Australas Phys Eng Sci Med (2013) 36:487–494 DOI 10.1007/s13246-013-0232-y
TECHNICAL PAPER
Developing a novel method to analyse Gafchromic EBT2 films in intensity modulated radiation therapy quality assurance Yunfei Hu • Yang Wang • Gerald Fogarty Guilin Liu
•
Received: 14 May 2013 / Accepted: 20 November 2013 / Published online: 30 November 2013 Ó Australasian College of Physical Scientists and Engineers in Medicine 2013
Abstract Recently individual intensity modulated radiation therapy quality assurances (IMRT QA) have been more and more performed with GafchromicTM EBT series films processed in red–green–blue (R–G–B) channel due to their extremely high spatial resolution. However, the efficiency of this method is relatively low, as for each box of film, a calibration curve must be established prior to the film being used for measurement. In this study, the authors find a novel method to process the GafchromicTM EBT series, that is, to use the 16-bit greyscale channel to process the exposed film rather than the conventional 48-bit R–G– B channel, which greatly increases the efficiency and even accuracy of the whole IMRT procedure. The main advantage is that when processed in greyscale channel, the GafchromicTM EBT2 films exhibits a linear relationship between the net pixel value and the dose delivered. This linear relationship firstly reduces the error in calibrationcurve fitting, and secondly saves the need of establishing a calibration curve for each box of films if it is only to be used for relative measurements. Clinical testing for this novel method was carried out in two radiation therapy centres that involved a total of 743 IMRT cases, and 740 cases passed the 3 mm 3 % gamma analysis criteria. The cases were also tested with small ionization chambers (cc13) and the results were convincing. Consequently the authors recommend the use of this novel method to
Y. Hu (&) Y. Wang Radiation Therapy Department, Genesis Cancer Care, Level A St Vincent’s Clinic, Darlinghurst, NSW 2010, Australia e-mail: [email protected] G. Fogarty G. Liu Radiation Therapy Department, Mater Hospital, 25 Rockland Rd, Crows Nest, NSW 2065, Australia
improve the accuracy and efficiency of individual IMRT QA procedure using Gafchromic EBT2 films. Keywords Gafchromic film Intensity modulated radiation therapy Dosimetry Greyscale Quality assurance Clinical testing
Introduction Radiochromic films are useful for radiotherapy quality assurance (QA) due to their high spatial resolution, low energy dependence in a variety of beam qualities, and near tissue equivalence, which are especially important for dose measurements with small segments and sharp gradients [1]. One good example of film applications is its use in individual plan verification for intensity modulated radiation therapy (IMRT) QA [5–9]. Gafchromic EBT film was initially designed to replace silver halide radiographic film for IMRT QA procedures [2]. However, it was soon reported that the EBT film had a relatively high measurement uncertainty, and this measurement uncertainty was mostly attributed to the nonuniformity of the sensitive layer of the film [3, 4]. To improve the uniformity, the manufacturer added a yellow dye to the sensitive layer, leading to the invention of the GafchromicTM EBT2 film. When analysing dose measurements using EBT2 film, the red channel, extracted from the red–green–blue (R–G– B) scanned images, is most often used, because it has the highest absorption and sensitivity (defined as the ratio of the change in optical density and the amount of dose) [10]. Historically the quantity used to analyse GafchromicTM film was mostly the optical density (OD). While the OD as a quantity of choice has been more or less inherited from
123
488
radiographic films, there have been attempts to use a sole pixel value (PV) from a document scanner. Compared to the OD, the use of PV (or raw data) always gives more pronounced non-linear behaviour, which will in turn lead to a higher uncertainty dur to the larger fitting error in establishing a calibration curve [1]. To establish a calibration curve and use the results for future measurements, the common way is to plot dose as a function of the measured net OD and then fit the data with an appropriate function [11]. For every particular radiochromic film dosimetry system, the sensitivity curve will be different [12]. If a linear response can be established between the PV and the dose, i.e., Dose = aPV ? b, where a and b are two measurable constants, the accuracy and efficiency of calibrating GafchromicTM EBT film can be significantly improved. The conventional R–G–B film analysis method requires an individual calibration curve to be established for every new box of Gafchromic film due to the film’s batch variation, and to establish an accurate polynomial fitting curve, a large number of data points need to be collected. On the other hand, when the calibration curve is linear, only a start and an end point are needed to establish an accurately fitting calibration curve, with a significantly reduced fitting error. With the introduction of the net PV, which is defined by subtracting the raw PV value with the background PV from an unexposed film piece, the b constant in the equation becomes zero, and the final equation between the Dose and the net PV now becomes Dose = anet PV. With this known linear relationship that goes through the origin, the PV can be directly compared to the dose value through a normalisation function (which determines the value for a), and there is no longer any need to establish a calibration curve for each box of film. The main advantages of establishing such a relationship include improved efficiency (only one normalisation point is needed instead of multiple calibration points) and accuracy (simpler fitting function), which are particularly important for individual IMRT QA. There have been a few studies working on establishing a linear function between dose and PV, mostly using red channel [11, 14, 16]. Todorovic et al. found that when the film was scanned in as a 14-bit greyscale image rather than the 42-bit RGB image, the resultant calibration curve was more linear [14]. Recently Devic et al. established a functional argument to linearize the inherently non-linear response of the Gafchromic EBT2 film, and was proved successful, but this argument was based on net OD instead of PV [16]. In this study, based on a number of experiments and clinical studies, the authors found that when the GafchromicTM EBT2 film is scanned by the greyscale channel of a flatbed scanner, a linear and sensitive PV-dose calibration
123
Australas Phys Eng Sci Med (2013) 36:487–494
Fig. 1 Set up EBT2 film for calibration. Film is aligned such that its centre part is located at the centre of the reference field. Build-up solid water is to be added after the film is aligned
curve can be established within a clinically useful dose range. If the net PV is used, the relationship between dose and net PV goes through the origin, and after normalisation, the net PV can be directly compared to the dose for evaluation. This method of analysing the dose measurements using GafchromicTM EBT2 film for individual IMRT QA has been clinically tested with a total of 743 IMRT patients’ plans with different anatomical sites at two clinics, and it was found that the criteria of 3 %/3 mm for an IMRT QA plan are passing almost 100 % gamma function points.
