Second window near-infrared dosimeter (NIR2D) system for radiation dosimetry

Abstract

Fiber-coupled scintillation dosimeters are a cost-effective alternative to the conventional ion chambers in radiation dosimetry. However, stem effects from optical fibers such as Cerenkov radiation incur significant errors in the readout signal. Here we introduce a second near-infrared window dosimeter, dubbed as NIR2D, that can potentially be used as real-time radiation detector for clinical megavoltage beams. Lanthanide-based rare-earth NaYF₄ nano-phosphors doped with both erbium and cerium elements were synthesized, and a compact 3D printed reader device integrated with a photodetector and data acquisition system was designed. The performance of the NIR2D was tested using a pre-clinical orthovoltage radiation source and a clinical megavoltage radiation source. The system was tested for dose linearity (100, 200, 600 MU), dose rate dependency (100, 200, 400, 600 MU min⁻¹), and energy dependency (6, 10, 15 MV). Test results with the clinical linear accelerator demonstrated excellent dose linearity and dose rate independency when exposed to 6 MV linac beams—both data follows a linear trendline with R² > 0.99. On the other hand, the NIR2D was energy dependent, where the readout dropped by 9% between 6 and 15 MV. For stem effects, we observed a finite Cerenkov contribution of 1%–3% when exposed between 100–600 MU min⁻¹ (6 MV) and 3%–6% when exposed between 5–15 MV (600 MU min⁻¹). While the stem effects were still observable, we expect that enhancing the current optical setup will simultaneously improve the scintillation signal and reduce the stem effects.

1. Introduction

Scintillation dosimetry is the quantification of radiation dose from clinical radiotherapy systems using organic/inorganic materials that are sensitive to ionizing radiation. Although the concept is not novel, certain negative artifacts prevented scintillators from being actively commercialized as dosimeters. The primary example is the stem effect or emission signals in the visible spectrum that originate from components other than the scintillator. For a 10 MV x-ray beam with a field size of 10 × 10 cm², the stem effect was reported to be between 1.9%–2.6% (Beddar et al 1992a, 1992b).

The stem effect for scintillation dosimetry is largely split into two components: Cerenkov radiation and fluorescence emission. Cerenkov radiation is an emission of visible light as charged particles passes through a dielectric medium at speeds faster than the phase velocity of that medium—it is known to be the main component of the stem effect in optical fibers (Lambert et al 2009, Jang et al 2013). Another component of the stem effect is fluorescence, or light emission due to fluors in the optical fiber in response to blue-violet excitation (Beaulieu et al 2013, Beaulieu and Beddar 2016). Recently, various advancements in data collection and processing techniques effectively reduced these two stem effects, thus enabling common materials such as plastic scintillators to be used as commercial dosimeters (Beddar 2006, 2015).

Among various strategies to minimize the stem effects, there are three representative approaches (De Boer et al 1993). The first strategy is to use optical filters that can filter out the ultraviolet wavelength range where Cerenkov emission is the largest. This approach was evaluated to be inefficient since a large portion of the scintillation signal was filtered out while the Cerenkov signal was still significantly large. The second approach uses a pair of optical fibers with identical lengths, one with a scintillator attached and the other left bare. This allows simultaneous acquisition and correction of Cerenkov signals. This third strategy uses a single fiber/scintillator assembly, but the signals are detected from two different color channels each connected to a separate photodetector. The two outputs are dynamically corrected by chromatically comparing the blue and green readout signals. The third technique is known to yield the most reliable results and was successfully implemented as plastic scintillation detectors (PSDs) for commercialization (Frelin et al 2005, Carrasco et al 2015, Ishikawa et al 2015, Mancosu et al 2017, Galavis et al 2019). However, the dual-channel method involves two photodetectors to interpret the raw data, potentially inducing an added source of error. In addition to the aforementioned spectral analyses techniques, other novel approaches include timing the linac beam to avoid the stem signal while detecting signals from scintillators with long-decay constants (Clift et al 2002, Archer et al 2017) or using air-core light guides to negate Cerenkov signals (Lambert et al 2008, 2010, Naseri et al 2010) have been suggested. However, these methods were difficult to replicate and implement in the clinic. As such, a simple single scintillator, optical fiber, and photodetector configuration is still an attractive solution.

