Incorporating multislice imaging into x-ray CT polymer gel dosimetry

Incorporating multislice imaging into x-ray CT polymer gel dosimetry H. Johnston, M. Hilts, and A. Jirasek Citation: Medical Physics 42, 1666 (2015); doi: 10.1118/1.4914419 View online: http://dx.doi.org/10.1118/1.4914419 View Table of Contents: http://scitation.aip.org/content/aapm/journal/medphys/42/4?ver=pdfcov Published by the American Association of Physicists in Medicine Articles you may be interested in Improved image quality for x-ray CT imaging of gel dosimeters Med. Phys. 38, 5130 (2011); 10.1118/1.3626487 Adaptive mean filtering for noise reduction in CT polymer gel dosimetry Med. Phys. 35, 344 (2008); 10.1118/1.2818742 X-ray CT dose in normoxic polyacrylamide gel dosimetry Med. Phys. 34, 1934 (2007); 10.1118/1.2732032 Investigations in x-ray computed tomography polyacrylamide gel dosimetry Med. Phys. 32, 3058 (2005); 10.1118/1.2000648 Image filtering for improved dose resolution in CT polymer gel dosimetry Med. Phys. 31, 39 (2004); 10.1118/1.1633106

Incorporating multislice imaging into x-ray CT polymer gel dosimetry H. Johnstona)

Department of Physics and Astronomy, University of Victoria, Victoria, British Columbia V8W 2Y2, Canada

M. Hiltsb)

Department of Physics and Astronomy, University of Victoria, Victoria, British Columbia V8W 2Y2, Canada and Medical Physics, BC Cancer Agency, Vancouver Island Centre, Victoria, British Columbia V8R 6V5, Canada

A. Jirasek

Department of Physics and Astronomy, University of Victoria, Victoria, British Columbia V8W 2Y2, Canada and Department of Physics, University of British Columbia—Okanagan Campus, Kelowna, British Columbia V1V 1V7, Canada

(Received 17 July 2014; revised 1 February 2015; accepted for publication 5 February 2015; published 17 March 2015) Purpose: To evaluate multislice computed tomography (CT) scanning for fast and reliable readout of radiation therapy (RT) dose distributions using CT polymer gel dosimetry (PGD) and to establish a baseline assessment of image noise and uniformity in an unirradiated gel dosimeter. Methods: A 16-slice CT scanner was used to acquire images through a 1 L cylinder filled with water. Additional images were collected using a single slice machine. The variability in CT number (NCT) associated with the anode heel e↵ect was evaluated and used to define a new slice-by-slice background subtraction artifact removal technique for CT PGD. Image quality was assessed for the multislice system by evaluating image noise and uniformity. The agreement in NCT for slices acquired simultaneously using the multislice detector array was also examined. Further study was performed to assess the e↵ects of increasing x-ray tube load on the constancy of measured NCT and overall scan time. In all cases, results were compared to the single slice machine. Finally, images were collected throughout the volume of an unirradiated gel dosimeter to quantify image noise and uniformity before radiation is delivered. Results: Slice-by-slice background subtraction e↵ectively removes the variability in NCT observed across images acquired simultaneously using the multislice scanner and is the recommended background subtraction method when using a multislice CT system. Image noise was higher for the multislice system compared to the single slice scanner, but overall image quality was comparable between the two systems. Further study showed NCT was consistent across image slices acquired simultaneously using the multislice detector array for each detector configuration of the slice thicknesses examined. In addition, the multislice system was found to eliminate variations in NCT due to increasing x-ray tube load and reduce scanning time by a factor of 4 when compared to imaging a large volume using a single slice scanner. Images acquired through an unirradiated, active gel revealed NCT varies between the top and bottom of the 1 L cylinder as well as across the diameter of the cylinder by up to 7 HU. Conclusions: Multislice CT imaging has been evaluated for CT PGD and found to be the superior technique compared to single slice imaging in terms of the time required to complete a scan and the tube load characteristics associated with each scanning method. The implementation of multislice scanning is straightforward and expected to facilitate routine gel dosimetry measurements for complex dose distributions in modern RT centers. C 2015 American Association of Physicists in Medicine. [http://dx.doi.org/10.1118/1.4914419] Key words: polymer gel dosimetry, x-ray computed tomography, radiation therapy, quality assurance