Materials and methods Calibrating EBT2 film using greyscale To calibrate the EBT2 film, three pieces of films from the same batch, each with a dimension of 254 mm 9 203 mm, were used. Each piece was sliced into twelve strips, all with a dimension of 20 mm 9 203 mm. The films were irradiated in solid water phantom under a linear accelerator (Siemens Artiste M5298). Each strip was irradiated independently with a known dose, as is shown in Fig. 1. Three pieces of films were used because the authors planned to calibrate the EBT2 film in three different clinically-useful dose ranges so as to determine the film’s useful dynamic dose ranges. Film 1 was irradiated with 0 cGy (background) to 55 cGy in an equivalent step of 5 cGy; film 2 was irradiated with 0 cGy to 550 cGy in an equivalent step of 50 cGy; and film 3 was irradiated with 0 cGy to 1,650 cGy in an equivalent step of 150 cGy. The Monitor Units (MUs) required to deliver the planned dose was calculated by CMS XiO radiation therapy planning system with the phantoms scanned into the system by a Computed Tomography (CT) scanner.
Australas Phys Eng Sci Med (2013) 36:487–494
489
Fig. 2 Example of a calibration film composed of strips irradiated with different doses; the amount of the dose delivered is marked at the bottom of each strip. The film can be analysed using either the conventional RGB channels or the novel Greyscale channel
Scanning and analysing the film Once all strips from the same piece of film were irradiated with the planned dose, the authors carefully reconstructed and taped together the original, complete sheet of film, as is shown in Fig. 2. The films were scanned using an EPSON Expression 10000XL flatbed scanner (EPSON Seiko Epson Corporation, Suwa, Japan) in both 48-bit colour and 16-bit greyscale format. The 48-bit colour image was split into the 16-bit red, green and blue channels. The scanned images were saved in JPEG format. TIFF format was the other option, but compared with JPEG, TIFF images has a much larger file size, which gives difficulties in image analysis and file storage for frequent clinical uses in the future. The disadvantage of JPEG is that as it does not support 16-bit images, when 16-bit image is saved it is automatically converted down to 8 bit. The authors are fully aware of this problem but for the purpose of this paper and potential clinical uses the 8-bit image data are considered adequate, so the scanned images are saved in JPEG format and used for analysis. To study the effect of film-self-developing [15] after irradiation on the accuracy of the result, films were scanned on the spot of irradiation (0 h), then 0.5, 1, 2, 12, 24 and 48 h after irradiation. The authors strictly controlled the consistency of other scan settings which may affect the final result, i.e. orientation, scan position, colour adjustments, brightness and contrast, and resolution. Software used to read the PV values from the calibration films was software developed by one of the authors (A. Prof Yang Wang) and approved by Australian Therapeutic
Fig. 3 Cylindrical PMMA phantom used for IMRT QA with GafchromicTM EBT2 film
Goods Administration (TGA), called Radiation Oncology Dosimetry Management System (RODOMS). The PV readings were taken at the centre of each film strip. Five readings were taken, averaged and subtracted by the background reading. The corrected PV reading was then plotted against dose in Excel to establish the calibration curve. Calibration curves from different channels (grey, red, green and blue) were compared and analysed. Clinical testing The novel method: greyscale channel was used to analyse the EBT2 film in IMRT QA, was then clinically tested. To
123
490
Fig. 4 Two supporting feet and upper half and lower half of the cylindrical phantom
simulate real patient’s geometry, a home-designed cylindrical PMMA phantom was used, as is shown in Figs. 3 and 4. The phantom is composed of two supporting feet, the upper half of the cylindrical and the lower half of the cylindrical. During QA, film sits in between the two halves horizontally. Before irradiation, the plane where the film sits is aligned at the iso-centre against the optical crosshair. The length and the diameter of the cylindrical is 30.0 and 24.0 cm separately. One piece of EBT2 film was irradiated in the phantom with the individual IMRT plan, scanned with exactly the same settings as the calibration film, and then compared to the dose map exported from the planning system in RODOMS. The film analysing part of RODOMS allowed the authors to compare the film in terms of horizontal/vertical profiles, iso-dose lines, point-to-point doses, and gamma function with different tolerances. The film analysing part of RODOMS software was designed based on the fact that when the Gafchromic EBT2 film is scanned in the greyscale channel, the resultant pixel value is linearly proportional to the dose (Dose = aPV ? b). Before direct comparison, three corrections need to be applied to the greyscale film: rescaling and co-registration, background subtraction, and normalisation. Rescaling and co-registration aims to rescale the scanned greyscale image to the same size of the planning dose map, and then coregister them so that the corresponding points are in the same coordinates. Then background subtraction is achieved by firstly choosing a background point in the unexposed area of the film, then the PV of this point subtracted from all the data points on the greyscale film to obtain the net PV so that now the relationship between dose and PV becomes: Dose = anet PV. Finally normalisation is performed by choosing a normalisation point in a uniform dose area inside the PTV; a normalisation factor (a) is
123
Australas Phys Eng Sci Med (2013) 36:487–494