In this paper, we introduce an application of rare-earth doped nano-phosphors (RENPs), which has emission peak specifically in the second near-infrared (NIR-II) window, as dosimeters for megavoltage clinical radiation sources. The use of infrared materials were previously explored as radiation detectors, including materials that scintillate at the standard near-infrared spectrum (650–1350 nm) to reduce Cerenkov signals (Takada 1998, Veronese et al 2014, 2017) or materials with broad infrared spectrum (1500–2500 nm) to detect radioluminescence from low energy clinical sources (Correia et al 2013) and nuclear reactors (Shikama et al 2006). In contrast, we suggest the use of modified RENPs that operate primarily between 1500–1700 nm. This emission range aligns closely to the sensitivity range of typical photodetectors (900–1700 nm) (Kindereit et al 2012, Vollmer et al 2015) while effectively minimizing the stem effect. As proof-of-concept, we evaluated the RENP sensor coupled to an in-house developed detector system using a pre-clinical irradiator for low energy response and a clinical megavoltage source for dose linearity, dose dependency, and energy dependency.

2. Materials and methods

2.1. Second window near-infrared dosimeter (NIR2D) system components

2.1.1. Down-conversion RENP

A series of lanthanide-based down-conversion nano-phosphors, NaYbF₄ doped with Er and Ce elements, are synthesized using a solvothermal decomposition method. The synthesis steps are modified from previously published methods (Naczynski et al 2013, Volotskova et al 2015, Zhong et al 2017, Dai et al 2019) and are detailed in supplementary materials (available online at stacks.iop.org/PMB/65/175013/mmedia). The resulting composition was NaYbF₄: Er, Ce at the core with NaYF₄ as the enveloping shell. These RENPs convert ionizing radiation into low-energy photons, which can in turn be measured with near-infrared photodetectors.

The phosphors were filled in a hollow glass tube with an inner diameter of 2.0 mm and length of 5.0 mm. One end of the glass tube has a semi-spherical shape, and the other end is open. The open end was sealed with epoxy resin after filling the tube with the RENPs. The glass tube was inserted in a 3D printed blackout cap, which was then attached to a 5 m long glass clad silica multimode optical fiber with low hydroxyl content (Thorlabs). Low-OH fibers are desirable for our experiment due to high transmission in the near infrared spectrum.

2.1.2. Characterization of RENPs

The elemental analyses of RENPs were performed using an inductively coupled plasma mass spectrometer. ICP standards for each element (Yttrium, Ytterbium, Erbium, Cerium, 1000 μg ml⁻¹) were obtained from Inorganic Ventures. The RENPs were characterized using a transmission electron microscope (TEM) operated at 200 kV. TEM samples were deposited and dried on 400-mesh Formvar-backed copper grids with ultrathin carbon support film, followed by plasma cleaning. NIH ImageJ (Version VI.46) and Gatan Digital Micrograph (Version 2.32) were used for TEM image processing, analysis, and measurement.

2.2. NIR2D reader

A portable reader device was designed to couple with the NIR2D. The reader device is comprised of 3D printed assembly parts, InGaAs amplified photodetector with power adapter (PDF10C, Thorlabs), focusing optics (F230SMA-C, AD11F, SM1L05-P5, and LA4052-C, Thorlabs), long-pass filter (FEL1400, Thorlabs), data acquisition (DAQ) board with 12-bit analog-to-digital converter (USB-204, Measurement Computing), and a 40 × 40 mm² cooling fan (NF-A4xlO-5V, Noctua).

