- Email: [email protected]
Time-resolved in vivo dosimetry for source tracking in brachytherapy
Brachytherapy
-
(2017)
-
Time-resolved in vivo dosimetry for source tracking in brachytherapy Jacob Graversen Johansen1,*, Susanne Rylander1, Simon Buus1, Lise Bentzen1, Steffen Bjerre Hokland1, Christian Skou Søndergaard1, Anders Karl Mikael With2, Gustavo Kertzscher3, Kari Tanderup1 1 Department of Oncology, Aarhus University Hospital, Aarhus C, Denmark € € Department of Medical Physics, Orebro University Hospital, Orebro, Sweden 3 Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX 2
ABSTRACT
PURPOSE: The purpose of this article is to demonstrate that brachytherapy source tracking can be realized with in vivo dosimetry. This concept could enable real-time treatment monitoring. METHODS: In vivo dosimetry was incorporated in the clinical routine during high-dose-rate prostate brachytherapy at Aarhus University Hospital. The dosimetry was performed with a radioluminescent crystal positioned in a dedicated brachytherapy needle in the prostate. The dose rate was recorded every 50e100 ms during treatment and analyzed retrospectively. The measured total delivered dose and dose rates for each dwell position with dwell times O0.7 s were compared with expected values. Furthermore, the distance between the source and dosimeter, which was derived from the measured dose rates, was compared with expected values. The measured dose rate pattern in each needle was used to determine the most likely position of the needle relative to the dosimeter. RESULTS: In total, 305 needles and 3239 dwell positions were analyzed based on 20 treatments. The measured total doses differed from the expected values by 4.7 8.4% (1SD) with range (17% to 12%). It was possible to determine needle shifts for 304 out of 305 needles. The mean radial needle shift between imaging and treatment was 0.2 1.1 mm (1SD), and the mean longitudinal shift was 0.3 2.0 mm (1SD). CONCLUSION: Time-resolved in vivo dosimetry can be used to provide geometric information about the treatment progression of afterloading brachytherapy. This information may provide a clear indication of errors and uncertainties during a treatment and, therefore, enables real-time treatment monitoring. Ó 2017 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved.
Keywords:
In vivo dosimetry; HDR brachytherapy; Source tracking; Dose rate measurement; Prostate cancer
Introduction Brachytherapy is a radiotherapy modality which involves a considerable number of manual procedures during treatment planning and delivery. Several investigations of radiation misadministrations in brachytherapy have shown that a substantial share of the events is caused by human errors (1e3). The identified misadministrations involved a Received 1 March 2017; received in revised form 11 August 2017; accepted 15 August 2017. Financial disclosure: This study was supported by an unrestricted research grant from Elekta and by Danish Cancer Society. * Corresponding author. Department of Oncology, Aarhus University Hospital, Norrebrogade 44, DK-8000, Aarhus C, Denmark. Tel.: þ45 78450000; fax: þ45 78464455. E-mail address: [email protected] (J.G. Johansen).
large range of different events including wrong connections of transfer tubes, reconstruction errors, and wrongly assigned source strength. Many events may remain unnoticed, and if detected, they are typically only identified posttreatment, due to the limited availability of commercial realtime treatment-monitoring systems. Furthermore, commercial software tools and algorithms that can provide guidance in the event of a treatment error are not available (4, 5). The development of treatment monitoring in brachytherapy has mainly focused on three approaches: imaging, tracking, and in vivo dosimetry (6). Imaging has the advantage of providing a direct link between source/applicator geometry and patient anatomy. However, extra acquisition and analysis of 3D imaging adds additional time to the treatment procedure. Furthermore, 3D imaging is typically not completely real-time, except for advanced
1538-4721/$ - see front matter Ó 2017 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.brachy.2017.08.009
2
J.G. Johansen et al. / Brachytherapy
setups as for example, suites including both afterloading units and MRI (7). Fluoroscopy provides real-time information but does not include soft tissue information, and the quality of the flouroscopy images is challenged by the radiation from the treatment source, at least during high-dose-rate (HDR) treatments (8, 9). This challenge has been overcome by increasing the X-ray strength and reducing the monitoring frequency (10). Electromagnetic tracking provides high geometric accuracy but is limited by the fact that the source is not tracked directly and that it does not link the geometric verification to the patient anatomy (11e14). In vivo dosimetry facilitates direct measurement of the absorbed dose in organs at risk (OAR) or close to the tumor. However, in vivo dosimetry is challenged by large positioning uncertainties of the dosimeters, limited utilization of real-time readouts, and that each dosimeter only provides the dose at a given point. Certain dosimeter types can only measure the accumulated dose (e.g., thermoluminescence dosimeter and alanine) while others can measure dose rates (e.g., diodes, metal-oxid-semiconductor field effect transistor and scintillators) (4, 5, 15, 16). Recent developments have reduced the size of detectors and the positional uncertainty and increased the sensitivity (agreement within 5%) and read-out frequency (O10 Hz) (17e23). This article will present in vivo dosimetry measurements performed with a system featuring high read out frequency, facilitating a study of individual dwell positions. The ability to analyze the dose rates for each dwell position opens new ways of identifying errors and uncertainties. Kertzscher et al. developed an adaptive algorithm for treatment monitoring, in which the most likely dosimeter position was determined (24). This article applies a methodology where dose rate for each dwell position is used to determine the positions of the individual treatment needles relative to the dosimeter. Source tracking with multipoint dosimeter has previously been explored both in phantoms (25e29) and in pulsed dose rate cervix treatments (30), but this is the first time a single dosimeter has been used for obtaining geometrical information for each individual needle in a considerable number of patients. The ability to track the source directly enables identification of misplacements of the source relative to the dosimeter. Detectable misplacements could be, for example, needle reconstruction errors, wrong connections of guide tubes or positional offsets by the afterloader, whereas an identical shift of both source and dosimeter relative to the anatomy cannot be detected.
Methods and materials The instruments and dose measurements The in vivo dosimetry was performed using a small (0.5 0.5 2 mm3) radioluminescent crystal made of Al2O3:C grown by Landauer Inc (USA) using the Czochralski
-
(2017)
-
technique (growth CZ#60). The crystal was coupled to a 15 m long optical fiber connecting to a data acquisition box (DAQ). The DAQ consisted of a bandpass filter (395e440 nm), a photo multiplier tube (PMT) (Hamamatsu h7360-01), and a DAQ card (NI9402). The bandpass filter was placed between the PMT and the optical fiber to accept the peak wavelength region of the Al2O3:C emission and remove the other wavelengths of the contaminating Cerenkov and fluorescence light that is emitted in the optical fiber, referred to as the stem effect (19, 31, 32). The PMT was connected to the DAQ card, which counted the number of detected photons in a given time interval. The crystal and optical fiber were shielded from light using a plastic tube. The outer diameter of the crystal and tube was 1 mm, which was thin enough to fit into the brachytherapy plastic needles used for the treatment. The DAQ was placed in a neighboring radiation shielded room to avoid false signals from radiation striking the PMT. The entire system is described in details in the study by Andersen et al. (19), which also contains a schematic drawing of the system and a detailed characterization of the crystal. The clinical procedure The in vivo dosimetry procedure was carried out by the clinical staff as an integrated part of the clinical brachytherapy procedure. Patients with D’Amico high-risk prostate cancer were treated with a combination of external beam radiotherapy (46 Gy in 23 fractions) and two individual fractions of HDR brachytherapy of each 8.5 Gy prescribed to the prostate þ 3 mm (33). Before implantation of needles in the patient, pre-irradiation and calibration of the crystal were performed by the medical physicist to account for potential changes in the detection efficiency of the system, for instance due to a recoupling of the optical fiber to the DAQ-box. This process was performed using a block of solid water with two parallel needles separated by 10.4 0.2 mm and took 15e30 minutes depending on the source strength (34). Transperineal insertion of plastic needles was performed under live transrectal ultrasound guidance. After insertion of all treatment needles, an additional needle, dedicated to the dosimeter, was positioned in the prostate as best feasible but preferentially in the center of the gland. Approximately, 1 h after needle implantation, the patient underwent an axial T2W MRIs acquired with a 2-mm slice thickness and a 1.4-mm transversal accuracy. The MRI was used for treatment planning, which involved contouring, needle reconstruction, dosimeter localization, and dose optimization (Oncentra Prostate 4.2.21, Elekta). Immediately before treatment delivery, an additional MR scan was obtained to detect and adjust for needle migration O3 mm, followed by reoptimization of the plan. Immediately before dose delivery, the dosimeter was placed in the dedicated needle and inserted to the end of the lumen, with the center of the luminescent crystal located 9.5 mm from the outer tip of the needle, based on X-ray exposure
J.G. Johansen et al. / Brachytherapy
of the crystal and fiber on radiochromic film. Throughout the delivery, the dosimeter measured and recorded the dose rate every 50e100 ms as well as accumulated dose. The planned total dose to the center of the crystal was evaluated in the treatment planning system. The brachytherapy treatment was delivered using the Flexitron afterloader (Elekta) with TCC2.1.3. A third MR scan was acquired immediately after delivery to evaluate the stability of the treatment needles (35).