1. INTRODUCTION Polymer gel dosimetry (PGD) is being actively developed as a tool for dose verification in modern radiation therapy (RT) to meet the growing need for three-dimensional (3D) evaluation of complex radiation treatments.1 The technique is expected to play a role in end-to-end dosimetry tests and as a commissioning tool for new treatment methods. Polymer gels consist of radiosensitive monomers infused in a gelatin matrix, which polymerize as a function of absorbed radiation 1666

Med. Phys. 42 (4), April 2015

dose. The recorded dose information is typically extracted from the gel using magnetic resonance imaging (MRI),2–6 optical computed tomography (optical-CT),7–10 or x-ray CT.11–15 Recent development of a prototype CT PGD system demonstrated the ability of this technique to verify spatially complex dose distributions in 3D.15 The system utilizes a new gel formulation16 designed to have enhanced radiation sensitivity as well as a robust new image postprocessing technique17 and gel calibration method.15,18 Given these recent technical advances and the availability of CT scanners in RT clinics, CT

0094-2405/2015/42(4)/1666/12/$30.00

© 2015 Am. Assoc. Phys. Med.

1666

1667

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

PGD is an attractive method for performing gel dosimetry in modern RT. However, with few exceptions,19,20 CT PGD has used single slice CT, not taking advantage of the availability of multislice machines in modern RT facilities. Use of a multislice scanner stands to make CT PGD more practical to implement in RT clinics due to the dramatic reduction in scan time expected for gel imaging. Detailed overviews of CT PGD have been provided previously.1,21 Briefly, image contrast observed in CT images of irradiated polymer gel is produced by the mass density changes that occur in the gel during polymerization.22–24 As the radiation dose increases, more monomer is converted to polymer until the monomer is used up, resulting in an increase in gel density at the location of dose deposition. For many polymer gels, the magnitude of the density change produced by exposure to ionizing radiation is small, resulting in low signal-to-noise CT images and poor dose resolution for CT PGD systems.22 For this reason, work in CT PGD has largely focused on improving the dose sensitivity of polymer gels by identifying new dose-sensitive monomers25,26 and finding methods to increase the concentration of cross-linking monomer in the gel system.16,27–29 The formulation used in the recent prototype dosimetry system was specifically designed to produce more polymer per unit dose compared to previous formulations.28,29 A greater polymer yield significantly improves the contrast of gel CT images and provides superior dose resolution for the system,16 as illustrated by the example dose image of the new gel irradiated with a 6 MV intensity modulated radiation therapy (IMRT) distribution shown in Fig. 1. This advancement, together with the reduction in scan time associated with multislice CT imaging, has the potential to make CT readout the preferred choice in PGD over MRI and optical-CT, as these techniques continue to be associated with several technical limitations3,6,8 as well as long scan times and lack of scanner availability.1

Fig. 1. A dose image of the new polymer gel designed to have enhanced radiation sensitivity compared to previous formulations (image scale is at bottom). The gel was irradiated with a 6 MV IMRT dose distribution and imaged using a multislice CT scanner. Medical Physics, Vol. 42, No. 4, April 2015

1667

Alongside gel development, incorporation of new image postprocessing techniques has improved the signal-to-noise of gel CT images, resulting in better image quality.17 However, even with a high-sensitivity gel and advanced image postprocessing techniques, CT PGD requires image averaging to reduce the stochastic noise in gel CT images.30 For a typical experiment, upwards of 30 images are required at each slice location in order to achieve optimal image quality. Depending on the imaging parameters used for the scan, this high number of images can place excessive load on the x-ray tube of the scanner. For example, a 12 cm long gel imaged using 3 mm slice thickness and 30 image averages per slice would required over 1000 CT images. For single slice machines, this level of scanning can overheat the x-ray tube and initiate a tube cooling period to prevent x-ray tube damage. The additional time required for tube cooling results in long scan times on the order of a few hours to collect images throughout a gel volume. In addition, we have observed significant variations in CT number (NCT) as the number of images collected increases if scanning is continued after multiple cooling periods with a single slice machine. The tube load e↵ects observed for single slice CT scanners may not be present for multislice systems. As multiple image slices are acquired simultaneously at each couch position, the load placed on the x-ray tube during scanning with a multislice scanner will be substantially reduced.31 Lowering the tube load has the potential to reduce the total time required for a scan from hours to minutes due to the more efficient method of collecting slice data and reduction in tube cooling periods during acquisition. Multislice scanning may also reduce or even eliminate the variations in NCT observed on single slice machines as the number of images acquired increases. Nevertheless, multislice systems present their own unique challenges when utilized for gel dosimetry. The use of a cone beam of x-rays may introduce significant variability in NCT across image slices acquired simultaneously using the multislice detector array due to the anode heel e↵ect.32 Moreover, the increase in scattered radiation associated with a cone-beam of x-rays may result in greater image noise 33 that may degrade the quality of gel CT images. For these reasons, it is imperative that multislice systems be thoroughly examined for CT PGD to ensure gel CT images are reliable and of sufficient quality for RT dose measurements. This paper presents studies performed to evaluate multislice CT scanning for CT PGD. The variability in NCT associated with the anode heel e↵ect is evaluated and used to define a new background subtraction technique for gel imaging. Relationships between image noise and several imaging parameters are established, and the uniformity of CT images is evaluated. The consistency of NCT across image slices is also assessed for each detector configuration. Further analysis is performed to evaluate the variability in NCT across a gel volume due to increasing x-ray tube load as well as the time required for volume scanning. Based on these results, images are collected throughout the volume of an unexposed, active gel to establish a baseline assessment of image noise and uniformity for a gel dosimeter before radiation is delivered.