Differential voltage is measured by connecting the positive and negative output from the InGaAs photodetector to two analog inputs in the DAQ board. To keep the photodetector’s temperature close to that of the patient room, which is regulated between 21 °C–23 °C, a 22 × 22 mm² heat sink (HS-BK-10, Gadgeter) is attached to the back of the photodetector and air is blown through the reader device using the cooling fan. This configuration is necessary since the photodetector is sensitive to temperature changes and outputs reliable signals when operated within its calibrated conditions (23 °C ± 5 °C).

The entire experimental apparatus is portable due to the compact form factor, with an exterior dimension of 180 × 140 × 90 mm³ and has a SMA optomechanical port on the side wall to connect with the optical fiber. A schematic diagram and photograph of the assembled device is shown in figure 1.

Figure 1.

(a) Schematic diagram and (b) photograph of the NIR2D assembly.

2.3. NIR2D evaluation

2.3.1. Spectral analysis with non-ionizing radiation source

Spectral response of the NIR2D was first evaluated with non-ionizing radiation source. This is because the luminescence emission spectrum of the RENP upon x-ray irradiation is expected to be similar to that when using a conventional laser excitation source. The yttrium and ytterbium in the host of the RENPs serve as x-ray absorbers which transfer energy to the luminescent dopant Er, thereby resulting in the emission corresponding to the characteristic Er peak at 1550 nm observed from the ⁴I_13/2 → ⁴I_15/2 transition. The fluorescence spectra of RENPs were acquired using a Horiba FluoroLog-3 Fluorimeter with an InGaAs IR detector (H10330B-75, 950–1700 nm).

2.3.2. Orthovoltage source

The NIR2D response to an orthovoltage source was evaluated using a 320 kVp cabinet x-ray system (X-Rad 320, Precision x-ray). The luminescence spectra of the phosphors upon x-ray irradiation were collected either in the regular UV–vis-NIR range (190–1100 nm) with a Ampex spectrometer, or in the NIR-II range (900–1700 nm) using an Acton spectrometer (SP2150, Princeton Instruments) coupled to an InGaAs camera (NIRvana 640, Princeton Instruments).

The accumulated dose and dose rate dependency of the NIR2D were measured using our reader device. For the accumulated dose test, the NIR2D was exposed for 1.5 min at a dose rate of 3.02 Gy min⁻¹. The dose rate of the x-ray system was controlled by changing the tube current from 5.0–12.5 mA, where 7.5 mA corresponds to a dose rate of 3.02 Gy min⁻¹ and 12.5 mA corresponds to 5.04 Gy min⁻¹.

2.3.3. Clinical megavoltage source

The NIR2D was tested with a Varian TrueBeam linear accelerator. The experimental procedure was designed so that it would be similar to performing dosimetry using a thimble ion chamber setup (figure 2(a)). Solid water phantoms (Solid Water®, CIRS) with horizontal dimensions of 30 × 30 cm² were placed on the patient couch. A pair of 5 cm thick solid waters were added on the bottom for backscatter layer, and 0.5 cm, 1.5 cm, or 2.0 cm thick solid water was added on the top to achieve maximum percentage depth dose (PDD) for 6 MV, 10 MV, or 15 MV, respectively. The thickness of the top solid water layer accounts for a 1 cm buildup thickness in the solid water with an ion chamber cavity. The fiber optic end with the sealed RENPs was inserted in the ion chamber cavity and aligned to the light field cross-hair center, and the couch was positioned at 100 cm source-to-surface distance (SSD). To maximize the Cerenkov emission, the linac beam size was set at the maximum size of 40 × 40 cm².

Figure 2.

Photograph of (a) scintillator and optical fiber assembly inserted in solid water and (b) reader device with shielding.

The in-house developed reader device was connected to the open end of the optical fiber, and data was transferred to a laptop in the control room using a 10 m USB 3.0 active extension cable (SIIG). Data was continuously acquired from start to finish for each dose delivery at a rate of 10 Hz, which corresponds to a maximum output frequency response with the current InGaAs photodetector. To minimize stray radiation from affecting the reader device, the device was placed on a cart 5 m away from the linac isocenter. Lead blocks (2.5 cm thickness) were covered around the reader device and a lead plate (0.5 cm) on top for added shielding (figure 2(b)). Radiation dose was delivered to the NIR2D using the parameters shown in table 1 to test for dose linearity, dose rate constancy, and energy constancy. For each parameter, the dose was delivered twice.

