Design and commissioning of a Silicon Photomultiplier-based dosimeter for Low Dose Rate (LDR) oncological brachytherapy

Brachytherapy is a radiotherapy procedure performed with radioactive sources implanted into the patient's body, close to the area affected by cancer. This is a reference procedure for the treatment of prostate and gynecologic cancer due to the reduction of the dose released close to organs at risk (e.g., rectum, bladder, colon). For this reason, real-time dose verification and source localisation are essential for an optimal treatment plan. The ORIGIN collaboration aims to achieve this goal through a 16-fibre sensor system, designed to house a small volume of scintillating material in a transparent fibre tip to enable point-like measurements. The selected scintillating materials feature a decay time of about 500 μs and the signal associated with the primary γ-ray interaction results in the emission of a sequence of single photons distributed over time. Therefore, the dosimeter requires a detector with single-photon sensitivity and a system designed to provide dose measurements by photon counting. Uniformity of fibre response, system stability and reproducibility of measurements are key features of the dosimeter. The characterisation of the 16-channel dosimeter system equipped with thermo-electrically cooled Silicon Photomultipliers, carried out in the laboratory using an X-ray cabinet, is discussed and the results are compared with an earlier version equipped with SiPMs operated at room temperature.


Introduction
ORIGIN is a project funded by the European Commission in the framework of the Horizon 2020 Research and Innovation Programme.Its aim is to develop an optical fibre-based in-vivo dosimeter for prostatic and gynecological brachytherapy (BT).Brachytherapy is a form of radiotherapy performed through the temporary or permanent implantation of -emitting seeds in the patient's body, that allows delivering the dose within the cancer volume and avoids undue exposure of nearby healthy organs [1,2].In particular, BT can be divided into High Dose Rate (HDR, > 12 Gy/h) and Low Dose Rate (LDR, from 0.4 Gy/h to 2 Gy/h).This work focuses on LDR-BT, employing a permanent implantation technique of different 125 I sources, called seeds, and it is mainly used to cure prostate cancer [3,4].For LDR-BT (specifications in table 1), the ORIGIN project aims to design a multi-point dosimeter sensitive in a distance range from the source within 5 to 30 mm, where 5 mm corresponds to the spacing between the LDR-BT template grid and 30 mm is linked to the average dimensions of the prostate gland [5].Sensitivity has to be guaranteed for the single seed, depositing a dose between 0.4 Gy/h to 2 Gy/h, and measurements have to provide a 5% statistical precision at 30 mm distance in 0.5 s, where the short time window is connected to the minimum time between two seed implants.

LDR -Specifications
Sensitivity to Dose from 5 mm to 30 mm Spatial Resolution 3 mm @ a distance of 30 mm Dose Rate Range 0.4-2 Gy/h Statistical Precision 5% in 0.5 s @ a distance of 30 mm Results obtained by irradiating with a 125 I source of 0.3 mCi activity the single-fibre prototype of the ORIGIN dosimeter, are reported in the work of M. Martyn et al. [5] and M. Caccia et al. [6].These studies also compare the counting rate as a function of the sensor to source distance with the prediction from the TG43-U1 protocol [7] indicating that there is no need for an "energy correction" -1 -factor to account for the non-water equivalence of the scintillating material.Therefore, the conversion factor from the measured signal to the absorbed dose is a single calibration coefficient.Nevertheless, the characterisation of the single-fibre prototype of the ORIGIN dosimeter showed that the fluctuations in the Dark Count Rate (DCR) for the SiPM in use (S13360-1375 from Hamamatsu) dominate the signals for measurements at distances larger than 25 mm, not compliant with the project specifications.For this reason, a new design was required to improve the dose sensitivity of the dosimeter by working on the two parameters that limit its performance: improving light collection and reducing background.In the new design, the thermo-electrically (TE) cooled SiPM was considered to reduce and keep the DCR (which is highly temperature dependent) constant and, at the same time, special attention was used in the design to improve light collection.
This work reports the full characterisation of the new ORIGIN dosimeter system comparing the results with the prototype based on sensors at room temperature.The comparison is mainly based on two figures of merit: Minimum Detectable Light (MDL) and sensitivity.The first parameter is defined as the minimum Photon Counting Rate (PCR, defined as the measured signal minus the DCR) exceeding 3 times the Poisson fluctuations of the DCR in time windows of 0.5 s.It represents the minimum detectable dose rate above the DCR background and is calculated as [6]: The sensitivity used to qualify the board is based on signals induced in the sensors with an X-ray tube (model 554-81 by Leybold) allowing to operate the system under stable and reproducible conditions.Sensitivity is defined as the minimum change in X-ray current () that induces a variation in PCR greater than three standard deviations () of a single measurement (ΔPCR ≥ 3).In the linear regime, the relation between PCR and X-ray current () is given by PCR =  • . (1. 2) The sensitivity of two systems can be compared using the slope () of the linear trend [6]: the higher the slope, the better the sensitivity.