1668

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

2. MATERIALS AND METHODS 2.A. Multislice CT scanner

The Optima CT580 multislice CT scanner (GE Medical Systems, Milwaukee, WI) used throughout this work is a 16slice, third generation, rotate–rotate machine that is similar to multislice CT scanners currently employed in many modern RT clinics. It is equipped with a Performix™ Pro VCT 100 x-ray tube with a maximum anode heat capacity of 5.7 MJ, more than twice that of a typical single slice machine. Two independently controlled tungsten cams provide prepatient beam collimation by tracking the x-ray focal spot during operation to ensure the narrowest possible field for a given slice thickness. The scanner also includes a HiLight® matrix detector composed of a solid scintillator material that provides 99% photon absorption efficiency. The detector is segmented into 24 cells along the direction parallel to the long axis of the patient to achieve postpatient collimation of the x-ray beam. The inner 16 rows are 0.625 mm wide and the outer 4 rows located on either side of the central matrix are 1.25 mm wide. The signals from each row can be used individually or combined using various detector configurations to adjust the slice thickness for up to 16 image slices. As such, the time required for a given patient scan can be reduced by up to a factor of 16 when compared to scanning using a single slice machine given the same technique parameters. When all the detector cells are used together, the system allows up to 20 mm of patient anatomy to be imaged simultaneously at a given couch position. 2.B. Imaging phantoms

All imaging phantoms used in this work were housed in 1 L polyethylene terephthalate cylinders (Modus Medical Devices, London, ON). For the anode heel e↵ect study, image quality investigations, and tube load experiment, the cylinder was filled with water at room temperature and remained in the CT simulator suite for 24 h prior to imaging to ensure a consistent water temperature throughout scanning. For the active gel investigation, a polymer gel containing 15% N-isopropylacrylamide (NIPAM), 4.5% N,N0methylenebisacrylamide (BIS), 5% gelatin, and 5 mM tetrakis hydroxymethyl phosphonium chloride (THPC) was manufactured and stored in a 1 L cylinder as described previously.15,16 The gel was not irradiated and remained in the refrigerator for the length of time typically required for the gel to become solid, radiation to be delivered, and polymerization to stabilize (i.e., ⇠27 h in total). An additional blank gel was manufactured as above using 5% gelatin, 0.5% NIPAM, and 0.5% BIS (no THPC) for background image subtraction. The blank gel was kept in the refrigerator with the active gel for the same amount of time before imaging.

on the CT couch using a custom-built head and neck phantom described previously.15 Details particular to each study are summarized below. 2.C.1. Reference protocol

A reference protocol was defined for imaging the waterfilled cylinder on the multislice scanner based on parameters shown to provide optimal image quality for the same phantom on a single slice machine.30,34 The resulting protocol acquires 8 image slices per couch position using the parameters indicated in bold in Table I. Images were collected using axial mode of acquisition unless otherwise indicated. 2.C.2. Anode heel effect and background subtraction