Table 1.

Linac dose delivery parameters for NIR2D.

Test type	Energy (MV)	Dose rate (MU min−1)	Delivered dose (MU)
Dose linearity	6	600	100
			200
			200
			600
Dose rate constancy	6	100	100
		200
		400
		600
Energy constancy	6	600	100
	10
	15

3. Results

3.1. Characterization of RENPs

The RENPs were synthesized as core/shell structures, which is consisted of an Er/Ce doped NaYbF₄ core surrounded by a NaYF₄ shell. The validity of the RENPs were visually confirmed with TEM micrographs (figure 3). It can be observed that the Er/Ce-doped NaYbF₄ core has uniform spherical morphology with narrow size distribution (~17 nm in diameter) and that 5 nm thick NaYF4 passive shells that surround the cores. These NaYF4 shells create a gap between lanthanide ions and surface quenchers.

Figure 3.

(a) TEM image of NaYbF₄: Er, Ce (core)/NaYF4 (shell). (b) Magnified image of a single RENP.

3.2. Non-ionizing fluorescence emission spectrum

The NIR2D was first investigated with a non-ionizing radiation source. Figure 4(a) shows the emission-excitation contour map of the NIR2D using the spectrometer setup. The resonant transfer of excitation energy from the sensitizers (Yb) to activator dopants (Er) in the RENPs resulted in the generation of NIR-II emission at ~1550 nm with 68 nm of full width at half maximum (FWHM), which is unique to the specific rare-earth activator (Er). The photo luminescence emission-excitation map further confirms that the Er-doped RENPs have a substantially large Stokes shift (~580 nm).

Figure 4.

Spectral data of the NIR2D acquired using a spectrometer. The data represents (a) emission-excitation contour and (b) emission spectra with a 980 nm laser source. The arrow represents the FWHM of the peak.

The fluorescence emission spectrum in response to a 980 nm excitation (figure 4(b)) was also investigated with an InGaAs IR detector. While there is a smaller emission peak between 950–1050 nm, a maximum emission peak can be clearly observed at 1550 nm.

3.3. Response to orthovoltage x-ray

The characteristic NIR-II emission of the Er dopant at 1550 nm was also observed using the pre-clinical x-ray. In figure 5(a), the luminescence emission spectrum of the RENP upon x-ray irradiation is similar to that from the laser source experiment. Moreover, the FWHM is measured to be 65 nm, which is also very close to that from the laser experiment (68 nm). The yttrium and ytterbium in the host of the RENPs serve as x-ray absorbers that transfer energy to the luminescent dopant Er, thereby resulting in an emission that corresponds to the characteristic Er peak at 1550 nm due to the ⁴I_13/2 → ⁴I_15/2 transition. This experiment allowed us to determine that a 1400 nm long-pass filter is an appropriate candidate for the NIR2D system. It should be noted that periodic valleys are observed in a figure. This may be due to alignment issues in the spectrometer setup. Since the spectrometer runs in step mode, the mirrors and dichroic mirror shifts for every step throughout the wavelength scan range. Thus, any mismatch between the optical fiber and the NIR-camera coupled spectrometer may cause such artifact. Despite the periodic valleys, the figure clearly shows that the maximum emission coincides at 1550 nm.

Figure 5.

RENP response to 320 kVp orthovoltage x-ray source. Graphs representing (a) spectral data ranging between 900–1700 nm, with arrows representing the FWHM, (b) dose rate dependency with error bars representing one standard deviation (N = 100), (c) and stability to prolonged radiation dose. Fluctuations in the x-ray tube current and hence the dose rate is assumed to be negligible (<1%).