calculated by comparing the PV of this normalisation point to the PV of the point at the same coordinate in the dose map. This normalisation value is then applied to all the data points on the greyscale film. These functions are only valid if PV is linearly proportional to the dose, and no extra calibration curve is required. The disadvantage of using a single-point background subtraction and normalisation correction functions is that the measurement error of the chosen background/normalisation point will be inherited by all the other points. This error can be reduced via: (a) carefully choosing the background/normalisation point which locates in a uniform dose area; and (b) use multiple correction points to reduce the error. After all correction functions are applied, the greyscale film is compared to the planning output in terms of: point dose comparison, iso-dose line matching, dose profile comparison and gamma-value function assessment. The 3 mm/3 % gamma function assessment is used as the major criteria for QA results. The clinical testing was carried out at the radiation therapy department at St Vincent’s Clinic (Darlinghurst, NSW, Australia) and Mater Hospital (Crows Nest, NSW, Australia). At St Vincent’s Clinic in total 215 IMRT patients’ plans were tested, including 95 Brain, Head and Neck cases, 111 Pelvis and Prostate cases, and 9 Lung and Breast cases. At Mater Hospital in total 528 IMRT patients’ plans were tested, including 118 Brain, Head and Neck cases, 356 Pelvis and Prostate cases, and 54 Lung and Breast cases. For all these cases, the absolute dose at the isocentre was also measured using a small thimble ionization chamber (cc-13) in a PMMA cylindrical phantom as a reference.
Results and discussion Comparing calibration curves using different colour channels Figures 5, 6 and 7 plot the EBT2 film calibration curves from greyscale, red, green and blue channels. On X-axis is the absolute dose delivered, and on Y-axis is the net PV. In high dose ranges (Fig. 5), all calibration curves end up at a saturated region beyond a certain point, where the pixel value stays at the maximum value of 256 no matter how the dose increases. The authors refer to this point as the saturation dose point, which defines the high end of the film’s dynamic range. The Gafchromic EBT2 film should not be used in a dose range over this point. It is noted from Fig. 5 that although for an 8-bit image the raw PV for all channels saturates at a maximum value of 256, the net PV does not necessarily saturate at the same value. This is
Australas Phys Eng Sci Med (2013) 36:487–494
491
Fig. 5 Gafchromic EBT2 film calibration curves in high dose range (0–1,650 cGy). The pixel value saturates at the saturation dose point
Fig. 6 Gafchromic EBT2 film calibration curves in medium dose range (0–550 cGy)
because the background PVs for different channels are different, and consequently when the raw saturation point is corrected by these different background values, the resultant net PV saturation points are also different. Below the saturation point the pixel value increases when dose delivered increases, but not necessarily in a linear relationship in all four channels. Only the greyscale channel shows a satisfactory linear calibration curve in all three dose ranges. Figure 8 combines all the data points from different dose ranges and plots them from 0 to 600 cGy, where none of the channels saturates. A linear interpolation was added to the data points for each channel, with the R2 value displayed. The plot shows that compared to the other threes, the
calibration curve from the greyscale image is much more linear, with R2 value nearly equal to 1.000 (0.999). Figure 9 shows the calibration curve from a greyscale film, but scanned at different times after the irradiation, up to 48 h. These curves are all linear. This indicates that for relative measurement, the Gafchromic EBT2 film can be scanned at any time after the irradiation up to 48 h post exposure. Clinical testing In the past two years the application of Gafchromic EBT2 film by greyscale channel in IMRT QA has been clinically tested in the radiation therapy departments in St Vincent’s
123
492
Australas Phys Eng Sci Med (2013) 36:487–494
Fig. 7 Gafchromic EBT2 film calibration curves in low dose range (0–55 cGy)
Fig. 8 Gafchromic EBT2 film calibration curves using greyscale, red, green and blue channel respectively in a dose range of 0–600 cGy. A linear interpolation was fit to each calibration curve and the R2 value of the linear interpolation was displayed
Clinic and Mater Hospital, Genesis Cancer Care. Table 1 shows the number of patients that were treated with IMRT at different sites in the two departments during the trial period and the number of patients that fail the QA when using the novel Gafchromic-film method. The QA pass line is chosen by the radiation oncologists as 3 mm and 3 %. In the 2-year trial 2 lung-and-breast patients and 1 pelvis-and-prostate patient out of a total of 743 patients failed the QA. The two lung-and-breast patient plans failed their QA assessment using the greyscale analysis of the EBT2 film because the cross sections of their target volume were much larger than the dimension of the Gafchromic EBT2 film and consequently no background-correction
123
point could be found on the measurement film, which reduced the accuracy of the measurement and caused the QA to fail. The pelvis-and-prostate patient plan failed its QA assessment using the greyscale analysis of the EBT2 film because of a staff mistake: the radiation therapist accidentally sent the wrong plan to the machine rather than the approved plan. This was found out by the QA procedure and was corrected in time before the treatment started, which would otherwise become a clinical incident. The QAs were also performed with small thimble ionisation chamber (CC-13) and compared to the point-dose results obtained from the films. It was observed that they agreed well within experimental errors.