The retrospective analysis The retrospective analysis of the dosimetry data was performed on three different levels: (1) Accumulated dose for entire treatment, (2) Dose rates in individual dwell positions, (3) Combination of all dose rates in each individual needle. Planned doses and dose rates were calculated by the TG43 formalism (36), using the geometry of dwell positions and the dosimeter. The dose rates were integrated over the length of the crystal to account for the finite size, and the geometrical values were obtained from DICOM files from the treatment planning system based on the needle reconstruction performed by the clinical staff. The total doses were obtained by accumulating dose rates throughout the treatment (Dmeas) and the relative offset to the expected dose (DTG43) was calculated: DD/ D5(Dmeas/DTG43-1)*100%. The dosimeter system provided 10e20 dose rate measurements per second. An average dose rate (Ḋmeas) was determined for each dwell position. The first 0.4 s and the last 0.1 s of each dwell position were omitted in the averaging, due to the movement of the source and a short stabilization time needed by the crystal. The analysis was only performed for dwell positions with dwell times longer than 0.7 s to allow for averaging over at least 0.2 s. The angle and distance required by the TG43 formalism were determined based on the coordinates for each dwell position given in the treatment plan, see Fig. 1. The relative offsets between measured and expected dose rate were determined as for the total dose: DḊ/Ḋ 5 (Ḋmeas/ḊTG43 1) 100%. Δr
Δradial
radial
r
longitudinal
θ
Δlongitudinal
Fig. 1. Schematic representation of a treatment needle with five dwell positions (circles) and the needle with dosimeter (square). The coordinates used in the analysis are also shown. Free parameters are shown in blue. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
-
(2017)
-
3
The discrepancy in dose rates DḊ was assumed to be due to an offset in the distance (Dr) between source and dosimeter, see Fig. 1. Under the condition that Dr was small, the change in anisotropy, scattering, absorption as well as on energy dependence of the crystal was assumed to be negligible, and an inverse square dependency was used in the analysis. In this case, a simple relation between the relative dose rate offsets and the distances could be obtained: 2 ð1Þ DD_ D_ 5 r2 ðr þ DrÞ 1 5 r2 rmeas 1 Here rmeas refers to the most likely distance between the source and the dosimeter based on the measured dose rate. This value was determined for all the dwell positions, by isolating rmeas in equation 1 and comparing to the expected distances (r). Furthermore, any uncertainty in the relative distances would be equal to an uncertainty in Dr. This would lead to the following relation between the standard deviation (SD) of the relative dose rate offset and the uncertainty of the distances, which also includes the internal uncertainty of the system (udosimeter): 2 2 SD DD_ D_ 5 ð2SDðDrÞ=rÞ þ udosimeter 2
ð2Þ
According to this equation, the SD of the relative offset has a distance dependency. The SD of the relative offset was therefore determined in distance intervals of 4 mm and fitted to a function based on equation 2 with SD (Dr) and udosimeter as free parameters. Equation 1 only provides the distance between the source and the dosimeter and not the direction. In fact all movements are assumed to be along the r-vector, see Fig. 1, whereas the largest shift is expected to be along the longitudinal needle direction since the uncertainty of the needle tip reconstruction and source positioning in the longitudinal direction is larger than in the radial direction and also because needle migration between imaging and treatment delivery occurs in the longitudinal direction. To account for this, a more thorough geometrical analysis was performed. It was assumed that needles could only move rigidly and that all dwell positions in a given needle would undergo common shifts. The principle behind the analysis was to shift each needle, and thereby any dwell position within the needle, to best fit the measured and TG43 dose rates for all dwell positions within the needle. This was done in three steps. First, all measured dose rates in a needle were plotted as a function of their distance from the afterloader indexer or index number (black dots in Fig. 2). Then, a 1D-function of the dose rate as a function of the index number in the needle was derived based on the TG43 formalism (ḊTG43(i)) (blue lines in Fig. 2). The function uses the geometry from the treatment plan to transform the index number to a geometric position and thereafter determine an r-vector, Fig. 1. The r-vector can be resolved into a longitudinal and a radial part. If a needle includes two or more source dwell positions, it is possible to determine one longitudinal shift (Dlongitudinal) and a radial shift (Dradial), which gives the best fit between ḊTG43(i) and Ḋmeas. The function ḊTG43(i)
4
J.G. Johansen et al. / Brachytherapy
b
×10 3
100 ×10
(2017)
-
3
c/s
c/s
a
-
50
90 45
80 70
40
60 35 50 30
40
25
30 20
20 10 15 145
150
155
160
165
170
175
0
180
150
155
160
165
170
175
180
index
185 index
Fig. 2. Two examples of the needle shift fit routine. The black dots are the measured count rates for the dwell positions, the blue line is the expected curve based on the planned positions and TG43, the red curve is after the shift in needle position. Panel ‘‘a’’ shows a needle with a longitudinal shift, and panel ‘‘b’’ shows a fit with a radial shift. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
was fitted to the experimental data, with Dlongitudinal and Dradial as free parameters. The fit was performed with the ROOT software (v. 6.06 CERN) (red lines in Fig. 2).
SD of the relative offsets are given at the bottom of the table. The low dose, below 8 Gy, in six of the treatments is due to the placement of the dosimeter just outside the prostate.
Results
Dose rates for individual source positions
Measurements were performed in a total of 22 treatments in 12 patients. In two cases, the dosimeter had slid out of the needle by 17 mm and 70 mm, respectively, during the treatment. These two treatments were omitted in the analysis leaving 20 treatments analyzed. In six of the remaining 20 treatments, the dosimeter was placed outside the prostate, close to the apex, because the dosimeter needle was not inserted deep enough into the prostate.
A total of 3239 of 3518 dwell positions had a dwell time longer than 0.7 s. The dose rates and the relative distance between the source and dosimeter (r) for each dwell position were determined. The average offset in dose rate was 4.3 13.4%, and the corresponding source dwell position shift was 2.1 2.3 mm. Figure 4a shows the relative offset between measured and expected dose rate as a function of the distance between dwell position and dosimeter. Figure 4b shows the distance determined from the in vivo dosimetry measurements (rmeas), calculated using equation 1 against the expected distance based on the treatment plan (r). To investigate the distance dependencies seen in Figs. 4a and 4b, the dwell positions were grouped according to the distance to the dosimeter in 4-mm intervals starting from 10 mm. The mean and SD of both the relative dose rate offset and the distance were determined for each of the groups. The mean values are given in Figs. 4c and 4d and the SDs in Figs. 4e and 4f. The SD of the relative offsets was fitted to equation 2 as described in the Materials and methods Section. The result of the fit is given in Fig. 4e. A strong linear relation was seen between the two distances, Fig. 4d. A linear fit was made, and the result indicated a general 10% increase in all distance as seen by the fit result in Fig. 4d.
Raw data Figure 3 shows dose measurements for three treatments, one with the dosimeter located close to the apex (a þ b) and two with the dosimeter in the center of the gland (cef). The left panels (a, c, and e) show the dose rates as a function of time and the right panels (b, d, and f) show the accumulated dose as a function of time. The inserts in a, c, and e show a close up of the dose rate pattern in one needle. The black and blue curves are the expected and measured values, respectively. The source progression into each individual needles was uniquely identifiable for all needles both from dose rate and dose accumulation plots. The stability of the dose rate measurements can be seen on the flat structures in the inserts. Total delivered dose The total measured doses for all 20 treatments are shown in Table 1 along with the expected delivered doses and the relative offset between the two (three first columns). The average and
Needle shifts The method was tested on the calibration data before the use on the clinical data. The calibration was performed
J.G. Johansen et al. / Brachytherapy
b Accum. dose [Gy]
Dose rate [Gy/s]
a 0.14 80000
0.12
70000 60000 50000 40000
0.1
30000 20000 10000
0.08
0 100
120
140
160
180
200
220
240
0.06
-
(2017)
-
5
5 4 3 2
0.04 1
0.02 0 0
400
600
800
×10
3
100
80
60
40
0.2
200
400
600
800
1000 time [s ]
d
0.3 0.25
0 0
1000 time [s ]
Accum. dose [Gy]
Dose rate [Gy/s]
c
200
20
0 120
140
160
180
200
220
240
260
280
12 10 8
0.15
6
0.1
4
0.05
2
0 0
200
400
600
800
1000
1200
e
1400 time [s ]
0 0
200
400
600
800
1000
1200
f
1400 time [s ]
×10
Accum. dose [Gy]
Dose rate [Gy/s]
0.3 3
120
0.25
100
80
60
0.2
40
20
0 40
50
60
70
80
90
100
110
10 8
0.15
6
0.1
4
0.05
2
0 0
100
200
300
400
500
600
700
800
900 time [s ]
0 0
100
200
300
400
500
600
700
800
900 time [s ]
Fig. 3. The measured data provided by the in vivo dosimetry system (blue lines) and the expected values based on the treatment plan (black lines). The measurements were updated every 50 or 100 ms during the treatments. The left panel (a, c, and e) is the dose rates, and the right panel (b, d, and f) is the accumulated dose. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
with 21 dwell positions separated by 3 mm in a needle with a minimum distance to the dosimeter of 10.4 0.2 mm (34). The calibration measurements agreed within 0.05 mm both radially and longitudinally. A total of 305 needles were analyzed, and radial and longitudinal shifts were determined for all except one needle, which only had one dwell position longer than 0.7 s. The average shift per patient is shown in Table 1, and a scatter plot of the radial (X) and longitudinal (Y) shifts for all needles in all the 20 treatments are given in Fig. 5. The average shifts for all needles were determined using a Gaussian fit (right panel in Fig. 5): Dradial 5 0.19 1.1 mm, Dlongitudinal 5 0.34 2.0 mm. The total displacement was Dr 5 (Dradial2þ Dlongitudinal2)½ 5 0.39 1.8 mm. The last row in Table 1 shows the overall average and the SD of the mean values across all treatments. The mean
shift across all needles within a given treatment can be interpreted as a potential shift (or reconstruction uncertainty) of the dosimeter. The average of the mean values is therefore an indicator of any general offset in the dosimeter position, whereas the SD of the mean values indicates the reproducibility of the dosimeter position. The values in the parentheses are the average and SD of the SDs determined for each treatment. The SDs of the needle shifts in a given treatment can be interpreted as the positional uncertainty of the treatment needles relative to the dosimeter. Figure 6 shows similar plots as in Figs. 4 and 5 for three selected treatments. The treatments represent the following scenarios: top: a treatment with stable geometry corresponding to needle shifts !3 mm and small dosimeter offset; middle: a treatment with offsets which can be explained by a slight systematic shift of the dosimeter of
6
J.G. Johansen et al. / Brachytherapy
-
(2017)
-
Table 1 Results from each of the 20 treatments Treatment
Dmeas (Gy)
DTG43 (Gy)
DD/DTG43 (%)
DRadial (mm)
DLongitudinal (mm)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Average SD
4.95 3.68 10.4 3.90 10.7 9.62 11.2 10.3 9.68 10.4 10.8 11.6 10.4 8.95 6.74 9.31 3.69 9.56 10.1 4.39
5.64 4.43 10.7 4.11 10.4 10.2 10.9 11.4 9.90 11.5 9.80 12.7 11.0 10.0 7.76 10.9 3.29 10.2 10.9 4.01
12.1 16.9 2.71 5.26 3.08 6.04 3.07 9.56 2.21 10.1 10.4 9.10 5.37 10.6 13.2 14.3 12.3 6.70 7.53 9.54 4.66 8.40
0.15 0.56 1.3 0.96 0.27 0.4 0.06 1.2 0.23 0.38 0.17 0.50 0.54 0.47 0.97 0.72 0.08 0.47 0.95 0.62 0.25 0.57 0.64 0.53 0.84 0.45 1.3 0.80 0.58 0.50 0.37 0.74 0.09 0.53 0.68 0.83 1.2 1.1 1.6 0.87 0.17 0.77 (0.67 0.24)
1.9 1.3 2.6 2.1 1.5 1.6 0.84 1.9 0.79 0.96 1.9 0.95 0.03 0.71 0.62 0.87 0.93 1.5 0.64 0.62 0.19 1.2 0.02 1.3 0.38 1.1 1.1 1.1 1.3 1.9 1.8 2.0 0.91 1.7 0.82 2.2 2.6 2.1 0.01 0.97 0.31 1.3 (1.40 0.50)
The first two columns are the measured and expected accumulated doses respectively, and the third column is the offset between the two. The last two columns are the mean and SD of the radial and longitudinal needle shifts in the treatment. The last row marked with bold letters shows the average and the standard deviations (SDs) for the mean values of the individual treatments. The values in the parentheses are the average and SD of the standard deviations of the treatments.