The ORIGIN dosimeter
The ORIGIN dosimeter consists of 16 polymethylmethacrylate (PMMA) optical fibres, each of them equipped with an inorganic scintillating tip (sensor in the following) made of doped gadolinium oxysulphide (Gd 2 O 2 S:Tb, also known as Gadox) or YVO (1Y 2 O 3 :Eu+4YVO 4 :Eu) [8][9][10][11].The scintillating material is produced in fine grain powder with an average grain size of 4 µm and a density of 7.5 g/cm 3 .The scintillating material is mixed with UV-curing liquid polymer adhesive (NOA61) to obtain the optimal sensor shape [11].The ratio of scintillating material to NOA61 and the shape used in production is a trade-off between light attenuation, optimal light collection, and feasibility of mass production.The scintillating material selected for the final production is Gadox, and an optimisation study is on-going to improve the shape reproducibility, the number of scintillating tips that can be produced in parallel, and the precision of bonding the sensor to the fibre tip.Nevertheless, the sensor used for this characterisation is the YVO from a pre-production, the only available at the time of the measurements.The sensor consists of a hemisphere of 500 µm in diameter glued on the tip of a -2 -PMMA optical fibre.The fibre is then covered by a protective jacket and it has an overall diameter below 1 mm to guarantee the compatibility with standard brachytherapy catheters.
The scintillating light produced by any -ray interacting with the YVO crystal peaks at 619 nm with a characteristic decay time of the order of 500 µs and a measured light yield1 of ≈ 48 photons/keV [11].As a consequence, the interaction of a single -ray (E  = 35.5 keV for 125 I) with the sensor produces a train of single photons (≈ 10 3 ) that needs to be detected.This working condition sets the requirement of using a light detector with single photon sensitivity and good detection efficiency in the wavelength region of interest.The natural choice is the use of Silicon Photo Multiplier (SiPM) operated in counting mode.
SiPMs are solid-state detectors that nowadays can be found in the market with different characteristics, good quality and low cost.For the specific application we were interested in SiPMs with good detection efficiency at 619 nm (which is not the peak sensitivity for standard devices), large gain and low DCR.The last two requirements are of great interest while operating the SiPMs in counting mode, because the first defines the amplification and the readout chain that guarantees the capability to count the single photon-electron occurrence, while the second defines the minimum counting rate distinguishable from the DCR.The SiPM identified for this application is the S13362 series produced by HAMA-MATSU.It has an effective photosensitive area of 1.3 × 1.3 mm 2 and a pixel pitch of 50 µm.This detector provides a Photon Detection Efficiency (PDE) of 25% at the wavelength of interest at the nominal operational voltage and a gain of ≈ 1.8 × 10 6 .Since this generation of SiPMs has demonstrated very low after pulses (cross-talk is not affecting counting measurements), the over-voltage ( ov ) can be further increased to get even higher detection efficiency and larger signals.The only limitation is the increase of the DCR.Indeed, to get this parameter under control, we selected thermo-electrically cooled SiPMs (TE-SiPMs) operated at −20 • C with a typical DCR of 5 kHz at the suggested operating voltage (+3 V).
Even if the small area TE-SiPMs allow to reduce the DCR and guarantee the operation in stable conditions, we have to consider that the current package increases the minimum distance achievable between the optical fibre and the sensitive area of the detector, impacting on the light collection.A series of measurements have been performed to quantify the light collection loss as a function of the distance, due to the optical aperture of the fibre in use.The measurements were performed using two setups: the fibre only and the optical system made of a focuser (F230SMA-A, 4.34 mm focal length) and an aspheric lens produced by ThorLab (assembly shown in figure 1 on the left).The light collection was measured with a monolithic pixel detector (MIMOTERA [12]).The beam profile was estimated by identifying the most intense pixel in the matrix, by defining the iso-lines at 10% of the matrix intensity and by fitting the points with a circle.Figure 2 shows the light spot diameter as a function of the distance measured with both configurations (e.g. with and without the focusing system).The plot compares the results and clearly indicates that the optical system improves the light collection with a minimum at ≈ 3 mm from the sensitive area of the detector, actually close to the distance between the entrance window of the package and the sensor position (3.2 mm).
The 16 SiPMs used in the ORIGIN dosimeter are mounted on a custom PCB board and each optical fibre, with its own focuser, is constrained in front of a SiPM.A precise mechanical system allows the fine adjustment of the distance between the fibre and the SiPM (procedure done at the assembly stage).Once the fine adjustment is made, the repeatability of the fibre position is ensured by the mechanics and has been measured to be better than 0.5%.
1Light yield is defined as the number of photons generated per unit deposited energy.
-3 - The SiPMs are operated with the DT5702 readout board produced by CAEN s.p.a.2The readout board is equipped with two Citiroc-1A ASIC3 [13] and can operate up to 64 SiPMs, even though we only use 16 channels.The system provides both the high voltage and the readout chain (an external power supply is used to drive the Peltier cells of the SiPMs).The signal produced by each SiPM feeds three charge amplifiers.Two amplifiers have tunable gain and tunable shaping times (Slow Shaper), and they are used to measure the light intensity when photons arrive in a short time window (≈ ns).The third amplifier has a fixed (15 ns) shaping time (Fast Shaper), and it is used either to trigger the read out for all the channels or to measure the number of triggers received by each SiPM in a defined time window (Counting Mode).The maximum counting rate achievable by the system is 20 MHz (deterministic signals) and the minimum trigger threshold that can be set at the discriminator is ≈ 1/3 photo-electrons (p.e.) when connected to SiPMs, providing a gain of 10 6 .Both requirements are fitting our needs.