Australas Phys Eng Sci Med (2013) 36:487–494
493
Fig. 9 Gafchromic EBT2 film calibration curves using greyscale–channel image. The film was scanned 0, 2, 12 and 48 h after irradiation
Table 1 IMRT QA clinical trial statistics using the novel method to process and analyse the Gafchromic EBT2 film Treatment site
Brain, head and neck
Pelvis and prostate
Lung and breast
Total no.
Total no.
Total no.
No. of failed cases
No. of failed cases
No. of failed cases
St Vincent’s
95
0
111
1
9
1
Mater
118
0
356
0
54
1
Conclusion The novel method of using greyscale channel to analyse Gafchromic EBT2 film in IMRT QA is confirmed as effective and efficient, because when the film is analysed in the greyscale channel, the pixel value is linearly proportional to the dose delivered, which saves the calibration procedure when the Gafchromic film is only used for relative comparisons. Clinical testing has shown consistent results, proving that this novel method may be used to replace the conventional R–G–B method to analyse the Gafchromic EBT series films in IMRT QA. Acknowledgments The authors would like to thank all staff in Radiation Oncology Associates, Genesis Cancer Care, for their full support during the experiment work.
References 1. Devic S (2011) Radiochromic film dosimetry: past, present, and future. Phys Med 27:122–134 2. Sankar A, Ayyangar KM, Nehru RM, Kurup PG, Murali V, Enke CA (2006) Comparison of Kodak EDR2 and Gafchromic EBT film for intensity-modulated radiation therapy dose distribution verification. Med Dosim 31:273–282
3. Van Battum LJ, Hoffmans D, Piersma H, Heukelom S (2008) Accurate dosimetry with Gafchromic EBT film of a 6MV photon beam in water: what level is achievable? Med Phys 35:704–716 4. Xu L, McEwen M, Cogocaru C, Faddegon B (2009) Measurement of lateral dose distributions using Gafchromic EBT films and PTW starcheck 2D array. Med Phys 36:2624 5. Trichter S, Trichter F, Sabbas A, Lovelock D, Kulidzhanov F, Nori D (2005) Evaluation of Gafchromic EBT film for IMRT QA using two different scanners. Med Phys 32:2167 6. Ding L, Urie M, Lo Y (2005) Radiochromic EBT film characterization and applicability for IMRT QA verification. Med Phys 32:1987 7. Zeidan OA, Sephenson SA, Meeks SL, Wagner TH, Willoughby TR, Kupelian PA et al (2006) Characterization and use of radiochromic film for IMRT dose verification. Med Phys 33:4064–4072 8. Van Wieringen N, Van Herten YRJ, Van de Kamer JB, Erogluer A, Bel A, Wiersma J (2007) Development of a QA procedure for IMRT plans using EBT radiochromic film. Radiother Oncol 84:S54 9. Mahata A, Ravindran PB (2007) Investigation of the use of Gafchromic EBT film for routine IMRT QA. Radiother Oncol 84:S54 10. Devic S, Tomic N, Pang Z, Seuntjens J, Podgorsak EB, Soares CG (2005) Absorption spectroscopy of EBT model GafchromicTM film. Med Phys 32:2167 11. Devic S, Seuntjens J, Sham E, Podgorsak EB, Kirov AS, Schmidtlein RC et al (2005) Precise radiochromic film dosimetry using a flat-bed scanner. Med Phys 32:1983 12. Devic S, Seuntjens J, Hegyi G, Podgorsak EB, Soares CG, Kirov AS (2004) Dosimetric properties of improved Gafchromic films for seven different digitizers. Med Phys 31:2392–2401
123
494 13. Devic S, Tomic N, Soares CG, Podgorsak EB (2009) Optimizing the dynamic range extension of a radiochromic film dosimetry system. Med Phys 36:429–437 14. Todorovic M, Fischer M, Cremers F, Thomas E, Schmidt R (2006) Evaluation of Gafchromic EBT prototype B for external beam dose verification. Med Phys 33:1321
123
Australas Phys Eng Sci Med (2013) 36:487–494 15. Chu R, Van Dyk G, Lewis D, O’Hara K, Buckland B, Dinelle F (1990) GafchromicTM dosimetry media: a high dose, thin film routine dosimeter and dose mapping tool. Radiat Phys Chem 35:767–773 16. Devic S, Tomic N, Aldelaijan S, DeBlois F, Seuntjens J, Chan MF, Lewis D (2012) Linearization of dose–response curve of the radiochromic film dosimetry system. Med Phys 39:4850
TECHNICAL PAPER
Developing a novel method to analyse Gafchromic EBT2 films in intensity modulated radiation therapy quality assurance Yunfei Hu • Yang Wang • Gerald Fogarty Guilin Liu
•
Received: 14 May 2013 / Accepted: 20 November 2013 / Published online: 30 November 2013 Ó Australasian College of Physical Scientists and Engineers in Medicine 2013
Abstract Recently individual intensity modulated radiation therapy quality assurances (IMRT QA) have been more and more performed with GafchromicTM EBT series films processed in red–green–blue (R–G–B) channel due to their extremely high spatial resolution. However, the efficiency of this method is relatively low, as for each box of film, a calibration curve must be established prior to the film being used for measurement. In this study, the authors find a novel method to process the GafchromicTM EBT series, that is, to use the 16-bit greyscale channel to process the exposed film rather than the conventional 48-bit R–G– B channel, which greatly increases the efficiency and even accuracy of the whole IMRT procedure. The main advantage is that when processed in greyscale channel, the GafchromicTM EBT2 films exhibits a linear relationship between the net pixel value and the dose delivered. This linear relationship firstly reduces the error in calibrationcurve fitting, and secondly saves the need of establishing a calibration curve for each box of films if it is only to be used for relative measurements. Clinical testing for this novel method was carried out in two radiation therapy centres that involved a total of 743 IMRT cases, and 740 cases passed the 3 mm 3 % gamma analysis criteria. The cases were also tested with small ionization chambers (cc13) and the results were convincing. Consequently the authors recommend the use of this novel method to