2 mm; and bottom: a treatment where two needles show specific larger offsets corresponding to 4 mm radial and 9 mm longitudinal shifts, respectively. The treatments are treatment 7, 6, and 2 respectively. Discussion This article presents in vivo dose rate measurements for 3239 dwell positions in 305 needles from 20 HDR treatments of prostate cancer patients. The dose rate measurements were analyzed both with regard to the total dose as well as according to dose rates from individual source positions. By exploiting the offsets between measured and expected dose rates, it was demonstrated that the dose rate information could be transformed into geometrical information of the source progression throughout the treatment. The total average offset of accumulated dose was 4.7% across the treatments with an 8.4% spread (1SD). Previous studies (22, 37e40) reporting on in vivo dosimetry in HDR brachytherapy have in general reported similar or higher offsets with typical ranges around 20% and up to 40% (around 10% SD), which is most often attributed to positional inaccuracies of the dosimeter. In a study by Mason et al. (41), the dosimeter (metal-oxid-semiconductor field effect transistor) was positioned in a needle, and the mean offset was 3.0% with a maximum of 11.2% similar to this study. One patient treatment in our study had a nominal offset in absolute dose above 15%, which is more than expected. This offset was identified as stemming from two needles with considerable offsets in the dose rates (bottom row in Fig. 6).
One of the advantages of dose rate measurements compared to total dose measurements is the possibility of identifying specifically which needles are related to geometrical offsets. The dose rate measurements presented in Fig. 6 show that offsets can occur if some treatment needles are shifted relative to the reconstructed position (bottom row in Fig. 6) or if the dosimeter has shifted (middle row in Fig. 6). The offsets in dose rates of the two treatments show significantly different patterns, which allows for a separation of the two types of events. With the ‘‘Adaptive Error Detection Algorithm’’ (AEDA) proposed by Kertzscher et al. (24), this strategy is further exploited. The reconstruction of the needles based on radial and longitudinal shifts, lead to an overall 1.8 mm uncertainty (SD). The main uncertainty was along the needle, which can be explained by larger reconstruction and source positioning uncertainties in the longitudinal direction than in lateral and anterioreposterior directions (42, 43). Reconstruction uncertainties are affecting positional uncertainties for both treatment needles and the dosimeter, and the contribution of each can be determined by looking at the individual treatments, Table 1. The mean shift across all needles in a single treatment is an indication of a shift in the dosimeter position, whereas the SD represents the uncertainty in the position of treatment needles. The SD of the mean shifts from all treatments was interpreted as an uncertainty in the positioning of the dosimeter and was determined to 0.77 mm (radial) and 1.3 mm (longitudinal). The average of the SDs across all treatments was on the other hand interpreted as the
J.G. Johansen et al. / Brachytherapy
10
rmeas[mm]
12
100
. .
(2017)
-
7
b
ΔD/D [%]
a
-
16
60
14
50
12
50
40
10
8 0 −50
8
30
6 4
20
2
10
6 4 2
−100 0
. .
20
30
40
50
60 r [mm]
0 0
0
d
0
Mean of r meas[mm]
Mean of ΔD/D [%]
c
10
−1 −2 −3 −4
10
20
30
40
50
60 r [mm]
0
60
y = 1.1r-0.75mm
50 40 30
−5
20
−6 10
−7
SD of ΔD/D [%]
e . .
10
20
30
40
50
0 0
60 r [mm]
f
25
y2 = (2.1.14mm/r)2 +7.2%2
20 15
SD of r meas[mm]
−8 0
1
30
40
50
60 r [mm]
40
50
60 r [mm]
10
20
30
40
50
60 r [mm]
3
5
20
30
4
2
10
20
5
10
0 0
10
0 0
Fig. 4. Panels ‘‘a’’ and ‘‘b’’ show scatter plots of the dose rate deviations and the recalculated distance as function of the expected distance respectively. The color bars indicate the number of dwell position within the pixel. The last four panels show the mean (c and d) and standard deviations (e and f) for given distance intervals between 10 mm and 50 mm. The panels ‘‘c’’ and ‘‘e’’ are for the dose rate offsets, and ‘‘d’’ and ‘‘f’’ are for the recalculated distances. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
positional uncertainty for the treatment needles. These were determined to 0.67 mm (radial) and 1.4 mm (longitudinal), Table 1. This magnitude of uncertainties is in good agreement with the 0.6 mm and 1.2 mm radial and longitudinal reconstruction uncertainties found by Haack et al. (42). The study of all dwell positions (Fig. 4) showed that the offsets in general could be explained by a general 7.2% uncertainty of the system overlaid with a 1.14 mm uncertainty (1SD) in the relative geometry (Fig. 4e). These observations are well in agreement with what is
expected from typical reconstruction uncertainties in MRI-based brachytherapy (44) and with the expected uncertainty of the dosimetry system. The dosimeter has previously been tested in pulsed dose rate cervix treatments (45, 46). Andersen et al. reported an 8% uncertainty for the dosimeter system, which includes reproducibility, energy response, angular dependence, stem effect, transmission, and calibration (19). The stem effect may be significant in configurations where the source is close to the optical fiber but far from the crystal. However, with our configuration, the stem effect is !1% for all dwell
N/0.2mm
J.G. Johansen et al. / Brachytherapy Δ Longitudinal [mm]
8
10
-
(2017)
-
30 25 20 15 10
5
5 0 −10
−8
−6
−4
−2
0
2
4
6
Δ Radial [mm]
−8
−6
−4
−2
0
2
4
6
Δ Longitudinal [mm]
N/0.2mm
0
−5
8
10
8
10
20 15 10
−10
5
−10
−5
0
5
10 Δ Radial [mm]
0 −10
Fig. 5. A representation of the radial and longitudinal shifts determined for all 304 needles. The left panel presents the longitudinal shift vs. the radial shift for all the 304 needles. The right panel is the individual histogram for the radial (top) and longitudinal (bottom) shifts. The red lines are the Gaussian fits. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
positions above the dosimeter, !2% for all positions within 3 cm below the dosimeter, and with larger separations than 1 cm to the fiber (47). Only for specific positions further than 3 cm away and with a distance to the fiber of 0.5e1 cm, the stem effect uncertainty may reach 5% (47). Using the uncertainty budget of Andersen et al. while reducing stem effect to 1% and correcting for energy dependence, the uncertainty (SD) of the system (udosimeter) reaches 5%, making it the dominating uncertainty compared to the uncertainties of count statistics. Furthermore, the systematic 4.7% measured underdosage could be explained by a 10% increase of all distances in the treatment or a systematic deviation in calibration. A contribution to the increase could be a swelling of the prostate. Buus et al. reported a 10% increase in the prostate volume between two fractions (33), which indicated the presence of swelling. Thorough phantom measurements to study the calibration and a study of the swelling of the prostate during a brachytherapy treatment are required to understand their relative contributions to the systematic offset. Figure 4f shows that the SDs of the dose rate offsets are transformed into a 2e3 mm SD in the distances. This is significantly larger than the 1.14 mm uncertainty in Dr found in the fit of Fig. 4e. The reason is that all dose rate uncertainties are assigned to the positional uncertainty (equation 1), and therefore, the 5% system uncertainty will lead to slightly overestimated positional shifts. This plays a more significant role at larger distances. For example, 5% dosimetric uncertainty equals a positional uncertainty of 0.5 mm at a 20 mm and 1 mm uncertainty at a 40 mm distance (equation 2). A reduction of the system uncertainty would therefore improve the geometrical precision, for example by improving the signal output, which will reduce the effect of background and stem signal.
Limitations Although the transformation from dose and into geometry is promising for real-time monitoring of brachytherapy treatments, it also has limitations. The most important limitation is that geometrical shifts are assessed relative to the dosimeter and not relative to the patient anatomy. Drifts of the entire template would not be detectable with this system, as it will lead to a common drift of both the dosimeter and the treatment needles and no change in relative distance. It is well known that potential drift of the entire implant is a significant challenge in prostate HDR (35, 48, 49). A precise determination of the dosimeter relative to the patient anatomy would therefore improve the potential of tracking with in vivo dosimetry. One way to overcome this problem could be to combine in vivo dosimetry with fluoroscopy. Secondly, the method only provides information about the size of the relative offset, hence the determined offset could stem from an offset in the dwell position, an offset in the dosimeter position, or a combination. For instance, 2 mm offsets caused by needle drifts, reconstruction errors, or by the afterloader cannot be distinguished. Finally, the internal uncertainty of the dosimeter has already been discussed, and it was shown that it may transform into up to ~1 mm geometrical uncertainty. Perspective Despite the limitations, in vivo dosimetry with subsecond dwell time read-out has a strong potential for realtime treatment monitoring. This article presents a retrospective study of the dose rates, but all the information shown in Fig. 6 can be generated in real time. In this way, any positional offsets of the source could be presented to the clinical staff within seconds after it occurs,
100
50
-
(2017)
60
-
4
10
50 40
2
9
Δ Longitudinal [mm]
. . Δ D/D [%]
rMEAS[mm]
J.G. Johansen et al. / Brachytherapy
0
30
−50
20
2
5
0
−5
10 −100
−10 20
30
40
50 60 r [mm]
0
. . Δ D/D [%]
2
100
0 0
10
20
30
40
50
60 r [mm]
60
0
4
40
0
30
1
2
20
−50
−10
−5
0
5 10 Δ Radial [mm]
−10
−5
0
5 10 Δ Radial [mm]
−10
−5
0
5 10 ΔRadial [mm]
10
50
50
ΔLongitudinal [mm]
10
rMEAS [mm]
0
5
0
−5
10 −100
−10 20
30
40
50 60 r [mm]
0
. . ΔD/D [%]
2
100
50
0 0
10
20
30
40
50
60 r [mm]
60
0
2
10
50 40
0
1
30
1
20
−50
ΔLongitudinal [mm]
10
rMEAS [mm]
0
5
0
−5
10 −100 0
−10 10
20
30
40
50 60 r [mm]
0
0 0
10
20
30
40
50
60 r [mm]
0
Fig. 6. Results from three selected treatments (7, 6, and 2 from Table 1). The left panels are the relative offsets in dose rates. The middle panels are the distance determined using the dose rates. The right panels are scatter plots of the radial and longitudinal needle shifts. The color bars indicate the number of dwell position within the pixel. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
and the treatment could be stopped for further investigation in case of unacceptably large offsets. This would require an implementation of a real-time dosimetry protocol in the clinical workflow. The protocol should include acceptable limits for the determined offsets. The limits should take into account that the determined offsets could include contributions from both offsets in source and dosimeter. For instance, by having both a maximum acceptable offset and a maximum on the spread in the offsets, similar to what is described in the study by Kertzscher et al. (24). In addition, a way to better fasten the dosimeter to the needle is required, to avoid drifting of the dosimeter. A further perspective is to iteratively reconstruct the needle geometry based on the dwell positions determined from the in vivo dosimetry measurements. Based on such a reconstruction, a delivered dose distribution could be generated; hence expanding the use of in vivo dosimetry
from the evaluation of a point dose to the evaluation of a 3D dose distribution. Conclusion Time-resolved in vivo dosimetry could be used to provide geometric information about the treatment progression of remote afterloading brachytherapy. Such source tracking has the potential to provide a clear indication of errors and uncertainties during a treatment, hence enabling real-time treatment monitoring using in vivo dosimetry. Acknowledgments We wish to thank Dr. M.S. Akselrod, Stillwater Crystal Growth Division, Landauer Inc, USA for the Al2O3:C crystals. This study was supported by an unrestricted research grant from Elekta and by Danish Cancer Society.