System characterisation
The system characterisation is a two step procedure.The first step characterises the SiPMs (i.e.breakdown voltage and DCR stability), while the second assesses the dosimeter performances in a controlled environment (i.e.channel equalisation and minimum detectable light).The first set of measurements is performed by keeping the SiPMs in darkness, while the second requires that the light produced in the fibre sensor is conveyed to the SiPM.To operate the system in stable and reproducible conditions, the characterisation is performed by irradiating the sensors with an X-ray beam, following the procedure outlined in [14].The X-ray cabinet in use emits photons in the 0-35 kVp range and provides currents up to 1 mA, tunable with a granularity of 0.01 mA.In order to guarantee the proper positioning of the fibre tip, a dedicated "hose" was designed and produced, and the clear fibre conveying the scintillating signal to the SiPM was routed through a cable duct to the detector.The system was operated at room temperature, while the SiPMs were at −20 • C thanks to the Peltier cells.For all measurements, the discriminator threshold was set at 0.5 p.e.

Breakdown voltage and DCR stability assessment
The most relevant characteristics for a SiPM (i.e.PDE, After Pulses and DCR) are voltage and temperature dependent [15].The optimal way to operate more SiPMs in similar and stable conditions is to set the same over-voltage in all detectors and to keep it constant during the operation.These requirement implies that the user measures the breakdown voltage (  ) for all SiPMs and has the possibility to adjust the high voltage channel by channel.A standard method used to measure   requires to collect data illuminating the detector with a small flux of photons and making histograms of the recorded signals (Multi-Photon spectrum) at different voltages.The measured peak-to-peak distances (Δ pp ) are expected to increase linearly with the applied voltage.For this reason, the fit of Δ pp as a function of the voltage allows to extrapolate the point where the avalanche mechanism is turned off (  ).
For the ORIGIN dosimeter, we use a method based on staircase plots4 avoiding the need of using external light.The SiPMs are covered with a cap and staircase plots are acquired at different high voltages.This procedure can be fully automatized and can be performed, simultaneously, for all SiPMs.In this case, the Δ pp is defined as the distance between the baseline and the first inflection point extracted from this fit function: where erfc is a complementary error function and  identifies the inflection point in the formula.The baseline is the center of the noise signal and it is fixed for all measurements.Also in this case, we can use the Δ pp as a function of the voltage to estimate the   .Figure 3 shows the result obtained for one of the SiPM in use.The typical error found for the   extrapolation is ≈ 20 mV and it is calculated by propagating the errors on the parameters obtained from the linear fit, taking into account 4Typical plots where the DCR is measured as a function of the threshold set at the discriminator.
-5 - the correlation between the slope and the offset of the fit.The average   calculated for all SiPMs is 49.03 ± 0.18 V (0.4% spread), with a maximum variation of 0.63 V.The difference in the   is compensated by regulating individually the high voltage for each SiPM; the maximum regulation range provided by the system is 4.5 V.During this characterisation we had one faulty connection compromising the operation of one channel.This SiPM is not considered in the analysis.