Y. Hu (&) Y. Wang Radiation Therapy Department, Genesis Cancer Care, Level A St Vincent’s Clinic, Darlinghurst, NSW 2010, Australia e-mail: [email protected] G. Fogarty G. Liu Radiation Therapy Department, Mater Hospital, 25 Rockland Rd, Crows Nest, NSW 2065, Australia
improve the accuracy and efficiency of individual IMRT QA procedure using Gafchromic EBT2 films. Keywords Gafchromic film Intensity modulated radiation therapy Dosimetry Greyscale Quality assurance Clinical testing
Introduction Radiochromic films are useful for radiotherapy quality assurance (QA) due to their high spatial resolution, low energy dependence in a variety of beam qualities, and near tissue equivalence, which are especially important for dose measurements with small segments and sharp gradients [1]. One good example of film applications is its use in individual plan verification for intensity modulated radiation therapy (IMRT) QA [5–9]. Gafchromic EBT film was initially designed to replace silver halide radiographic film for IMRT QA procedures [2]. However, it was soon reported that the EBT film had a relatively high measurement uncertainty, and this measurement uncertainty was mostly attributed to the nonuniformity of the sensitive layer of the film [3, 4]. To improve the uniformity, the manufacturer added a yellow dye to the sensitive layer, leading to the invention of the GafchromicTM EBT2 film. When analysing dose measurements using EBT2 film, the red channel, extracted from the red–green–blue (R–G– B) scanned images, is most often used, because it has the highest absorption and sensitivity (defined as the ratio of the change in optical density and the amount of dose) [10]. Historically the quantity used to analyse GafchromicTM film was mostly the optical density (OD). While the OD as a quantity of choice has been more or less inherited from
123
488
radiographic films, there have been attempts to use a sole pixel value (PV) from a document scanner. Compared to the OD, the use of PV (or raw data) always gives more pronounced non-linear behaviour, which will in turn lead to a higher uncertainty dur to the larger fitting error in establishing a calibration curve [1]. To establish a calibration curve and use the results for future measurements, the common way is to plot dose as a function of the measured net OD and then fit the data with an appropriate function [11]. For every particular radiochromic film dosimetry system, the sensitivity curve will be different [12]. If a linear response can be established between the PV and the dose, i.e., Dose = aPV ? b, where a and b are two measurable constants, the accuracy and efficiency of calibrating GafchromicTM EBT film can be significantly improved. The conventional R–G–B film analysis method requires an individual calibration curve to be established for every new box of Gafchromic film due to the film’s batch variation, and to establish an accurate polynomial fitting curve, a large number of data points need to be collected. On the other hand, when the calibration curve is linear, only a start and an end point are needed to establish an accurately fitting calibration curve, with a significantly reduced fitting error. With the introduction of the net PV, which is defined by subtracting the raw PV value with the background PV from an unexposed film piece, the b constant in the equation becomes zero, and the final equation between the Dose and the net PV now becomes Dose = anet PV. With this known linear relationship that goes through the origin, the PV can be directly compared to the dose value through a normalisation function (which determines the value for a), and there is no longer any need to establish a calibration curve for each box of film. The main advantages of establishing such a relationship include improved efficiency (only one normalisation point is needed instead of multiple calibration points) and accuracy (simpler fitting function), which are particularly important for individual IMRT QA. There have been a few studies working on establishing a linear function between dose and PV, mostly using red channel [11, 14, 16]. Todorovic et al. found that when the film was scanned in as a 14-bit greyscale image rather than the 42-bit RGB image, the resultant calibration curve was more linear [14]. Recently Devic et al. established a functional argument to linearize the inherently non-linear response of the Gafchromic EBT2 film, and was proved successful, but this argument was based on net OD instead of PV [16]. In this study, based on a number of experiments and clinical studies, the authors found that when the GafchromicTM EBT2 film is scanned by the greyscale channel of a flatbed scanner, a linear and sensitive PV-dose calibration
123
Australas Phys Eng Sci Med (2013) 36:487–494
Fig. 1 Set up EBT2 film for calibration. Film is aligned such that its centre part is located at the centre of the reference field. Build-up solid water is to be added after the film is aligned
curve can be established within a clinically useful dose range. If the net PV is used, the relationship between dose and net PV goes through the origin, and after normalisation, the net PV can be directly compared to the dose for evaluation. This method of analysing the dose measurements using GafchromicTM EBT2 film for individual IMRT QA has been clinically tested with a total of 743 IMRT patients’ plans with different anatomical sites at two clinics, and it was found that the criteria of 3 %/3 mm for an IMRT QA plan are passing almost 100 % gamma function points.