To examine the e↵ects of the anode heel e↵ect, a set of 25 test images were collected using the reference protocol through the center of the water-filled cylinder at each of the eight slice positions in the detector array. An additional set of 25 images were collected at each slice location using the same protocol to examine the ability of a new slice-by-slice background subtraction technique to remove the heel e↵ect variability. The new slice-by-slice image processing technique is described in detail in Sec. 2.D.1. 2.C.3. Image noise and uniformity

Image noise was evaluated using a series of scans that varied each imaging parameter from the reference protocol as summarized in Table I. For each parameter setting, one test image and one background image were acquired through the center of the water phantom at each slice location. The same set of test and background images collected using the reference protocol was used to examine image uniformity. An additional set of 25 test images and 25 background images were acquired using the multislice scanner and a HiSpeed Fx/i single slice machine (GE Medical Systems, Milwaukee, WI) to compare the image noise and uniformity associated with each system during a typical CT PGD experiment. The protocol used on the single slice scanner15 contained the same imaging parameters as the reference protocol with the exception of slice thickness (3 mm for the single slice scanner, 2.5 mm for the multislice machine).

Table I. The imaging parameters independently varied from the reference protocol (shown in bold) to examine image noise for the Optima CT580 multislice scanner. Parameter

2.C. CT imaging

CT imaging for all studies was performed using the Optima CT580 multislice scanner described in Sec. 2.A unless otherwise indicated. Each 1 L cylinder was positioned Medical Physics, Vol. 42, No. 4, April 2015

1668

Tube voltage (kV) Tube current (mA) Gantry rotation time (s) Slice thickness (mm)

Values examined 80, 100, 120, 140 100, 150, 200, 250, 300, 330 0.5, 0.6, 0.7, 0.8, 0.9, 1, 2, 3, 4 0.625, 1.25, 2.5, 3.75, 5, 7.5, 10

1669

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

2.C.4. Consistency in NCT for multislice scanning

The consistency in NCT for image slices acquired simultaneously was examined for each slice thickness between 0.625 and 10 mm and its associated detector combinations (see Table IV). This was achieved by collecting images of the water phantom using scanning protocols that varied the detector configuration for each slice thickness. All other imaging parameters were consistent with the reference protocol. An additional set of background images were collected for each detector arrangement for slice-by-slice background subtraction. 2.C.5. Effects of increasing tube load

The e↵ects of increasing x-ray tube load were investigated using the reference protocol to collect a set of 25 test images at each of 48 adjacent slice locations along the length of the 1 L water-filled cylinder (i.e., over 12 cm). This resulted in 1200 test images in total and required 6 independent, 20 mm wide array acquisitions at 6 distinct couch positions. An additional set of 1200 images were collected using the same protocol for background subtraction. The experiment was then repeated using 4D CT with the same technique parameters (each image was reconstructed using one full rotation of projection measurements). A set of 525 images were also acquired at 21 slice locations (i.e., over 6 cm) using the single slice scanner for comparison with the multislice machine. Overheating of the x-ray tube prevented acquisition of images along the entire length of the cylinder when using the single slice scanner. The time required for each scan was recorded to compare the total time required for volume scanning using single and multislice machines. 2.C.6. Effects of imaging an active gel

The e↵ects of imaging an active gel before radiation is delivered were evaluated by imaging the 1 L active gel along the entire length of the cylinder using the reference protocol. Twenty-five images were collected at each slice location. A second set of images were acquired through the 1 L blank gel using the same protocol for background subtraction. 2.D. Image processing and analyses

All image processing and data analyses were performed using matlab R2010b (MathWorks, Natick, MA). Image averaging was performed for both test and background images prior to further image processing for all studies where more than one image was acquired at a given slice position. Background subtraction was performed using the new slice-by-slice technique (described below) unless otherwise indicated. Further details on image processing and analysis for each investigation are given below. 2.D.1. Anode heel effect and background subtraction

Background subtraction was performed using a single slice and a new slice-by-slice technique. Using the single Medical Physics, Vol. 42, No. 4, April 2015

1669

slice method, the averaged test image at the center of the multislice detector array (slice 5 of 8) was subtracted from the remaining, averaged test images. A similar procedure was used in previous work for a single slice scanner and was shown to be an e↵ective method for removing artifacts in gel CT images.11 The new slice-by-slice technique involves subtracting the averaged background image at each slice location from the averaged test image at the corresponding slice position in the detector array. The processed images for each subtraction method were analyzed by extracting the mean and standard deviation of NCT within a square 50 ⇥ 50 pixel region of interest (ROI) at the center of the 1 L cylinder. Mean NCT were then plotted as a function of slice position with respect to the CT room coordinate system. The resulting curve was fit to a linear function to assess the consistency in NCT across the detector array achieved using each background subtraction technique. 2.D.2. Image noise and uniformity