10
J.G. Johansen et al. / Brachytherapy
References [1] ICRP. Prevention of high-dose-rate brachytherapy accidents. ICRP Publication. Ann ICRP97 2005;35:1e51. [2] IAEA. Lessons learned from accidental exposures in radiotherapy. IAEA Safety Report Series 17. Vienna: IAEA; 2000. [3] Richardson S. A 2-year review of recent Nuclear Regulatory Commission events: what errors occur in the modern brachytherapy era? Pract Radiat Oncol 2012;2:157e163. [4] Kertzscher G, Rosenfeld A, Beddar S, et al. In vivo dosimetry: trends and prospects for brachytherapy. Br J Radiol 2014;87:20140206. [5] Tanderup K, Beddar S, Andersen CE, et al. In vivo dosimetry in brachytherapy. Med Phys 2013;40:070902. [6] Bert C, Kellermeier M, Tanderup K. Electromagnetic tracking for treatment verification in interstitial brachytherapy. J Contemp Brachytherapy 2016;8:448e453. [7] de Leeuw H, Moerland MA, van Vulpen M, et al. A dual-plane co-RASOR technique for accurate and rapid tracking and position verification of an Ir-192 source for single fraction HDR brachytherapy. Phys Med Biol 2013;58:7829e7839. [8] Sheikh-Bagheri D, Munro P. A Monte Carlo study of verification imaging in high dose rate brachytherapy. Med Phys 1998;25: 404e414. [9] van’t Riet A, Mak ACA, te Loo HJ, et al. Verification of source position in interstitial and intraluminal brachytherapy. Radiother Oncol 1994;33:167e170. [10] Nose T, Chatani M, Otani Y, et al. Real-time verification of a highdose-rate Ir-192 source position using a modified C-arm fluoroscope. Int J Radiat Oncol Biol Phys 2017;97:858e865. [11] Bharat S, Kung C, Dehghan E, et al. Electromagnetic tracking for catheter reconstruction in ultrasound-guided high-dose-rate brachytherapy of the prostate. Brachytherapy 2014;13:640e650. [12] Poulin E, Racine E, Binnekamp D, et al. Fast, automatic, and accurate catheter reconstruction in HDR brachytherapy using an electromagnetic 3D tracking system. Med Phys 2015;42:1227. [13] Boutaleb S, Racine E, Fillion O, et al. Performance and suitability assessment of a real-time 3D electromagnetic needle tracking system for interstitial brachytherapy. J Contemp Brachytherapy 2015;7:280e289. [14] Beaulieu L, Racine E, Han DY, et al. Real-time electromagnetic trackingebased treatment platform for high-dose-rate prostate brachytherapy: clinical workflows and end-to-end validation. Brachytherapy 2017. http://dx.doi.org/10.1016/j.brachy.2017.04.247 [Epub ahead of print]. [15] Therriault-Proulx F, Beaulieu L, Beddar S. Validation of plastic scintillation detectors for applications in low-dose-rate brachytherapy. Brachytherapy 2017;16:903e909. [16] Van Gellekom MPR, Canters RAM, Dankers FJWM, et al. In vivo dosimetry in gynecological applicationsdA feasibility study. Brachytherapy 2017. http://dx.doi.org/10.1016/j.brachy.2017.04.240 [Epub ahead of print]. [17] Kertzscher G, Beddar S. Ruby-based inorganic scintillation detectors for 192Ir brachytherapy. Phys Med Biol 2016;61:7744e7764. [18] Lambert J, Yin Y, McKenzie DR, et al. A prototype scintillation dosimeter customized for small and dynamic megavoltage radiation fields. Phys Med Biol 2010;55:1115e1126. [19] Andersen CE, Nielsen SK, Greilich S, et al. Characterization of a fiber-coupled Al2O3:C luminescence dosimetry system for online in vivo dose verification during 192Ir brachytherapy. Med Phys 2009;36:708e718. [20] Therriault-Proulx F, Archambault L, Beaulieu L, et al. Development of a novel multi-point plastic scintillation detector with a single optical transmission line for radiation dose measurement. Phys Med Biol 2012;57:7147e7159. [21] Cherpak AJ, Cygler JE, Choan E, et al. Real-time measurement of urethral dose and position during permanent seed implantation for prostate brachytherapy. Brachytherapy 2013;13:169e177.
-
(2017)
-
[22] Carrara M, Tenconi C, Rossi G, et al. Rectal wall measurements during HDR prostate brachytherapy with MO dosimeters integrated on a trans-rectal US probe: Comparison with planned and reconstructed doses. Radiother Oncol 2016;118:148e153. [23] Carrara M, Romanyukha A, Tenconi C, et al. Clinical application of MOSkin dosimeters to rectal wall in vivo dosimetry in gynecological HDR brachytherapy. Phys Med 2017;41:5e12. [24] Kertzscher G, Andersen CE, Tanderup K. Adaptive error detection for HDR/PDR brachytherapy: guidance for decision making during real-time in vivo point dosimetry. Med Phys 2014;41:052102. [25] Therriault-Proulx F, Beddar S, Beaulieu L. On the use of a single-fiber multipoint plastic scintillation detector for 192Ir high-dose-rate brachytherapy. Med Phys 2013;40:062101. [26] Nakano T, Suchowerska N, McKenzie DR, et al. Real-time verification of HDR brachytherapy source location: implementation of detector redundancy. Phys Med Biol 2005;50:319e327. [27] Smith RL, Haworth A, Panettieri V, et al. A method for verification of treatment delivery in HDR prostate brachytherapy using a flat panel detector for both imaging and source tracking. Med Phys 2016;43:2435e2442. [28] Guiral P, Ribouton J, Jalade P, et al. Design and testing of a phantom and instrumented gynecological applicator based on GaN dosimeter for use in high dose rate brachytherapy quality assurance. Med Phys 2016;43:5240e5251. [29] Fonseca G, Podesta M, Bellezzo M, et al. Online pretreatment verification of high-dose rate brachytherapy using an imaging panel. Phys Med Biol 2017;62:5449e5461. [30] Tanderup K, Christensen JJ, Granfeldt J, et al. Geometric stability of intracavitary pulsed dose rate brachytherapy monitored by in vivo rectal dosimetry. Radiother Oncol 2006;79:87e93. [31] Beddar AS, Mackie TR, Attix FH. Cerenkov light generated in optical fibres and other light pipes irradiated by electron beams. Phys Med Biol 1992;37:925e935. [32] de Boer SF, Beddar AS, Rawlinson JA. Optical filtering and spectral measurements of radiation-induced light in plastic scintillation dosimetry. Med Phys 1999;38:945e958. [33] Buus S, Rylander S, Hokland S, et al. Learning curve of MRI-based planning for high-dose-rate brachytherapy for prostate cancer. Brachytherapy 2016;15:426e434. [34] Kertzscher G, Andersen CE, Tanderup K. Fast, simple, and accurate: in vivo dosimeter calibration for HDR/PDR brachytherapy. Int J Radiat Oncol Biol Phys 2014;90:S742. [35] Buus S, Lizondo SM, Hokland S, et al. Needle migration and dosimetric impact in high dose rate brachytherapy for prostate cancer evaluated by repeated MRI. Brachytherapy 2017;. [this issue]. [36] Nath R, Anderson LL, Luxton G, et al. Dosimetry of interstitial brachytherapy sources: recommendations of the AAPM radiation Therapy Committee Task Group No. 43. Med Phys 1995;22:209e234. [37] Nose T, Koizumi M, Yoshida K, et al. In vivo dosimetry of high-doserate interstitial brachytherapy in the pelvic region: use of a radiophotoluminescence glass dosimeter for measurement of 1004 points in 66 patients with pelvic malignancy. Int J Radiat Oncol Biol Phys 2008; 70:626e633. [38] Anagnostopoulos G, Baltas D, Geretschlaeger A, et al. In-vivo thermoluminescence dosimetry dose verification of transperineal 192Ir high-dose rate brachytherapy using CT-based planning for the treatment of prostate cancer. Int J Radiat Oncol Biol Phys 2003;57:1183e1191. [39] Suchowerska N, Jackson M, Lambert J, et al. Clinical trials of a urethral dose measurement system in brachytherapy using scintillation detectors. Int J Radiat Oncol Biol Phys 2011;79:609e615. [40] Seymour EL, Downes SJ, Fogarty GB, et al. In vivo real-time dosimetric verification in high dose rate prostate brachytherapy. Med Phys 2011;38:4785e4794. [41] Mason J, Mamo A, Al-Qaisieh B, et al. Real-time in vivo dosimetry in high dose rate prostate brachytherapy. Radiother Oncol 2016;120: 333e338.
J.G. Johansen et al. / Brachytherapy [42] Haack S, Nielsen SK, Lindegaard JC, et al. Applicator reconstruction in MRI 3D image-based dose planning of brachytherapy for cervical cancer. Radiother Oncol 2009;91:187e193. [43] Tanderup K, P€ otter R, Lindegaard JC, et al. PTV margins should not be used to compensate for uncertainties in 3D image guided intracavitary brachytherapy. Radiother Oncol 2010;97: 495e500. [44] Kirisits C, Rivard MJ, Baltas D, et al. Review of clinical brachytherapy uncertainties: analysis guidelines of GEC-ESTRO and the AAPM. Radiother Oncol 2014;110:199e212. [45] Andersen CE, Damkjær SMS, Kertzscher G, et al. Fiber-coupled radioluminescence dosimetry with saturated Al2O3:C crystals: characterization in 6 and 18 MV photon beams. Radiat Meas 2011;46: 1090e1098.
-
(2017)
-
11
[46] Kertzscher G, Andersen CA, Siebert FA, et al. Identifying afterloading PDR and HDR brachytherapy errors using real-time fiber-coupled Al2O3:C dosimetry and a novel statistical error decision criterion. Radiother Oncol 2011;100:456e462. [47] Kertzscher G, Andersen CE, Edmund J, et al. Stem signal suppression in fiber-coupled Al2O3:C dosimetry for 192Ir brachytherapy. Radiat Meas 2011;46:2020e2024. [48] Kawakami S, Ishiyama H, Terazaki T, et al. Catheter displacement prior to the delivery of high-dose-rate brachytherapy in the treatment of prostate cancer patients. J Contemp Brachytherapy 2014;6:161e166. [49] Simnor T, Li S, Lowe G, et al. Justification for inter-fraction correction of catheter movement in fractionated high dose-rate brachytherapy treatment of prostate cancer. Radiother Oncol 2009; 93:253e258.