Materials and methods Calibrating EBT2 film using greyscale To calibrate the EBT2 film, three pieces of films from the same batch, each with a dimension of 254 mm 9 203 mm, were used. Each piece was sliced into twelve strips, all with a dimension of 20 mm 9 203 mm. The films were irradiated in solid water phantom under a linear accelerator (Siemens Artiste M5298). Each strip was irradiated independently with a known dose, as is shown in Fig. 1. Three pieces of films were used because the authors planned to calibrate the EBT2 film in three different clinically-useful dose ranges so as to determine the film’s useful dynamic dose ranges. Film 1 was irradiated with 0 cGy (background) to 55 cGy in an equivalent step of 5 cGy; film 2 was irradiated with 0 cGy to 550 cGy in an equivalent step of 50 cGy; and film 3 was irradiated with 0 cGy to 1,650 cGy in an equivalent step of 150 cGy. The Monitor Units (MUs) required to deliver the planned dose was calculated by CMS XiO radiation therapy planning system with the phantoms scanned into the system by a Computed Tomography (CT) scanner.
Australas Phys Eng Sci Med (2013) 36:487–494
489
Fig. 2 Example of a calibration film composed of strips irradiated with different doses; the amount of the dose delivered is marked at the bottom of each strip. The film can be analysed using either the conventional RGB channels or the novel Greyscale channel
Scanning and analysing the film Once all strips from the same piece of film were irradiated with the planned dose, the authors carefully reconstructed and taped together the original, complete sheet of film, as is shown in Fig. 2. The films were scanned using an EPSON Expression 10000XL flatbed scanner (EPSON Seiko Epson Corporation, Suwa, Japan) in both 48-bit colour and 16-bit greyscale format. The 48-bit colour image was split into the 16-bit red, green and blue channels. The scanned images were saved in JPEG format. TIFF format was the other option, but compared with JPEG, TIFF images has a much larger file size, which gives difficulties in image analysis and file storage for frequent clinical uses in the future. The disadvantage of JPEG is that as it does not support 16-bit images, when 16-bit image is saved it is automatically converted down to 8 bit. The authors are fully aware of this problem but for the purpose of this paper and potential clinical uses the 8-bit image data are considered adequate, so the scanned images are saved in JPEG format and used for analysis. To study the effect of film-self-developing [15] after irradiation on the accuracy of the result, films were scanned on the spot of irradiation (0 h), then 0.5, 1, 2, 12, 24 and 48 h after irradiation. The authors strictly controlled the consistency of other scan settings which may affect the final result, i.e. orientation, scan position, colour adjustments, brightness and contrast, and resolution. Software used to read the PV values from the calibration films was software developed by one of the authors (A. Prof Yang Wang) and approved by Australian Therapeutic
Fig. 3 Cylindrical PMMA phantom used for IMRT QA with GafchromicTM EBT2 film
Goods Administration (TGA), called Radiation Oncology Dosimetry Management System (RODOMS). The PV readings were taken at the centre of each film strip. Five readings were taken, averaged and subtracted by the background reading. The corrected PV reading was then plotted against dose in Excel to establish the calibration curve. Calibration curves from different channels (grey, red, green and blue) were compared and analysed. Clinical testing The novel method: greyscale channel was used to analyse the EBT2 film in IMRT QA, was then clinically tested. To
123
490
Fig. 4 Two supporting feet and upper half and lower half of the cylindrical phantom
simulate real patient’s geometry, a home-designed cylindrical PMMA phantom was used, as is shown in Figs. 3 and 4. The phantom is composed of two supporting feet, the upper half of the cylindrical and the lower half of the cylindrical. During QA, film sits in between the two halves horizontally. Before irradiation, the plane where the film sits is aligned at the iso-centre against the optical crosshair. The length and the diameter of the cylindrical is 30.0 and 24.0 cm separately. One piece of EBT2 film was irradiated in the phantom with the individual IMRT plan, scanned with exactly the same settings as the calibration film, and then compared to the dose map exported from the planning system in RODOMS. The film analysing part of RODOMS allowed the authors to compare the film in terms of horizontal/vertical profiles, iso-dose lines, point-to-point doses, and gamma function with different tolerances. The film analysing part of RODOMS software was designed based on the fact that when the Gafchromic EBT2 film is scanned in the greyscale channel, the resultant pixel value is linearly proportional to the dose (Dose = aPV ? b). Before direct comparison, three corrections need to be applied to the greyscale film: rescaling and co-registration, background subtraction, and normalisation. Rescaling and co-registration aims to rescale the scanned greyscale image to the same size of the planning dose map, and then coregister them so that the corresponding points are in the same coordinates. Then background subtraction is achieved by firstly choosing a background point in the unexposed area of the film, then the PV of this point subtracted from all the data points on the greyscale film to obtain the net PV so that now the relationship between dose and PV becomes: Dose = anet PV. Finally normalisation is performed by choosing a normalisation point in a uniform dose area inside the PTV; a normalisation factor (a) is
123
Australas Phys Eng Sci Med (2013) 36:487–494