The e↵ects of varying scanning parameters on image noise were determined by computing the standard deviation of NCT ( NCT) within a 50 ⇥ 50 pixel square ROI at the center of the 1 L cylinder for each processed image. Relationships between image noise and each parameter were established by plotting NCT as a function of the parameter values and fitting each of the resulting curves to a quadratic function. Image uniformity was assessed for one representative CT image by extracting the mean and standard deviation of NCT from 36 individual 20 ⇥ 20 pixel square ROIs at the center of the 1 L cylinder and 15 concentric, 3 pixel wide ring ROIs extending radially over the 1 L cylinder. A ring ROI analysis was used in addition to the square ROI evaluation to examine the e↵ects of residual ring artifacts that may be present following background subtraction. In addition, the standard deviation of mean NCT ( Nmean CT) was also computed over the grid and ring ROIs of each slice in the array. The resulting Nmean CT values were plotted as a function of slice position with respect to the CT room coordinate system to assess the consistency in uniformity for image slices acquired simultaneously. Both NCT and Nmean CT were also calculated for one slice acquired using the single and multislice scanners to compare the performance of the two systems. 2.D.3. Consistency in NCT for multislice scanning

The consistency in NCT across image slices acquired simultaneously was examined by extracting the mean and standard deviation of NCT within a square 50 ⇥ 50 pixel ROI at the center of the 1 L cylinder for each processed image. Mean NCT were then plotted as a function of slice position with respect to the CT room coordinate system for each slice thickness and its associated detector configurations. The minimum, maximum, and range (maximum–minimum value) of NCT and NCT were also computed for each configuration where more than one slice was acquired simultaneously.

1670

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

2.D.4. Effects of increasing tube load

To examine the e↵ects of increasing tube load for the axial and 4D CT scans, slice-by-slice background subtraction was performed using a central array of images within the set of test images as well as a separate set of images collected over an entire volume. The mean and standard deviation of NCT within a square 50 ⇥ 50 pixel ROI were extracted from the center of the 1 L cylinder for each processed image of each scan. Mean NCT were then plotted as a function of slice position (with respect to the CT room coordinate system) to assess the variability in NCT associated with tube load when imaging the full length of a 1 L cylinder using di↵erent scanning techniques and background subtraction methods. In addition, mean NCT for the slices of a given array (i.e., data from eight slices) were fit to a linear function. The slope and intercept for each array were then plotted as a function of array number to evaluate the variability in NCT due to changes in the anode heel e↵ect caused by increasing x-ray tube load. Tube load characteristics were also compared for the single and multislice systems. For the single slice scanner, background subtraction was performed by subtracting the first averaged image slice from the remaining slices collected, while for the multislice machine, slice-by-slice background subtraction was performed using a second set of images. Each data set was then fit to a linear function to assess the tube load over the entire scanned volume. 2.D.5. Effects of imaging an active gel

Slice-by-slice background subtraction was performed using the additional set of averaged background images at each slice location. The resulting images were then filtered using an adaptive mean filter (3 ⇥ 3, n = 1) and remnant artifact removal using a span = 5, degree = 3 and 2 iterations.17,35,36 Similar to the studies described above, the mean and standard deviation of NCT within a square 50⇥50 pixel ROI at the center of the 1 L cylinder were extracted and used to plot mean NCT as a function of slice position to assess the consistency in NCT over the entire volume of the active gel. The uniformity of gel CT images was assessed over the length of the 1 L cylinder using the grid and ring ROI analyses outlined in Sec. 2.D.2. Uniformity was also evaluated using row and column profiles. Seven adjacent NCT profiles were extracted through the diameter of the 1 L cylinder at 5 slice locations spaced 3.5 cm apart along the length of the cylinder. Each profile was 190 pixels long and 1 pixel wide. The mean and standard deviation of NCT were then computed over the seven pixels in the perpendicular direction of the profiles for both the row and column data.