-
(2017)
-
Time-resolved in vivo dosimetry for source tracking in brachytherapy Jacob Graversen Johansen1,*, Susanne Rylander1, Simon Buus1, Lise Bentzen1, Steffen Bjerre Hokland1, Christian Skou Søndergaard1, Anders Karl Mikael With2, Gustavo Kertzscher3, Kari Tanderup1 1 Department of Oncology, Aarhus University Hospital, Aarhus C, Denmark € € Department of Medical Physics, Orebro University Hospital, Orebro, Sweden 3 Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX 2
ABSTRACT
PURPOSE: The purpose of this article is to demonstrate that brachytherapy source tracking can be realized with in vivo dosimetry. This concept could enable real-time treatment monitoring. METHODS: In vivo dosimetry was incorporated in the clinical routine during high-dose-rate prostate brachytherapy at Aarhus University Hospital. The dosimetry was performed with a radioluminescent crystal positioned in a dedicated brachytherapy needle in the prostate. The dose rate was recorded every 50e100 ms during treatment and analyzed retrospectively. The measured total delivered dose and dose rates for each dwell position with dwell times O0.7 s were compared with expected values. Furthermore, the distance between the source and dosimeter, which was derived from the measured dose rates, was compared with expected values. The measured dose rate pattern in each needle was used to determine the most likely position of the needle relative to the dosimeter. RESULTS: In total, 305 needles and 3239 dwell positions were analyzed based on 20 treatments. The measured total doses differed from the expected values by 4.7 8.4% (1SD) with range (17% to 12%). It was possible to determine needle shifts for 304 out of 305 needles. The mean radial needle shift between imaging and treatment was 0.2 1.1 mm (1SD), and the mean longitudinal shift was 0.3 2.0 mm (1SD). CONCLUSION: Time-resolved in vivo dosimetry can be used to provide geometric information about the treatment progression of afterloading brachytherapy. This information may provide a clear indication of errors and uncertainties during a treatment and, therefore, enables real-time treatment monitoring. Ó 2017 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved.
Keywords:
In vivo dosimetry; HDR brachytherapy; Source tracking; Dose rate measurement; Prostate cancer
Introduction Brachytherapy is a radiotherapy modality which involves a considerable number of manual procedures during treatment planning and delivery. Several investigations of radiation misadministrations in brachytherapy have shown that a substantial share of the events is caused by human errors (1e3). The identified misadministrations involved a Received 1 March 2017; received in revised form 11 August 2017; accepted 15 August 2017. Financial disclosure: This study was supported by an unrestricted research grant from Elekta and by Danish Cancer Society. * Corresponding author. Department of Oncology, Aarhus University Hospital, Norrebrogade 44, DK-8000, Aarhus C, Denmark. Tel.: þ45 78450000; fax: þ45 78464455. E-mail address: [email protected] (J.G. Johansen).
large range of different events including wrong connections of transfer tubes, reconstruction errors, and wrongly assigned source strength. Many events may remain unnoticed, and if detected, they are typically only identified posttreatment, due to the limited availability of commercial realtime treatment-monitoring systems. Furthermore, commercial software tools and algorithms that can provide guidance in the event of a treatment error are not available (4, 5). The development of treatment monitoring in brachytherapy has mainly focused on three approaches: imaging, tracking, and in vivo dosimetry (6). Imaging has the advantage of providing a direct link between source/applicator geometry and patient anatomy. However, extra acquisition and analysis of 3D imaging adds additional time to the treatment procedure. Furthermore, 3D imaging is typically not completely real-time, except for advanced
1538-4721/$ - see front matter Ó 2017 American Brachytherapy Society. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.brachy.2017.08.009
2
J.G. Johansen et al. / Brachytherapy
setups as for example, suites including both afterloading units and MRI (7). Fluoroscopy provides real-time information but does not include soft tissue information, and the quality of the flouroscopy images is challenged by the radiation from the treatment source, at least during high-dose-rate (HDR) treatments (8, 9). This challenge has been overcome by increasing the X-ray strength and reducing the monitoring frequency (10). Electromagnetic tracking provides high geometric accuracy but is limited by the fact that the source is not tracked directly and that it does not link the geometric verification to the patient anatomy (11e14). In vivo dosimetry facilitates direct measurement of the absorbed dose in organs at risk (OAR) or close to the tumor. However, in vivo dosimetry is challenged by large positioning uncertainties of the dosimeters, limited utilization of real-time readouts, and that each dosimeter only provides the dose at a given point. Certain dosimeter types can only measure the accumulated dose (e.g., thermoluminescence dosimeter and alanine) while others can measure dose rates (e.g., diodes, metal-oxid-semiconductor field effect transistor and scintillators) (4, 5, 15, 16). Recent developments have reduced the size of detectors and the positional uncertainty and increased the sensitivity (agreement within 5%) and read-out frequency (O10 Hz) (17e23). This article will present in vivo dosimetry measurements performed with a system featuring high read out frequency, facilitating a study of individual dwell positions. The ability to analyze the dose rates for each dwell position opens new ways of identifying errors and uncertainties. Kertzscher et al. developed an adaptive algorithm for treatment monitoring, in which the most likely dosimeter position was determined (24). This article applies a methodology where dose rate for each dwell position is used to determine the positions of the individual treatment needles relative to the dosimeter. Source tracking with multipoint dosimeter has previously been explored both in phantoms (25e29) and in pulsed dose rate cervix treatments (30), but this is the first time a single dosimeter has been used for obtaining geometrical information for each individual needle in a considerable number of patients. The ability to track the source directly enables identification of misplacements of the source relative to the dosimeter. Detectable misplacements could be, for example, needle reconstruction errors, wrong connections of guide tubes or positional offsets by the afterloader, whereas an identical shift of both source and dosimeter relative to the anatomy cannot be detected.
Methods and materials The instruments and dose measurements The in vivo dosimetry was performed using a small (0.5 0.5 2 mm3) radioluminescent crystal made of Al2O3:C grown by Landauer Inc (USA) using the Czochralski
-
(2017)
-
technique (growth CZ#60). The crystal was coupled to a 15 m long optical fiber connecting to a data acquisition box (DAQ). The DAQ consisted of a bandpass filter (395e440 nm), a photo multiplier tube (PMT) (Hamamatsu h7360-01), and a DAQ card (NI9402). The bandpass filter was placed between the PMT and the optical fiber to accept the peak wavelength region of the Al2O3:C emission and remove the other wavelengths of the contaminating Cerenkov and fluorescence light that is emitted in the optical fiber, referred to as the stem effect (19, 31, 32). The PMT was connected to the DAQ card, which counted the number of detected photons in a given time interval. The crystal and optical fiber were shielded from light using a plastic tube. The outer diameter of the crystal and tube was 1 mm, which was thin enough to fit into the brachytherapy plastic needles used for the treatment. The DAQ was placed in a neighboring radiation shielded room to avoid false signals from radiation striking the PMT. The entire system is described in details in the study by Andersen et al. (19), which also contains a schematic drawing of the system and a detailed characterization of the crystal. The clinical procedure The in vivo dosimetry procedure was carried out by the clinical staff as an integrated part of the clinical brachytherapy procedure. Patients with D’Amico high-risk prostate cancer were treated with a combination of external beam radiotherapy (46 Gy in 23 fractions) and two individual fractions of HDR brachytherapy of each 8.5 Gy prescribed to the prostate þ 3 mm (33). Before implantation of needles in the patient, pre-irradiation and calibration of the crystal were performed by the medical physicist to account for potential changes in the detection efficiency of the system, for instance due to a recoupling of the optical fiber to the DAQ-box. This process was performed using a block of solid water with two parallel needles separated by 10.4 0.2 mm and took 15e30 minutes depending on the source strength (34). Transperineal insertion of plastic needles was performed under live transrectal ultrasound guidance. After insertion of all treatment needles, an additional needle, dedicated to the dosimeter, was positioned in the prostate as best feasible but preferentially in the center of the gland. Approximately, 1 h after needle implantation, the patient underwent an axial T2W MRIs acquired with a 2-mm slice thickness and a 1.4-mm transversal accuracy. The MRI was used for treatment planning, which involved contouring, needle reconstruction, dosimeter localization, and dose optimization (Oncentra Prostate 4.2.21, Elekta). Immediately before treatment delivery, an additional MR scan was obtained to detect and adjust for needle migration O3 mm, followed by reoptimization of the plan. Immediately before dose delivery, the dosimeter was placed in the dedicated needle and inserted to the end of the lumen, with the center of the luminescent crystal located 9.5 mm from the outer tip of the needle, based on X-ray exposure
J.G. Johansen et al. / Brachytherapy
of the crystal and fiber on radiochromic film. Throughout the delivery, the dosimeter measured and recorded the dose rate every 50e100 ms as well as accumulated dose. The planned total dose to the center of the crystal was evaluated in the treatment planning system. The brachytherapy treatment was delivered using the Flexitron afterloader (Elekta) with TCC2.1.3. A third MR scan was acquired immediately after delivery to evaluate the stability of the treatment needles (35).
