6 μM, 101.9 μM, and 302.81 μM for the 50 μM, 100 μM, and 300 μM tissue-simulating phantoms, respectively.

3.3. Correlation analysis

Pearson correlation coefficients were calculated for four conditions. The first coefficient was cumulative MSOLD signal and SOED-calculated cumulative [1O₂]_rx for all three PS concentration (Figure 4 (a)), resulting in an r² = 0.9207. Additionally, we calculated correlations for three time periods during illumination: pre, during, and post-PDT (Figure 4 (b)) “instantaneous” signals. These time periods were defined as the first, middle-most, and final 10 seconds of illumination, respectively. Calculated correlation coefficients were r² = 0.86311, r² = 0.98868, and r² = 0.99462 for pre, during, and post-PDT time periods, respectively. We observed increasing correlation, but a decreasing slope for later time periods means there is no direct correlation of the MSOLD signal to the singlet oxygen because the raw MSOLD singal includes a strong background (see Fig. 2). The pre-, during-, and post-PDT full MSOLD spectra measurements are processed by subtracting a virtual background to eliminate the intercepts and the time dependent slope (m) in Fig. 4 to extrapolate the MSOLD component that were directly proportional to singlet oxygen (Fig. 5).

Figure 4.

(a) Cumulative MSOLD signal versus SOED-calculated [¹O₂]_rx for all liquid phantom PS concentrations. (b) Pre, during, and post-PDT MSOLD signal versus SOED-calculated [¹O₂]_rx for all liquid phantom PS concentrations. Pre, during, and post-PDT conditions were defined as the first, middle-most, and final 10 seconds of treatment. Best-fit linear curves (dashed lines), slopes (m) and Pearson correlation coefficients accompany all considered conditions.

Figure 5.

(a) Cumulative MSOLD signal versus SOED-calculated [¹O₂]_rx for all liquid phantom PS concentrations after accounting for concentration-dependent virtual background. (b) Pre, during, and post-PDT MSOLD signal versus SOED-calculated [¹O₂]_rx for all liquid phantom PS concentrations after accounting for concentration-dependent background.

4. DISCUSSION

4.1. MSOLD spectral measurements and calibration curve

MSOLD signal shows a linearly increasing signal for increasing concentration for both 1270nm peak-intensity and area under the curve, allowing the calculation of a calibration curve relating MSOLD signal and photofrin concentration in liquid phantoms. Area under the curve shows a slightly higher slope than 1270nm peak-intensity, though the effect is only significant at very high PS concentrations that are unlikely to be observed in a clinical setting. Increasing the amount of spectral sampling points (filters) to provide better, more accurate fits for the background and [¹O₂] signals would likely address this minor discrepancy between the area under the curve signal and the background-subtracted 1270nm peak intensity curve.

4.2. MSOLD/SOED correlation analysis

During the SOED and MSOLD comparison experiments, [³O₂] and PS concentration decreased over illumination time, with PS closely following expected SOED-calculated PS levels, demonstrating an induced PDT effect in the liquid phantoms. The MSOLD signal showed good correlation with the SOED-calculated [¹O₂]_rx for cumulative signal, as well as pre, during, and post-PDT time periods. However, the decreasing slope for the later time periods shows the effect that PS concentration-dependent autophosphorescence and instability of InGaAs photodetector leakage at very high gain settings. In a clinical or pre-clinical setting, where stability and reproducibility are required for typically longer illumination times, care must be taken to take these effects into consideration for accurate dosimetry. This can be done by accounting for background and PSAF, whether by a virtual background subtraction as we did here, or by monitoring of the full MSOLD spectra over the entire treatment with a multichannel system.

5. CONCLUSIONS

MSOLD signal shows a linear increase with PS concentration for both 1270 nm peak-intensity and area under the curve signals, allowing a calibration curve that related these signals with PS concentration to be established. Additionally, MSOLD shows good correlation with SOED-calculated [¹O₂]_rx for cumulative and instantaneous signals using an increased gain signal from an InGaAs detector in tissue-simulating liquid phantoms. However, care must be taken when using this method due to concentration-dependent autophosphorescence from the PS, as well as the increased instability over long treatment times of the signal from the InGaAs detector when used at such high gain settings. A multichannel system that monitors the full MSOLD signal during treatment would be the best approach to compensate for these effects.