calculated by comparing the PV of this normalisation point to the PV of the point at the same coordinate in the dose map. This normalisation value is then applied to all the data points on the greyscale film. These functions are only valid if PV is linearly proportional to the dose, and no extra calibration curve is required. The disadvantage of using a single-point background subtraction and normalisation correction functions is that the measurement error of the chosen background/normalisation point will be inherited by all the other points. This error can be reduced via: (a) carefully choosing the background/normalisation point which locates in a uniform dose area; and (b) use multiple correction points to reduce the error. After all correction functions are applied, the greyscale film is compared to the planning output in terms of: point dose comparison, iso-dose line matching, dose profile comparison and gamma-value function assessment. The 3 mm/3 % gamma function assessment is used as the major criteria for QA results. The clinical testing was carried out at the radiation therapy department at St Vincent’s Clinic (Darlinghurst, NSW, Australia) and Mater Hospital (Crows Nest, NSW, Australia). At St Vincent’s Clinic in total 215 IMRT patients’ plans were tested, including 95 Brain, Head and Neck cases, 111 Pelvis and Prostate cases, and 9 Lung and Breast cases. At Mater Hospital in total 528 IMRT patients’ plans were tested, including 118 Brain, Head and Neck cases, 356 Pelvis and Prostate cases, and 54 Lung and Breast cases. For all these cases, the absolute dose at the isocentre was also measured using a small thimble ionization chamber (cc-13) in a PMMA cylindrical phantom as a reference.
Results and discussion Comparing calibration curves using different colour channels Figures 5, 6 and 7 plot the EBT2 film calibration curves from greyscale, red, green and blue channels. On X-axis is the absolute dose delivered, and on Y-axis is the net PV. In high dose ranges (Fig. 5), all calibration curves end up at a saturated region beyond a certain point, where the pixel value stays at the maximum value of 256 no matter how the dose increases. The authors refer to this point as the saturation dose point, which defines the high end of the film’s dynamic range. The Gafchromic EBT2 film should not be used in a dose range over this point. It is noted from Fig. 5 that although for an 8-bit image the raw PV for all channels saturates at a maximum value of 256, the net PV does not necessarily saturate at the same value. This is
Australas Phys Eng Sci Med (2013) 36:487–494
491
Fig. 5 Gafchromic EBT2 film calibration curves in high dose range (0–1,650 cGy). The pixel value saturates at the saturation dose point
Fig. 6 Gafchromic EBT2 film calibration curves in medium dose range (0–550 cGy)
because the background PVs for different channels are different, and consequently when the raw saturation point is corrected by these different background values, the resultant net PV saturation points are also different. Below the saturation point the pixel value increases when dose delivered increases, but not necessarily in a linear relationship in all four channels. Only the greyscale channel shows a satisfactory linear calibration curve in all three dose ranges. Figure 8 combines all the data points from different dose ranges and plots them from 0 to 600 cGy, where none of the channels saturates. A linear interpolation was added to the data points for each channel, with the R2 value displayed. The plot shows that compared to the other threes, the
calibration curve from the greyscale image is much more linear, with R2 value nearly equal to 1.000 (0.999). Figure 9 shows the calibration curve from a greyscale film, but scanned at different times after the irradiation, up to 48 h. These curves are all linear. This indicates that for relative measurement, the Gafchromic EBT2 film can be scanned at any time after the irradiation up to 48 h post exposure. Clinical testing In the past two years the application of Gafchromic EBT2 film by greyscale channel in IMRT QA has been clinically tested in the radiation therapy departments in St Vincent’s
123
492
Australas Phys Eng Sci Med (2013) 36:487–494
Fig. 7 Gafchromic EBT2 film calibration curves in low dose range (0–55 cGy)
Fig. 8 Gafchromic EBT2 film calibration curves using greyscale, red, green and blue channel respectively in a dose range of 0–600 cGy. A linear interpolation was fit to each calibration curve and the R2 value of the linear interpolation was displayed
Clinic and Mater Hospital, Genesis Cancer Care. Table 1 shows the number of patients that were treated with IMRT at different sites in the two departments during the trial period and the number of patients that fail the QA when using the novel Gafchromic-film method. The QA pass line is chosen by the radiation oncologists as 3 mm and 3 %. In the 2-year trial 2 lung-and-breast patients and 1 pelvis-and-prostate patient out of a total of 743 patients failed the QA. The two lung-and-breast patient plans failed their QA assessment using the greyscale analysis of the EBT2 film because the cross sections of their target volume were much larger than the dimension of the Gafchromic EBT2 film and consequently no background-correction
123
point could be found on the measurement film, which reduced the accuracy of the measurement and caused the QA to fail. The pelvis-and-prostate patient plan failed its QA assessment using the greyscale analysis of the EBT2 film because of a staff mistake: the radiation therapist accidentally sent the wrong plan to the machine rather than the approved plan. This was found out by the QA procedure and was corrected in time before the treatment started, which would otherwise become a clinical incident. The QAs were also performed with small thimble ionisation chamber (CC-13) and compared to the point-dose results obtained from the films. It was observed that they agreed well within experimental errors.