The retrospective analysis The retrospective analysis of the dosimetry data was performed on three different levels: (1) Accumulated dose for entire treatment, (2) Dose rates in individual dwell positions, (3) Combination of all dose rates in each individual needle. Planned doses and dose rates were calculated by the TG43 formalism (36), using the geometry of dwell positions and the dosimeter. The dose rates were integrated over the length of the crystal to account for the finite size, and the geometrical values were obtained from DICOM files from the treatment planning system based on the needle reconstruction performed by the clinical staff. The total doses were obtained by accumulating dose rates throughout the treatment (Dmeas) and the relative offset to the expected dose (DTG43) was calculated: DD/ D5(Dmeas/DTG43-1)*100%. The dosimeter system provided 10e20 dose rate measurements per second. An average dose rate (Ḋmeas) was determined for each dwell position. The first 0.4 s and the last 0.1 s of each dwell position were omitted in the averaging, due to the movement of the source and a short stabilization time needed by the crystal. The analysis was only performed for dwell positions with dwell times longer than 0.7 s to allow for averaging over at least 0.2 s. The angle and distance required by the TG43 formalism were determined based on the coordinates for each dwell position given in the treatment plan, see Fig. 1. The relative offsets between measured and expected dose rate were determined as for the total dose: DḊ/Ḋ 5 (Ḋmeas/ḊTG43 1) 100%. Δr
Δradial
radial
r
longitudinal
θ
Δlongitudinal
Fig. 1. Schematic representation of a treatment needle with five dwell positions (circles) and the needle with dosimeter (square). The coordinates used in the analysis are also shown. Free parameters are shown in blue. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
-
(2017)
-
3
The discrepancy in dose rates DḊ was assumed to be due to an offset in the distance (Dr) between source and dosimeter, see Fig. 1. Under the condition that Dr was small, the change in anisotropy, scattering, absorption as well as on energy dependence of the crystal was assumed to be negligible, and an inverse square dependency was used in the analysis. In this case, a simple relation between the relative dose rate offsets and the distances could be obtained: 2 ð1Þ DD_ D_ 5 r2 ðr þ DrÞ 1 5 r2 rmeas 1 Here rmeas refers to the most likely distance between the source and the dosimeter based on the measured dose rate. This value was determined for all the dwell positions, by isolating rmeas in equation 1 and comparing to the expected distances (r). Furthermore, any uncertainty in the relative distances would be equal to an uncertainty in Dr. This would lead to the following relation between the standard deviation (SD) of the relative dose rate offset and the uncertainty of the distances, which also includes the internal uncertainty of the system (udosimeter): 2 2 SD DD_ D_ 5 ð2SDðDrÞ=rÞ þ udosimeter 2
ð2Þ
According to this equation, the SD of the relative offset has a distance dependency. The SD of the relative offset was therefore determined in distance intervals of 4 mm and fitted to a function based on equation 2 with SD (Dr) and udosimeter as free parameters. Equation 1 only provides the distance between the source and the dosimeter and not the direction. In fact all movements are assumed to be along the r-vector, see Fig. 1, whereas the largest shift is expected to be along the longitudinal needle direction since the uncertainty of the needle tip reconstruction and source positioning in the longitudinal direction is larger than in the radial direction and also because needle migration between imaging and treatment delivery occurs in the longitudinal direction. To account for this, a more thorough geometrical analysis was performed. It was assumed that needles could only move rigidly and that all dwell positions in a given needle would undergo common shifts. The principle behind the analysis was to shift each needle, and thereby any dwell position within the needle, to best fit the measured and TG43 dose rates for all dwell positions within the needle. This was done in three steps. First, all measured dose rates in a needle were plotted as a function of their distance from the afterloader indexer or index number (black dots in Fig. 2). Then, a 1D-function of the dose rate as a function of the index number in the needle was derived based on the TG43 formalism (ḊTG43(i)) (blue lines in Fig. 2). The function uses the geometry from the treatment plan to transform the index number to a geometric position and thereafter determine an r-vector, Fig. 1. The r-vector can be resolved into a longitudinal and a radial part. If a needle includes two or more source dwell positions, it is possible to determine one longitudinal shift (Dlongitudinal) and a radial shift (Dradial), which gives the best fit between ḊTG43(i) and Ḋmeas. The function ḊTG43(i)
4
J.G. Johansen et al. / Brachytherapy
b
×10 3
100 ×10
(2017)
-
3
c/s
c/s
a
-
50
90 45
80 70
40
60 35 50 30
40
25
30 20
20 10 15 145
150
155
160
165
170
175
0
180
150
155
160
165
170
175
180
index
185 index
Fig. 2. Two examples of the needle shift fit routine. The black dots are the measured count rates for the dwell positions, the blue line is the expected curve based on the planned positions and TG43, the red curve is after the shift in needle position. Panel ‘‘a’’ shows a needle with a longitudinal shift, and panel ‘‘b’’ shows a fit with a radial shift. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
was fitted to the experimental data, with Dlongitudinal and Dradial as free parameters. The fit was performed with the ROOT software (v. 6.06 CERN) (red lines in Fig. 2).
SD of the relative offsets are given at the bottom of the table. The low dose, below 8 Gy, in six of the treatments is due to the placement of the dosimeter just outside the prostate.
Results
Dose rates for individual source positions
Measurements were performed in a total of 22 treatments in 12 patients. In two cases, the dosimeter had slid out of the needle by 17 mm and 70 mm, respectively, during the treatment. These two treatments were omitted in the analysis leaving 20 treatments analyzed. In six of the remaining 20 treatments, the dosimeter was placed outside the prostate, close to the apex, because the dosimeter needle was not inserted deep enough into the prostate.
A total of 3239 of 3518 dwell positions had a dwell time longer than 0.7 s. The dose rates and the relative distance between the source and dosimeter (r) for each dwell position were determined. The average offset in dose rate was 4.3 13.4%, and the corresponding source dwell position shift was 2.1 2.3 mm. Figure 4a shows the relative offset between measured and expected dose rate as a function of the distance between dwell position and dosimeter. Figure 4b shows the distance determined from the in vivo dosimetry measurements (rmeas), calculated using equation 1 against the expected distance based on the treatment plan (r). To investigate the distance dependencies seen in Figs. 4a and 4b, the dwell positions were grouped according to the distance to the dosimeter in 4-mm intervals starting from 10 mm. The mean and SD of both the relative dose rate offset and the distance were determined for each of the groups. The mean values are given in Figs. 4c and 4d and the SDs in Figs. 4e and 4f. The SD of the relative offsets was fitted to equation 2 as described in the Materials and methods Section. The result of the fit is given in Fig. 4e. A strong linear relation was seen between the two distances, Fig. 4d. A linear fit was made, and the result indicated a general 10% increase in all distance as seen by the fit result in Fig. 4d.
Raw data Figure 3 shows dose measurements for three treatments, one with the dosimeter located close to the apex (a þ b) and two with the dosimeter in the center of the gland (cef). The left panels (a, c, and e) show the dose rates as a function of time and the right panels (b, d, and f) show the accumulated dose as a function of time. The inserts in a, c, and e show a close up of the dose rate pattern in one needle. The black and blue curves are the expected and measured values, respectively. The source progression into each individual needles was uniquely identifiable for all needles both from dose rate and dose accumulation plots. The stability of the dose rate measurements can be seen on the flat structures in the inserts. Total delivered dose The total measured doses for all 20 treatments are shown in Table 1 along with the expected delivered doses and the relative offset between the two (three first columns). The average and
Needle shifts The method was tested on the calibration data before the use on the clinical data. The calibration was performed
J.G. Johansen et al. / Brachytherapy
b Accum. dose [Gy]
Dose rate [Gy/s]
a 0.14 80000
0.12
70000 60000 50000 40000
0.1
30000 20000 10000
0.08
0 100
120
140
160
180
200
220
240
0.06
-
(2017)
-
5
5 4 3 2
0.04 1
0.02 0 0
400
600
800
×10
3
100
80
60
40
0.2
200
400
600
800
1000 time [s ]
d
0.3 0.25
0 0
1000 time [s ]
Accum. dose [Gy]
Dose rate [Gy/s]
c
200
20
0 120
140
160
180
200
220
240
260
280
12 10 8
0.15
6
0.1
4
0.05
2
0 0
200
400
600
800
1000
1200
e
1400 time [s ]
0 0
200
400
600
800
1000
1200
f
1400 time [s ]
×10
Accum. dose [Gy]
Dose rate [Gy/s]
0.3 3
120
0.25
100
80
60
0.2
40
20
0 40
50
60
70
80
90
100
110
10 8
0.15
6
0.1
4
0.05
2
0 0
100
200
300
400
500
600
700
800
900 time [s ]
0 0
100
200
300
400
500
600
700
800
900 time [s ]
Fig. 3. The measured data provided by the in vivo dosimetry system (blue lines) and the expected values based on the treatment plan (black lines). The measurements were updated every 50 or 100 ms during the treatments. The left panel (a, c, and e) is the dose rates, and the right panel (b, d, and f) is the accumulated dose. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
with 21 dwell positions separated by 3 mm in a needle with a minimum distance to the dosimeter of 10.4 0.2 mm (34). The calibration measurements agreed within 0.05 mm both radially and longitudinally. A total of 305 needles were analyzed, and radial and longitudinal shifts were determined for all except one needle, which only had one dwell position longer than 0.7 s. The average shift per patient is shown in Table 1, and a scatter plot of the radial (X) and longitudinal (Y) shifts for all needles in all the 20 treatments are given in Fig. 5. The average shifts for all needles were determined using a Gaussian fit (right panel in Fig. 5): Dradial 5 0.19 1.1 mm, Dlongitudinal 5 0.34 2.0 mm. The total displacement was Dr 5 (Dradial2þ Dlongitudinal2)½ 5 0.39 1.8 mm. The last row in Table 1 shows the overall average and the SD of the mean values across all treatments. The mean
shift across all needles within a given treatment can be interpreted as a potential shift (or reconstruction uncertainty) of the dosimeter. The average of the mean values is therefore an indicator of any general offset in the dosimeter position, whereas the SD of the mean values indicates the reproducibility of the dosimeter position. The values in the parentheses are the average and SD of the SDs determined for each treatment. The SDs of the needle shifts in a given treatment can be interpreted as the positional uncertainty of the treatment needles relative to the dosimeter. Figure 6 shows similar plots as in Figs. 4 and 5 for three selected treatments. The treatments represent the following scenarios: top: a treatment with stable geometry corresponding to needle shifts !3 mm and small dosimeter offset; middle: a treatment with offsets which can be explained by a slight systematic shift of the dosimeter of
6
J.G. Johansen et al. / Brachytherapy
-
(2017)
-
Table 1 Results from each of the 20 treatments Treatment
Dmeas (Gy)
DTG43 (Gy)
DD/DTG43 (%)
DRadial (mm)
DLongitudinal (mm)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Average SD
4.95 3.68 10.4 3.90 10.7 9.62 11.2 10.3 9.68 10.4 10.8 11.6 10.4 8.95 6.74 9.31 3.69 9.56 10.1 4.39
5.64 4.43 10.7 4.11 10.4 10.2 10.9 11.4 9.90 11.5 9.80 12.7 11.0 10.0 7.76 10.9 3.29 10.2 10.9 4.01
12.1 16.9 2.71 5.26 3.08 6.04 3.07 9.56 2.21 10.1 10.4 9.10 5.37 10.6 13.2 14.3 12.3 6.70 7.53 9.54 4.66 8.40
0.15 0.56 1.3 0.96 0.27 0.4 0.06 1.2 0.23 0.38 0.17 0.50 0.54 0.47 0.97 0.72 0.08 0.47 0.95 0.62 0.25 0.57 0.64 0.53 0.84 0.45 1.3 0.80 0.58 0.50 0.37 0.74 0.09 0.53 0.68 0.83 1.2 1.1 1.6 0.87 0.17 0.77 (0.67 0.24)
1.9 1.3 2.6 2.1 1.5 1.6 0.84 1.9 0.79 0.96 1.9 0.95 0.03 0.71 0.62 0.87 0.93 1.5 0.64 0.62 0.19 1.2 0.02 1.3 0.38 1.1 1.1 1.1 1.3 1.9 1.8 2.0 0.91 1.7 0.82 2.2 2.6 2.1 0.01 0.97 0.31 1.3 (1.40 0.50)
The first two columns are the measured and expected accumulated doses respectively, and the third column is the offset between the two. The last two columns are the mean and SD of the radial and longitudinal needle shifts in the treatment. The last row marked with bold letters shows the average and the standard deviations (SDs) for the mean values of the individual treatments. The values in the parentheses are the average and SD of the standard deviations of the treatments.