JINST 19 C03025
Stability in time of the DCR is another important requisite for this dosimeter.If stable, we can consider to measure the DCR before the treatment and subtract a constant value from the measured counting rate.DCR measurements are performed counting the number of triggers received by each channel in time windows of 0.5 s when all SiPMs are covered by the caps.The DCR stability is assessed by fitting the data points with a line and by checking the compatibility of the angular coefficient with 0. This condition is achieved with a chi-square per degrees of freedom below 1.2.

Channels equalisation
Next step in the ORIGIN dosimeter characterisation is the assessment of the channel response uniformity.Since we know that PDE, Δ pp and DCR are over-voltage dependent, we firstly adjusted the high voltage to guarantee the same over-voltage for all SiPMs (HV equalisation) and then we measured the DCR.The operating voltage was pushed to a value of +6 V over the break-down voltage, where an improvement by ≈ 20% was measured in the MDL value compared to the nominal +3 V excess voltage.In this condition, the average DCR was measured to be 13.1 kHz,5 with a standard deviation of 2.9 kHz over the 16 SiPMs.This result strongly impacts the channel by channel variation of the MDL (1.1).For this reason we decided to compare the performance using a different equalisation method, namely adjusting the HV to equalise the DCR (DCR equalisation).This obviously allowed to reduce the spread of the DCR down to 0.4 kHz, equalising the MDL (see figure 4).On the other hand, the DCR equalisation required to modify the  ov of few SiPMs, possibly changing their PDE.In fact, accordingly to the datasheet, we could expect a spread in PDE up to ∼5% that, together with the geometrical acceptance, impacts the sensitivity.
5This value is higher than the DCR quoted by the company, because it is measured at higher  ov .
-6 - The ORIGIN system sensitivity, as defined in (1.2), was measured by illuminating the scintillating tip of the fibre with X-rays at different current settings.6 We characterised all channels by using the same fibre, knowing that the reproducibility of the measurement is better than 0.5% (value obtained by repeating 10 times the measurement after reconnecting the fibre to the same SiPM).Each run is 1 minute long and we record the trigger counts in time windows of 0.5 s, extracting the average value and its standard deviation.A DCR measurement is performed for all channels at the beginning of the scan and the constant term is subtracted for all measurements.The PCR, as function of the current settings, is used to get the angular coefficient () from the linear fit.The spread in the sensitivity, due to the PDE and the geometrical acceptance, is shown in figure 5 (left).To compensate for this spread, an equalisation coefficient is estimated for all channels:

JINST 19 C03025
where PCR 1 and PCR i are respectively the Photon Counting Rates of the first and i-th SiPM measured at 0.2 mA, which is a point in the middle of the current scan.The data collected at different X-ray currents are multiplied by the respective   and the new sensitivity is determined by fitting the corrected data.Figure 5 (right) shows the leftover spread after the equalisation.