Australas Phys Eng Sci Med (2013) 36:487–494
493
Fig. 9 Gafchromic EBT2 film calibration curves using greyscale–channel image. The film was scanned 0, 2, 12 and 48 h after irradiation
Table 1 IMRT QA clinical trial statistics using the novel method to process and analyse the Gafchromic EBT2 film Treatment site
Brain, head and neck
Pelvis and prostate
Lung and breast
Total no.
Total no.
Total no.
No. of failed cases
No. of failed cases
No. of failed cases
St Vincent’s
95
0
111
1
9
1
Mater
118
0
356
0
54
1
Conclusion The novel method of using greyscale channel to analyse Gafchromic EBT2 film in IMRT QA is confirmed as effective and efficient, because when the film is analysed in the greyscale channel, the pixel value is linearly proportional to the dose delivered, which saves the calibration procedure when the Gafchromic film is only used for relative comparisons. Clinical testing has shown consistent results, proving that this novel method may be used to replace the conventional R–G–B method to analyse the Gafchromic EBT series films in IMRT QA. Acknowledgments The authors would like to thank all staff in Radiation Oncology Associates, Genesis Cancer Care, for their full support during the experiment work.
References 1. Devic S (2011) Radiochromic film dosimetry: past, present, and future. Phys Med 27:122–134 2. Sankar A, Ayyangar KM, Nehru RM, Kurup PG, Murali V, Enke CA (2006) Comparison of Kodak EDR2 and Gafchromic EBT film for intensity-modulated radiation therapy dose distribution verification. Med Dosim 31:273–282
3. Van Battum LJ, Hoffmans D, Piersma H, Heukelom S (2008) Accurate dosimetry with Gafchromic EBT film of a 6MV photon beam in water: what level is achievable? Med Phys 35:704–716 4. Xu L, McEwen M, Cogocaru C, Faddegon B (2009) Measurement of lateral dose distributions using Gafchromic EBT films and PTW starcheck 2D array. Med Phys 36:2624 5. Trichter S, Trichter F, Sabbas A, Lovelock D, Kulidzhanov F, Nori D (2005) Evaluation of Gafchromic EBT film for IMRT QA using two different scanners. Med Phys 32:2167 6. Ding L, Urie M, Lo Y (2005) Radiochromic EBT film characterization and applicability for IMRT QA verification. Med Phys 32:1987 7. Zeidan OA, Sephenson SA, Meeks SL, Wagner TH, Willoughby TR, Kupelian PA et al (2006) Characterization and use of radiochromic film for IMRT dose verification. Med Phys 33:4064–4072 8. Van Wieringen N, Van Herten YRJ, Van de Kamer JB, Erogluer A, Bel A, Wiersma J (2007) Development of a QA procedure for IMRT plans using EBT radiochromic film. Radiother Oncol 84:S54 9. Mahata A, Ravindran PB (2007) Investigation of the use of Gafchromic EBT film for routine IMRT QA. Radiother Oncol 84:S54 10. Devic S, Tomic N, Pang Z, Seuntjens J, Podgorsak EB, Soares CG (2005) Absorption spectroscopy of EBT model GafchromicTM film. Med Phys 32:2167 11. Devic S, Seuntjens J, Sham E, Podgorsak EB, Kirov AS, Schmidtlein RC et al (2005) Precise radiochromic film dosimetry using a flat-bed scanner. Med Phys 32:1983 12. Devic S, Seuntjens J, Hegyi G, Podgorsak EB, Soares CG, Kirov AS (2004) Dosimetric properties of improved Gafchromic films for seven different digitizers. Med Phys 31:2392–2401
123
494 13. Devic S, Tomic N, Soares CG, Podgorsak EB (2009) Optimizing the dynamic range extension of a radiochromic film dosimetry system. Med Phys 36:429–437 14. Todorovic M, Fischer M, Cremers F, Thomas E, Schmidt R (2006) Evaluation of Gafchromic EBT prototype B for external beam dose verification. Med Phys 33:1321
123
Australas Phys Eng Sci Med (2013) 36:487–494 15. Chu R, Van Dyk G, Lewis D, O’Hara K, Buckland B, Dinelle F (1990) GafchromicTM dosimetry media: a high dose, thin film routine dosimeter and dose mapping tool. Radiat Phys Chem 35:767–773 16. Devic S, Tomic N, Aldelaijan S, DeBlois F, Seuntjens J, Chan MF, Lewis D (2012) Linearization of dose–response curve of the radiochromic film dosimetry system. Med Phys 39:4850