2 mm; and bottom: a treatment where two needles show specific larger offsets corresponding to 4 mm radial and 9 mm longitudinal shifts, respectively. The treatments are treatment 7, 6, and 2 respectively. Discussion This article presents in vivo dose rate measurements for 3239 dwell positions in 305 needles from 20 HDR treatments of prostate cancer patients. The dose rate measurements were analyzed both with regard to the total dose as well as according to dose rates from individual source positions. By exploiting the offsets between measured and expected dose rates, it was demonstrated that the dose rate information could be transformed into geometrical information of the source progression throughout the treatment. The total average offset of accumulated dose was 4.7% across the treatments with an 8.4% spread (1SD). Previous studies (22, 37e40) reporting on in vivo dosimetry in HDR brachytherapy have in general reported similar or higher offsets with typical ranges around 20% and up to 40% (around 10% SD), which is most often attributed to positional inaccuracies of the dosimeter. In a study by Mason et al. (41), the dosimeter (metal-oxid-semiconductor field effect transistor) was positioned in a needle, and the mean offset was 3.0% with a maximum of 11.2% similar to this study. One patient treatment in our study had a nominal offset in absolute dose above 15%, which is more than expected. This offset was identified as stemming from two needles with considerable offsets in the dose rates (bottom row in Fig. 6).
One of the advantages of dose rate measurements compared to total dose measurements is the possibility of identifying specifically which needles are related to geometrical offsets. The dose rate measurements presented in Fig. 6 show that offsets can occur if some treatment needles are shifted relative to the reconstructed position (bottom row in Fig. 6) or if the dosimeter has shifted (middle row in Fig. 6). The offsets in dose rates of the two treatments show significantly different patterns, which allows for a separation of the two types of events. With the ‘‘Adaptive Error Detection Algorithm’’ (AEDA) proposed by Kertzscher et al. (24), this strategy is further exploited. The reconstruction of the needles based on radial and longitudinal shifts, lead to an overall 1.8 mm uncertainty (SD). The main uncertainty was along the needle, which can be explained by larger reconstruction and source positioning uncertainties in the longitudinal direction than in lateral and anterioreposterior directions (42, 43). Reconstruction uncertainties are affecting positional uncertainties for both treatment needles and the dosimeter, and the contribution of each can be determined by looking at the individual treatments, Table 1. The mean shift across all needles in a single treatment is an indication of a shift in the dosimeter position, whereas the SD represents the uncertainty in the position of treatment needles. The SD of the mean shifts from all treatments was interpreted as an uncertainty in the positioning of the dosimeter and was determined to 0.77 mm (radial) and 1.3 mm (longitudinal). The average of the SDs across all treatments was on the other hand interpreted as the
J.G. Johansen et al. / Brachytherapy
10
rmeas[mm]
12
100
. .
(2017)
-
7
b
ΔD/D [%]
a
-
16
60
14
50
12
50
40
10
8 0 −50
8
30
6 4
20
2
10
6 4 2
−100 0
. .
20
30
40
50
60 r [mm]
0 0
0
d
0
Mean of r meas[mm]
Mean of ΔD/D [%]
c
10
−1 −2 −3 −4
10
20
30
40
50
60 r [mm]
0
60
y = 1.1r-0.75mm
50 40 30
−5
20
−6 10
−7
SD of ΔD/D [%]
e . .
10
20
30
40
50
0 0
60 r [mm]
f
25
y2 = (2.1.14mm/r)2 +7.2%2
20 15
SD of r meas[mm]
−8 0
1
30
40
50
60 r [mm]
40
50
60 r [mm]
10
20
30
40
50
60 r [mm]
3
5
20
30
4
2
10
20
5
10
0 0
10
0 0
Fig. 4. Panels ‘‘a’’ and ‘‘b’’ show scatter plots of the dose rate deviations and the recalculated distance as function of the expected distance respectively. The color bars indicate the number of dwell position within the pixel. The last four panels show the mean (c and d) and standard deviations (e and f) for given distance intervals between 10 mm and 50 mm. The panels ‘‘c’’ and ‘‘e’’ are for the dose rate offsets, and ‘‘d’’ and ‘‘f’’ are for the recalculated distances. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
positional uncertainty for the treatment needles. These were determined to 0.67 mm (radial) and 1.4 mm (longitudinal), Table 1. This magnitude of uncertainties is in good agreement with the 0.6 mm and 1.2 mm radial and longitudinal reconstruction uncertainties found by Haack et al. (42). The study of all dwell positions (Fig. 4) showed that the offsets in general could be explained by a general 7.2% uncertainty of the system overlaid with a 1.14 mm uncertainty (1SD) in the relative geometry (Fig. 4e). These observations are well in agreement with what is
expected from typical reconstruction uncertainties in MRI-based brachytherapy (44) and with the expected uncertainty of the dosimetry system. The dosimeter has previously been tested in pulsed dose rate cervix treatments (45, 46). Andersen et al. reported an 8% uncertainty for the dosimeter system, which includes reproducibility, energy response, angular dependence, stem effect, transmission, and calibration (19). The stem effect may be significant in configurations where the source is close to the optical fiber but far from the crystal. However, with our configuration, the stem effect is !1% for all dwell
N/0.2mm
J.G. Johansen et al. / Brachytherapy Δ Longitudinal [mm]
8
10
-
(2017)
-
30 25 20 15 10
5
5 0 −10
−8
−6
−4
−2
0
2
4
6
Δ Radial [mm]
−8
−6
−4
−2
0
2
4
6
Δ Longitudinal [mm]
N/0.2mm
0
−5
8
10
8
10
20 15 10
−10
5
−10
−5
0
5
10 Δ Radial [mm]
0 −10
Fig. 5. A representation of the radial and longitudinal shifts determined for all 304 needles. The left panel presents the longitudinal shift vs. the radial shift for all the 304 needles. The right panel is the individual histogram for the radial (top) and longitudinal (bottom) shifts. The red lines are the Gaussian fits. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
positions above the dosimeter, !2% for all positions within 3 cm below the dosimeter, and with larger separations than 1 cm to the fiber (47). Only for specific positions further than 3 cm away and with a distance to the fiber of 0.5e1 cm, the stem effect uncertainty may reach 5% (47). Using the uncertainty budget of Andersen et al. while reducing stem effect to 1% and correcting for energy dependence, the uncertainty (SD) of the system (udosimeter) reaches 5%, making it the dominating uncertainty compared to the uncertainties of count statistics. Furthermore, the systematic 4.7% measured underdosage could be explained by a 10% increase of all distances in the treatment or a systematic deviation in calibration. A contribution to the increase could be a swelling of the prostate. Buus et al. reported a 10% increase in the prostate volume between two fractions (33), which indicated the presence of swelling. Thorough phantom measurements to study the calibration and a study of the swelling of the prostate during a brachytherapy treatment are required to understand their relative contributions to the systematic offset. Figure 4f shows that the SDs of the dose rate offsets are transformed into a 2e3 mm SD in the distances. This is significantly larger than the 1.14 mm uncertainty in Dr found in the fit of Fig. 4e. The reason is that all dose rate uncertainties are assigned to the positional uncertainty (equation 1), and therefore, the 5% system uncertainty will lead to slightly overestimated positional shifts. This plays a more significant role at larger distances. For example, 5% dosimetric uncertainty equals a positional uncertainty of 0.5 mm at a 20 mm and 1 mm uncertainty at a 40 mm distance (equation 2). A reduction of the system uncertainty would therefore improve the geometrical precision, for example by improving the signal output, which will reduce the effect of background and stem signal.
Limitations Although the transformation from dose and into geometry is promising for real-time monitoring of brachytherapy treatments, it also has limitations. The most important limitation is that geometrical shifts are assessed relative to the dosimeter and not relative to the patient anatomy. Drifts of the entire template would not be detectable with this system, as it will lead to a common drift of both the dosimeter and the treatment needles and no change in relative distance. It is well known that potential drift of the entire implant is a significant challenge in prostate HDR (35, 48, 49). A precise determination of the dosimeter relative to the patient anatomy would therefore improve the potential of tracking with in vivo dosimetry. One way to overcome this problem could be to combine in vivo dosimetry with fluoroscopy. Secondly, the method only provides information about the size of the relative offset, hence the determined offset could stem from an offset in the dwell position, an offset in the dosimeter position, or a combination. For instance, 2 mm offsets caused by needle drifts, reconstruction errors, or by the afterloader cannot be distinguished. Finally, the internal uncertainty of the dosimeter has already been discussed, and it was shown that it may transform into up to ~1 mm geometrical uncertainty. Perspective Despite the limitations, in vivo dosimetry with subsecond dwell time read-out has a strong potential for realtime treatment monitoring. This article presents a retrospective study of the dose rates, but all the information shown in Fig. 6 can be generated in real time. In this way, any positional offsets of the source could be presented to the clinical staff within seconds after it occurs,
100
50
-
(2017)
60
-
4
10
50 40
2
9
Δ Longitudinal [mm]
. . Δ D/D [%]
rMEAS[mm]
J.G. Johansen et al. / Brachytherapy
0
30
−50
20
2
5
0
−5
10 −100
−10 20
30
40
50 60 r [mm]
0
. . Δ D/D [%]
2
100
0 0
10
20
30
40
50
60 r [mm]
60
0
4
40
0
30
1
2
20
−50
−10
−5
0
5 10 Δ Radial [mm]
−10
−5
0
5 10 Δ Radial [mm]
−10
−5
0
5 10 ΔRadial [mm]
10
50
50
ΔLongitudinal [mm]
10
rMEAS [mm]
0
5
0
−5
10 −100
−10 20
30
40
50 60 r [mm]
0
. . ΔD/D [%]
2
100
50
0 0
10
20
30
40
50
60 r [mm]
60
0
2
10
50 40
0
1
30
1
20
−50
ΔLongitudinal [mm]
10
rMEAS [mm]
0
5
0
−5
10 −100 0
−10 10
20
30
40
50 60 r [mm]
0
0 0
10
20
30
40
50
60 r [mm]
0
Fig. 6. Results from three selected treatments (7, 6, and 2 from Table 1). The left panels are the relative offsets in dose rates. The middle panels are the distance determined using the dose rates. The right panels are scatter plots of the radial and longitudinal needle shifts. The color bars indicate the number of dwell position within the pixel. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
and the treatment could be stopped for further investigation in case of unacceptably large offsets. This would require an implementation of a real-time dosimetry protocol in the clinical workflow. The protocol should include acceptable limits for the determined offsets. The limits should take into account that the determined offsets could include contributions from both offsets in source and dosimeter. For instance, by having both a maximum acceptable offset and a maximum on the spread in the offsets, similar to what is described in the study by Kertzscher et al. (24). In addition, a way to better fasten the dosimeter to the needle is required, to avoid drifting of the dosimeter. A further perspective is to iteratively reconstruct the needle geometry based on the dwell positions determined from the in vivo dosimetry measurements. Based on such a reconstruction, a delivered dose distribution could be generated; hence expanding the use of in vivo dosimetry
from the evaluation of a point dose to the evaluation of a 3D dose distribution. Conclusion Time-resolved in vivo dosimetry could be used to provide geometric information about the treatment progression of remote afterloading brachytherapy. Such source tracking has the potential to provide a clear indication of errors and uncertainties during a treatment, hence enabling real-time treatment monitoring using in vivo dosimetry. Acknowledgments We wish to thank Dr. M.S. Akselrod, Stillwater Crystal Growth Division, Landauer Inc, USA for the Al2O3:C crystals. This study was supported by an unrestricted research grant from Elekta and by Danish Cancer Society.