Results
The ORIGIN dosimeter has been characterised and shown to work under stable conditions.Of the two methods used to equalise the channel response, DCR equalisation gave better results.This equalisation reduces the channel-to-channel variation of the MDL, which depends on the SiPM DCR and cannot be corrected off-line.The MDL spread was reduced from 10.5% (obtained with the HV-equalisation) to 1.6% (see table 2).Although a different over voltage induces a different PDE in SiPMs (50 to 55% in the peak sensitivity), the response to light could be equalised by a scaling factor accounting as well for possible acceptance variations due to mechanical uncertainties in the optical receptacle alignment with respect to the SiPM.These coefficients were calculated by characterising all channels with the same sensor.The procedure resulted in a spread in sensitivity, after the equalisation, that is well below 1%.The summary of the obtained results is given in table 2.
Table 2. Comparison between HV and DCR equalisation.For the MDL and Sensitivity is also quoted the channel-to-channel variation.The results quoted for the sensitivity are the ones obtained after the renormalisation described in the test.
HV-equalisation DCR-equalisation 483 ± 50 Hz (10.5%) 490 ± 7.9 Hz (1.6%) Sensitivity 1.9•10 6 ± 5.04•10 3 Hz/mA (0.3%) 1.81•10 6 ± 7.2•10 3 Hz/mA (0.4%) Finally, we measured the figures of merit with respect to the same sensors operated at room temperature and with the fibre tip at minimum distance from the sensor surface, a solution found to be non compliant with the ORIGIN specifications.Comparing the results obtained by the DCR equalised TE-cooled system to the ones from the room-temperature board (table 3), it can be noticed that the new board provides a remarkable enhancement in performance.In particular, MDL is halved and the sensitivity is tripled, which are results that should enable the collaboration to meet the project specifications.A measurement campaign is ongoing to verify the system sensitivity up to 3 cm, which was the only missing specification of the previous system.

Conclusion
The ORIGIN project aims to develop an in vivo real-time dosimeter for oncological brachytherapy, with dose mapping and source localisation capabilities.The proposed solution is based on 16 clear fibres incorporating a small scintillator at the tip to enable point-like measurements.The scintillating material considered in the project has a decay time of the order of 500 µs that produces a series of single photons detected by SiPMs.A single-probe prototype of the ORIGIN system was tested with low-dose-rate brachytherapy sources and the system was found to comply with TG43-U1 protocol, but the sensitivity was limited by the dark count rate values.In order to overcome this problem, the collaboration proposed the use of thermo-electrically cooled SiPMs to reduce the dark count rate when operating at low temperature.
The 16-channel dosimeter, equipped with thermo-electrically cooled SiPMs and a focusing system to improve the light collection, has been commissioned.All but one of the channels (not working for a faulty connection) could operate under stable conditions, and it was shown that, after the equalisation, the system achieved a minimum detectable light of 490 ± 7.9 Hz, a sensitivity of 1.81•10 6 ± 7.2•10 3 Hz/mA and a spread of these parameters of 1.6% and 0.4% respectively.Finally, the system has been shown to perform better than the previous version in terms of both minimum detectable light (two times better) and sensitivity (three times better).At the time of writing, the system is being qualified in the clinical environment for final performance evaluation.

Figure 1 .Figure 2 .
Figure 1.The focusing system and the mechanical drawing of its coupling with the TE-cooled SiPMs are shown on the left side of the picture, while a picture of the ORIGIN dosimeter with the readout board and the 16 sensors connected is shown on the right.

Figure 3 .
Figure 3.Typical staircase obtained with a SiPM (left) and Δ pp as a function of the voltage (right) used to estimate the   .

Figure 5 .
Figure 5. Sensitivity measured before and after the equalisation procedure (left and the right respectively).The high voltage is adjusted to have same DCR in all channels.

Table 1 .
Specifications of the ORIGIN project, for LDR.

Table 3 .
Comparison between room-temperature board and DCR equalised TE-cooled board.The numbers quoted for the room-temperature SiPMs are referred to one single channel.The errors quoted for the MDL and the sensitivity, in this case, refer to the error propagation of the measurement.± 9.21•10 3 Hz/mA 1.81•10 6 ± 7.2•10 3 Hz/mA (0.4%)