3. RESULTS AND DISCUSSION 3.A. Anode heel effect and background subtraction

The consistency of NCT across eight slices acquired simultaneously using the multislice scanner is shown in Fig. 2 for images processed using (a) no background subtraction, Medical Physics, Vol. 42, No. 4, April 2015

1670

(b) background subtraction using a single image slice, and (c) background subtraction using the new slice-by-slice technique. Error bars represent the standard deviation of NCT at the center of the 1 L cylinder. Significant variability in NCT is present due to the anode heel e↵ect when no background subtraction is employed and this variability remains after background subtraction is performed using a single slice from the center of the array. Using the new slice-by-slice background subtraction method, the variability in NCT is e↵ectively removed. This can also be seen from the linear fit parameters computed for each subtraction method summarized in Table II along with the corresponding 95% confidence bounds. No change in slope is observed when going from no background subtraction to background subtraction using a single image slice. However, using the new slice-by-slice technique, the slope is reduced by an order of magnitude and is consistent with a horizontal line. Based on the results of Fig. 2 and the computed fit parameters summarized in Table II, it is recommended that background subtraction be performed using the slice-by-slice technique when acquiring gel images using a multislice CT scanner. This method of background subtraction e↵ectively accounts for variations in NCT due to the anode heel e↵ect that may otherwise introduce systematic variations in gel CT measurements. Use of the slice-by-slice background subtraction technique also represents a major change in the CT PGD process as multiple background images must be acquired when using a multislice CT scanner but a single background image may be used when a single slice scanner is employed. The additional image collection required for slice-by-slice background subtraction is not expected to significantly increase scan time as the time required to collect image data is dramatically reduced when using a multislice system. Further discussion on the time requirements for CT PGD using a multislice scanner is given in Sec. 3.D. 3.B. Image noise and uniformity

Figure 3 illustrates the e↵ects of varying the (a) x-ray tube voltage (kV), (b) tube current (mA), (c) gantry rotation time ( t), and (d) slice thickness ( x) on image noise for the multislice CT scanner. As tube voltage increases, CT image noise decreases as (kV) 1.2, while as tube current increases, the image noise decreases as (mA) 0.5. Similarly, image noise was found to depend on the gantry rotation time and slice thickness as ( t) 0.5 and ( x) 0.5, respectively. Increased signal to noise was also observed as slice thickness, tube current, and tube voltage increase for a multislice scanner examined by Hill et al.19 Moreover, since the number of photons (N) incident on the 1 L cylinder increases linearly with tube current, gantry rotation time, and slice p thickness, the relationships found here agree with the 1/ N reduction in image noise predicted by theory.37 Each relationship also agrees with that reported for a typical single slice scanner30 with the exception of tube voltage which has a smaller impact on image noise for the multislice machine. However, as for the single slice system, tube voltage has the greatest influence on

1671

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

1671

Fig. 2. The e↵ect of background subtraction on the anode heel e↵ect. Images were processed using (a) no background subtraction, (b) background subtraction using a central image slice, and (c) the new slice-by-slice technique where the averaged background image at each slice location is subtracted from the averaged test image at the corresponding slice position. Error bars represent the standard deviation of NCT at the center of the 1 L cylinder. The data are fit to a linear function for each background subtraction technique.

CT image noise compared to the other technique parameters examined. As such, the recommendations for reducing image noise for CT PGD using a single slice machine30 can be applied to the multislice system. Specifically, as the x-ray tube load depends on the tube voltage, current, and gantry rotation time equally, the most efficient way to reduce noise in gel CT images is to increase the x-ray tube voltage.30 Figure 4 shows mean NCT plotted as a function of ROI number for (a) 36 individual square ROIs at the center of the 1 L cylinder and (b) 15 concentric ring ROIs extending radially over the 1 L cylinder for one representative CT slice. Error bars represent the standard deviation of NCT for each ROI evaluated. Image uniformity was found to be excellent as both the mean and standard deviation of NCT are consistent over the ROIs positioned throughout the 1 L cylinder. Image uniformity was also consistent across image slices acquired simultaneously [Fig. 4(c)] as Nmean CT for both the grid and ring ROIs of each image agree to within 0.4 HU. The excellent image uniformity found for the multislice scanner using both grid and ring ROI analyses indicates CT measurements using the multislice system are spatially uniform and reliable for CT PGD. Table III summarizes the noise and uniformity measured for the single and multislice scanners. The noise associated with the multislice machine is 69% higher than that determined for the single slice scanner, but no significant di↵erences in image uniformity were found between the two systems. The increase in image noise observed for the multislice scanner is in part explained by the decrease in slice thickness from 3 to 2.5 mm. Using the relationship derived above [i.e., NCT / ( x) 0.5], the increase in noise due to change of slice thickness is expected to be approximately