The NIR2D’s response to the orthovoltage source was also tested using the in-house developed reader device. In the dose rate dependency test, 100 data points were recorded for each current setting of the x-ray source and the tube current was set between 0 mA and 12.5 mA, which corresponds to a dose rate of 0 and 5.04 Gy min⁻¹, respectively. The readout values were subtracted by the average dark signal, or the signal measured when the x-ray tube was off. Results showed a linear response to the tube current, with standard deviation ranging between 2.187 × 10⁻³ V and 3.101 × 10⁻³ V.

For the dose accumulation test (figure 5(c)), the NIR2D was exposed to a 3.02 Gy min⁻¹ beam for ~90 s or a total dose of 4.5 Gy. During the duration of the test, the readout signal was consistent, indicating that the accumulated readout voltage increases linearly with delivered dose. Additional studies will be performed in the future to evaluate the radiation degradation effects for prolonged x-ray exposure (e.g. 1 kGy).

It should be noted that no stem effects were observed using the current setup (Supplemental figure 1). This is likely because the orthovoltage source was not energetic enough to induce Cerenkov signal and the indigenous fluorescence of the low-OH (hydroxyl) silica fibers, which has an emission peak of 450 nm and FWHM of 25 nm (Simiele and Dewerd 2018), was effectively filtered out using the 1400 nm long-pass filter.

3.4. Response to megavoltage x-ray

The megavoltage response of the NIR2D was studied using the Varian TrueBeam, and was tested for dose accumulation, dose rate response, and energy response. Reference background signal was acquired by removing the optical fiber and blocking the fiber port without delivering any dose. It should be noted that there were no significant changes in the background signal when the linac delivered dose at 6 MV and 600 MU min⁻¹ (supplementary figure S2).

In the dose accumulation test, the readout values were integrated into a single value at the end of each exposure. As can be observed in the raw data (figure 6(a)), the NIR2D reader shows good repeatability during multiple exposures. Moreover, the accumulated voltage plot (figure 6(b)) clearly demonstrates a linear dose response with negligible offset from the axis of origin. The two trendlines represent two separate exposures, and since the data acquisition rate was set to 10 Hz, 600 data points were detected for both trials during the exposure time of 60 s.

Figure 6.

(a) Raw data of differential voltage values collected at 6 MV and 600 MU min⁻¹ and results for (b) dose linearity, (c) dose rate dependency, and (d) energy dependency. All the results were subtracted by the baseline signal or readout values when the linac beam was off.

For dose rate dependency test, the readout values were averaged for each dose rate setting (N = 80), which were also repeated twice for each parameter, and the energy and prescribed dose were fixed at 6 MV and 100 MU. Figure 6(c) shows that the average readout voltage increases linearly with dose rate, and the standard deviation ranges between 9.993 × 10⁻³ V and 1.276 × 10⁻² V. The results suggest that the readout signal will accumulate proportionally to the delivered dose while being independent of the dose rate. For Cerenkov measurements, the same experimental parameters were repeated by removing the scintillator-filled glass tube from the optical fiber and blackout cap assembly. Results show that the Cerenkov signal increases with dose rate, although the slope was substantially smaller than that of the NIR2D by 97.22%. The standard deviation ranged between 8.758 × 10⁻³ V and 1.347 × 10⁻² V.

Lastly, the NIR2D response to beam energy was evaluated. For this experiment, the dose rate and dose settings were fixed at 600 MU min⁻¹ and 100 MU, respectively. Results (figure 6(d)) show a finite dependency between different energy settings (6, 10, and 15 MV), and the readout signal drops by ~9% between 6 and 15 MV. Also, the Cerenkov signal increases with energy, although the absolute readout value is close to background level (between 0 mV and 44 mV). The range of the standard deviations were 9.996 × 10⁻³ V–1.212 × 10⁻² V and 7.878 × 10⁻³ V–1.169 × 10⁻² V for the NIR2D and Cerenkov radiation readouts, respectively.