10
J.G. Johansen et al. / Brachytherapy
References [1] ICRP. Prevention of high-dose-rate brachytherapy accidents. ICRP Publication. Ann ICRP97 2005;35:1e51. [2] IAEA. Lessons learned from accidental exposures in radiotherapy. IAEA Safety Report Series 17. Vienna: IAEA; 2000. [3] Richardson S. A 2-year review of recent Nuclear Regulatory Commission events: what errors occur in the modern brachytherapy era? Pract Radiat Oncol 2012;2:157e163. [4] Kertzscher G, Rosenfeld A, Beddar S, et al. In vivo dosimetry: trends and prospects for brachytherapy. Br J Radiol 2014;87:20140206. [5] Tanderup K, Beddar S, Andersen CE, et al. In vivo dosimetry in brachytherapy. Med Phys 2013;40:070902. [6] Bert C, Kellermeier M, Tanderup K. Electromagnetic tracking for treatment verification in interstitial brachytherapy. J Contemp Brachytherapy 2016;8:448e453. [7] de Leeuw H, Moerland MA, van Vulpen M, et al. A dual-plane co-RASOR technique for accurate and rapid tracking and position verification of an Ir-192 source for single fraction HDR brachytherapy. Phys Med Biol 2013;58:7829e7839. [8] Sheikh-Bagheri D, Munro P. A Monte Carlo study of verification imaging in high dose rate brachytherapy. Med Phys 1998;25: 404e414. [9] van’t Riet A, Mak ACA, te Loo HJ, et al. Verification of source position in interstitial and intraluminal brachytherapy. Radiother Oncol 1994;33:167e170. [10] Nose T, Chatani M, Otani Y, et al. Real-time verification of a highdose-rate Ir-192 source position using a modified C-arm fluoroscope. Int J Radiat Oncol Biol Phys 2017;97:858e865. [11] Bharat S, Kung C, Dehghan E, et al. Electromagnetic tracking for catheter reconstruction in ultrasound-guided high-dose-rate brachytherapy of the prostate. Brachytherapy 2014;13:640e650. [12] Poulin E, Racine E, Binnekamp D, et al. Fast, automatic, and accurate catheter reconstruction in HDR brachytherapy using an electromagnetic 3D tracking system. Med Phys 2015;42:1227. [13] Boutaleb S, Racine E, Fillion O, et al. Performance and suitability assessment of a real-time 3D electromagnetic needle tracking system for interstitial brachytherapy. J Contemp Brachytherapy 2015;7:280e289. [14] Beaulieu L, Racine E, Han DY, et al. Real-time electromagnetic trackingebased treatment platform for high-dose-rate prostate brachytherapy: clinical workflows and end-to-end validation. Brachytherapy 2017. http://dx.doi.org/10.1016/j.brachy.2017.04.247 [Epub ahead of print]. [15] Therriault-Proulx F, Beaulieu L, Beddar S. Validation of plastic scintillation detectors for applications in low-dose-rate brachytherapy. Brachytherapy 2017;16:903e909. [16] Van Gellekom MPR, Canters RAM, Dankers FJWM, et al. In vivo dosimetry in gynecological applicationsdA feasibility study. Brachytherapy 2017. http://dx.doi.org/10.1016/j.brachy.2017.04.240 [Epub ahead of print]. [17] Kertzscher G, Beddar S. Ruby-based inorganic scintillation detectors for 192Ir brachytherapy. Phys Med Biol 2016;61:7744e7764. [18] Lambert J, Yin Y, McKenzie DR, et al. A prototype scintillation dosimeter customized for small and dynamic megavoltage radiation fields. Phys Med Biol 2010;55:1115e1126. [19] Andersen CE, Nielsen SK, Greilich S, et al. Characterization of a fiber-coupled Al2O3:C luminescence dosimetry system for online in vivo dose verification during 192Ir brachytherapy. Med Phys 2009;36:708e718. [20] Therriault-Proulx F, Archambault L, Beaulieu L, et al. Development of a novel multi-point plastic scintillation detector with a single optical transmission line for radiation dose measurement. Phys Med Biol 2012;57:7147e7159. [21] Cherpak AJ, Cygler JE, Choan E, et al. Real-time measurement of urethral dose and position during permanent seed implantation for prostate brachytherapy. Brachytherapy 2013;13:169e177.
-
(2017)
-
[22] Carrara M, Tenconi C, Rossi G, et al. Rectal wall measurements during HDR prostate brachytherapy with MO dosimeters integrated on a trans-rectal US probe: Comparison with planned and reconstructed doses. Radiother Oncol 2016;118:148e153. [23] Carrara M, Romanyukha A, Tenconi C, et al. Clinical application of MOSkin dosimeters to rectal wall in vivo dosimetry in gynecological HDR brachytherapy. Phys Med 2017;41:5e12. [24] Kertzscher G, Andersen CE, Tanderup K. Adaptive error detection for HDR/PDR brachytherapy: guidance for decision making during real-time in vivo point dosimetry. Med Phys 2014;41:052102. [25] Therriault-Proulx F, Beddar S, Beaulieu L. On the use of a single-fiber multipoint plastic scintillation detector for 192Ir high-dose-rate brachytherapy. Med Phys 2013;40:062101. [26] Nakano T, Suchowerska N, McKenzie DR, et al. Real-time verification of HDR brachytherapy source location: implementation of detector redundancy. Phys Med Biol 2005;50:319e327. [27] Smith RL, Haworth A, Panettieri V, et al. A method for verification of treatment delivery in HDR prostate brachytherapy using a flat panel detector for both imaging and source tracking. Med Phys 2016;43:2435e2442. [28] Guiral P, Ribouton J, Jalade P, et al. Design and testing of a phantom and instrumented gynecological applicator based on GaN dosimeter for use in high dose rate brachytherapy quality assurance. Med Phys 2016;43:5240e5251. [29] Fonseca G, Podesta M, Bellezzo M, et al. Online pretreatment verification of high-dose rate brachytherapy using an imaging panel. Phys Med Biol 2017;62:5449e5461. [30] Tanderup K, Christensen JJ, Granfeldt J, et al. Geometric stability of intracavitary pulsed dose rate brachytherapy monitored by in vivo rectal dosimetry. Radiother Oncol 2006;79:87e93. [31] Beddar AS, Mackie TR, Attix FH. Cerenkov light generated in optical fibres and other light pipes irradiated by electron beams. Phys Med Biol 1992;37:925e935. [32] de Boer SF, Beddar AS, Rawlinson JA. Optical filtering and spectral measurements of radiation-induced light in plastic scintillation dosimetry. Med Phys 1999;38:945e958. [33] Buus S, Rylander S, Hokland S, et al. Learning curve of MRI-based planning for high-dose-rate brachytherapy for prostate cancer. Brachytherapy 2016;15:426e434. [34] Kertzscher G, Andersen CE, Tanderup K. Fast, simple, and accurate: in vivo dosimeter calibration for HDR/PDR brachytherapy. Int J Radiat Oncol Biol Phys 2014;90:S742. [35] Buus S, Lizondo SM, Hokland S, et al. Needle migration and dosimetric impact in high dose rate brachytherapy for prostate cancer evaluated by repeated MRI. Brachytherapy 2017;. [this issue]. [36] Nath R, Anderson LL, Luxton G, et al. Dosimetry of interstitial brachytherapy sources: recommendations of the AAPM radiation Therapy Committee Task Group No. 43. Med Phys 1995;22:209e234. [37] Nose T, Koizumi M, Yoshida K, et al. In vivo dosimetry of high-doserate interstitial brachytherapy in the pelvic region: use of a radiophotoluminescence glass dosimeter for measurement of 1004 points in 66 patients with pelvic malignancy. Int J Radiat Oncol Biol Phys 2008; 70:626e633. [38] Anagnostopoulos G, Baltas D, Geretschlaeger A, et al. In-vivo thermoluminescence dosimetry dose verification of transperineal 192Ir high-dose rate brachytherapy using CT-based planning for the treatment of prostate cancer. Int J Radiat Oncol Biol Phys 2003;57:1183e1191. [39] Suchowerska N, Jackson M, Lambert J, et al. Clinical trials of a urethral dose measurement system in brachytherapy using scintillation detectors. Int J Radiat Oncol Biol Phys 2011;79:609e615. [40] Seymour EL, Downes SJ, Fogarty GB, et al. In vivo real-time dosimetric verification in high dose rate prostate brachytherapy. Med Phys 2011;38:4785e4794. [41] Mason J, Mamo A, Al-Qaisieh B, et al. Real-time in vivo dosimetry in high dose rate prostate brachytherapy. Radiother Oncol 2016;120: 333e338.
J.G. Johansen et al. / Brachytherapy [42] Haack S, Nielsen SK, Lindegaard JC, et al. Applicator reconstruction in MRI 3D image-based dose planning of brachytherapy for cervical cancer. Radiother Oncol 2009;91:187e193. [43] Tanderup K, P€ otter R, Lindegaard JC, et al. PTV margins should not be used to compensate for uncertainties in 3D image guided intracavitary brachytherapy. Radiother Oncol 2010;97: 495e500. [44] Kirisits C, Rivard MJ, Baltas D, et al. Review of clinical brachytherapy uncertainties: analysis guidelines of GEC-ESTRO and the AAPM. Radiother Oncol 2014;110:199e212. [45] Andersen CE, Damkjær SMS, Kertzscher G, et al. Fiber-coupled radioluminescence dosimetry with saturated Al2O3:C crystals: characterization in 6 and 18 MV photon beams. Radiat Meas 2011;46: 1090e1098.
-
(2017)
-
11
[46] Kertzscher G, Andersen CA, Siebert FA, et al. Identifying afterloading PDR and HDR brachytherapy errors using real-time fiber-coupled Al2O3:C dosimetry and a novel statistical error decision criterion. Radiother Oncol 2011;100:456e462. [47] Kertzscher G, Andersen CE, Edmund J, et al. Stem signal suppression in fiber-coupled Al2O3:C dosimetry for 192Ir brachytherapy. Radiat Meas 2011;46:2020e2024. [48] Kawakami S, Ishiyama H, Terazaki T, et al. Catheter displacement prior to the delivery of high-dose-rate brachytherapy in the treatment of prostate cancer patients. J Contemp Brachytherapy 2014;6:161e166. [49] Simnor T, Li S, Lowe G, et al. Justification for inter-fraction correction of catheter movement in fractionated high dose-rate brachytherapy treatment of prostate cancer. Radiother Oncol 2009; 93:253e258.