Using the data from figures 6(c) and (d), the Cerenkov contribution was extracted by normalizing the Cerenkov signal with the scintillation signal. The Cerenkov contribution as a function of dose rate and beam energy are shown in figures 7(a) and (b). Since the data is a function of both Cerenkov and scintillation readout signals, propagation of error was used to calculate the error bars:

$$\sigma \left(\frac{\overline{C}}{\overline{S}}\right)=\left|\frac{\overline{C}}{\overline{S}}\right|\sqrt{{\left(\frac{\sigma \left(C\right)}{\overline{C}}\right)}^2+{\left(\frac{\sigma \left(S\right)}{\overline{S}}\right)}^2}$$ (1)

where $\overline{C}/\overline{S}$ is the averaged Cerenkov signal normalized by the average scintillation signal, σ (C) is one standard deviation of the Cerenkov signal, and σ (S) is one standard deviation of the scintillation signal.

Figure 7.

Cerenkov contribution for NIR2D tested at varying (a) dose rates and (b) beam energy.

4. Discussion

We demonstrated the performance of the NIR2D using a pre-clinical x-ray source and a clinical linac. Although stem effects were observed in the linac experiments, operating in the second window near-infrared spectrum shows potential in negating the need of multiple photodetectors and optical filters. Some relevant discussions about the stem effects and strategies to reduce its contribution are discussed below.

4.1. Stem effect contribution

As indicated in the previous section, the raw Cerenkov signals were indeed close to background level. However, since the scintillation signal was also relatively dim, despite having excellent linear response to accumulated dose and dose rate, the Cerenkov contribution was 1%–3% when exposed to 6 MV beam (figure 7(a)). While the value is similar to previously reported stem effects (1.9%–2.6% at 10 MV, 10 × 10 cm² beam), the results are promising since the packaging of the NIR2D was not optimized.

One of the main issues was the packaging material itself—the RENPs were filled in a hollow glass tube in free powder form and was capped with epoxy resin. The powder freely moved within the glass encapsulant, leading to inconsistencies and reduction of light emission in the optical axis as the NIR2D was placed horizontally in the solid water. Moreover, a thick glass layer separated the RENP and the optical fiber. Since the glass wall was significantly thick (~1.0 mm) and long (5.0 mm), and since the glass tube section was positioned directly under the linac beam, this may not only reduce the scintillation signal but also potentially induce additional Cerenkov emission. Once the bulk synthesis procedures are optimized, several strategies will be implemented in the future to boost the scintillation signal while reducing the stem effect at the same time.

One potential solution in optimizing the phosphor packaging is to replace the glass container with a thin polymer film, similar to those used for fabricating x-ray plates. By uniformly and densely packing the material in a tighter space, and by coupling the embedded scintillator to the optical fiber with optical cement, both the overall stem signal and the uncertainty due to inconsistent shifting of the nanoparticles will be minimized. In addition, the packaging material may be coated with reflective paint to redirect stray visible photons towards the optical fiber (Archer et al 2020).

For the stem effects in the optical fiber, the Cerenkov and fluorescence signals may be reduced by optimizing the optical filter in the dose reader. Since the base of the NIR2D’s emission curve starts at 1450 nm, the current 1400 nm filter could be replaced with a precision-grade 1450 nm long-pass filter. This will enhance the scintillation signal while simultaneously filtering out the stem signal—the Cerenkov component that depreciates by ~λ⁻³ (Beaulieu and Beddar 2016) and the fluorescence component that has a peak emission of 450 nm and a narrow FWHM of 25 nm—from the optical fiber. Although the fiber fluorescence emission was not considered in this study due to the use of the 1400 nm long-pass filter, further studies will be performed based on previously reported techniques (Lee et al 2007) to clearly identify the stem effect in detail. Despite the observable signals, it should be noted that the stem effect was maximized in this experiment by using the largest possible beam field size of 40 × 40 cm² at 100 SSD and that the error bars of the stem effects were close to the background signal.

4.2. Energy dependency

Although the NIR2D demonstrated excellent dose linearity and dose rate constancy, some dependence to energy was observed (figure 7(b)). Between 6 MV and 15 MV, the NIR2D signal dropped by ~9%. Due to this energy dependency, the variation of the Cerenkov contribution was also relatively large (3%–6%).

One potential source of error for the energy dependency is electron contamination, which is due to secondary electrons/positrons originating from the linac treatment head, air, etc. The secondary electron effects were less significant for the dose accumulation and dose rate constancy tests since the experimental parameters such as linac energy, solid water thicknesses, and NIR2D position were kept constant. However, in the energy dependency test, the build-up thicknesses of the solid water were varied (0.5 cm, 1.5 cm, and 2.0 cm) to achieve maximum dose (d-max). In future studies, the NIR2D performance will also be evaluated at depth well beyond the maximum range of electron contamination (e.g. 10 cm of solid water).

4.3. Reducing sources of uncertainty

As observed in figure 7, the uncertainty of the Cerenkov to scintillation ratio was significantly large. This is because the error bars accounted for errors in both the Cerenkov and scintillation signals (equation (1)). One potential source of uncertainty, including the uncertainties observed in figures 6(c) and (d), maybe due to the 10 m USB long active extension cord that was used to connect the NIR2D reader in the patient room to the computer in the control room. A comparison of dark signals collected using an active 10 m extension cord vs. no extension cord shows 50% increase in standard deviation (5.5 mV to 8.2 mV) as well as 20% increase in average readout values (supplementary figure 3). Although the absolute value is small, the stem effect readout values are particularly sensitive to noise since they are close to background level. While this is a rudimentary evaluation, it can be suggested that embedding a single board computer directly in the NIR2D reader device instead of using an extension cable may reduce the uncertainty of the reader device.

On the sensor side, the uncertainty can be further reduced if the scintillation signal from the nano-phosphors is enhanced. The coefficient of variance $(COV=\frac{\sigma (Readout)}{avg(Readout)}\cdot 100)$ of the scintillation signal ranges between 5.1%–42.0% in the dose rate dependency test (figure 6(c)) and 5.1%–6.9% in the energy dependency test (figure 6(d)). While the COV is relatively small in the latter experiment, the value is substantially large when the NIR2D is exposed to 100 MU min⁻¹ in the dose rate dependency experiment (42.0%). This is because the readout voltage itself was low and the fluctuations in the voltage readout was relatively large compared to the averaged value. For the COV to drop well below 10%, the signal response from 100 MU min⁻¹ must be at least close to that of 400 MU min⁻¹ (COV = 7.7%). This means that the emission of the scintillator should be ~2.8 times brighter than the current configuration. Further studies to boost the scintillator signal, as discussed in section 4.1, is expected to reduce the uncertainty and improve the overall accuracy of the system.

5. Conclusion

In summary, the use of rare-earth doped nano-phosphors as NIR2D dosimeter was demonstrated in this study. Operating in the second window near-infrared spectrum shows promise that the system may only require a single InGaAs photodetector coupled to an optical cut-off filter and focusing lens assembly. The performance of the dosimeter was evaluated using a pre-clinical orthovoltage source and a clinical linear accelerator.

The NIR2D was evaluated first with a pre-clinical orthovoltage system, where results confirmed an emission peak of 1550 nm and a linear response to accumulated dose and dose rate. For clinical megavoltage source, the NIR2D showed consistent readout voltage under multiple exposures (6 MV and 600 MU min⁻¹) and had a linear response to accumulated dose (100–600 MU) and dose rate (100–600 MU min⁻¹). Although the Cerenkov contribution was similar to the previously reported spectral removal technique using PSDs (1.9%–2.6%), further improvements in future iterations will enhance the scintillation signal while simultaneously reducing the stem effects. Potential strategies in improving the system include replacing the container with a thin polymer film to minimize the phosphor/optical fiber distance, using optical adhesives to maximize optical coupling, optimizing the phosphor packing density to maximize the scintillation intensity, and replacing the reader system with a precision-grade 1450 nm long-pass filter.

While further investigation will be performed, we successfully demonstrated here that the NIR2D may be a suitable candidate for performing real-time relative dosimetry under various applications, including dosimetry under small radiation fields.