10%. The additional increase in image noise observed for the multislice machine may be attributed to the larger amount of scattered radiation associated with the multislice scanner cone beam x-ray system.33 Even so, this work has established di↵erences in noise levels in x-ray CT PGD using multislice CT as compared to original single slice CT work.30 Noise di↵erences can a↵ect resulting dose calibration accuracy and precision as reported previously.18 Hence, a full evaluation of error propagation analysis is stressed when implementing a new multislice CT technique. 3.C. Consistency in NCT for multislice scanning

Figure 5 shows mean NCT plotted as a function of slice position (with respect to the CT room coordinate system) for slice thicknesses ranging from 0.625 to 10 mm and their

Table II. Linear fit parameters and corresponding 95% confidence bounds determined by plotting mean NCT as a function of slice position for each background subtraction technique. Background subtraction None Single slice Slice-by-slice

Slope (HU/mm) 0.09 ± 0.03 0.09 ± 0.03 0.004 ± 0.005

Medical Physics, Vol. 42, No. 4, April 2015

Intercept (HU) 0.3 ± 0.2 0.2 ± 0.2 0.15 ± 0.03

Fig. 3. The e↵ects of varying the (a) x-ray tube voltage, (b) tube current, (c) gantry rotation time, and (d) slice thickness on CT image noise for the multislice scanner. The data are fit to a quadratic function for each parameter investigated.

1672

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

1672

Fig. 4. Image uniformity for one representative image slice computed using (a) 36 square ROIs at the center of the water-filled cylinder and (b) 15 concentric ring ROIs extending radially over the cylinder. The standard deviation of mean NCT for the grid and ring ROIs for eight slices is shown in (c) to illustrate the consistency in uniformity for slices acquired simultaneously.

associated detector configurations. Error bars represent the standard deviation of NCT at the center of the 1 L cylinder and therefore the noise ( NCT) present in each image. The 3.75 and 7.5 mm data are plotted together as only one detector configuration is available for each slice thickness. The mean NCT and NCT associated with each image slice of a given detector configuration are consistent across the array for all available slice thicknesses. Mean NCT and NCT are also consistent for image slices acquired using di↵erent detector configurations for the same slice thickness, with the exception of 0.625 mm, where a small o↵set in NCT is observed between the 2i and 16i values. These results can also be seen by examining the range (maximum–minimum value) of NCT and NCT for each detector configuration and slice thickness summarized in Table IV. While an increase in the range of NCT and NCT is observed as the number of images in the array increases for each slice thickness, range values for slice thicknesses between 1.25 and 10 mm vary by less than 0.3 and 0.6 HU for the mean NCT and NCT, respectively. The wider range of NCT and NCT values observed as the number of images in the array increases is likely due to the increase in image noise that results from the additional scattered radiation associated with a larger cone beam of x-rays. As image noise also increases as slice thickness decreases, this e↵ect is more pronounced for the 0.625 mm data. Nevertheless, the results presented here indicate excellent consistency across the detector array for all slice thicknesses, further supporting the removal of the anode heel e↵ect using slice-by-slice background subtraction and indicating multiple images of a gel dosimeter can be reliably acquired simultaneously using various slice thicknesses. These results also indicate the detector configurations for

slice thicknesses between 1.25 and 10 mm can be used interchangeably, allowing some flexibility in the number of images acquired simultaneously for a given experiment. For example, dose information can be collected for a small region of gel using four 2.5 mm slices or a larger region using eight 2.5 mm slices with no significant di↵erences in mean NCT or NCT expected in the resulting images.

Table III. Image noise and uniformity for one image slice (25 averages) for the single and multislice CT scanners. Note images were acquired using 3.0 mm slice thickness for the single slice machine and 2.5 mm thickness for the multislice scanner. Image quality Noise (HU) Grid uniformity (HU) Ring uniformity (HU)

Single slice scanner

Multislice scanner

0.5 0.04 0.05

0.9 0.07 0.06

Medical Physics, Vol. 42, No. 4, April 2015

Fig. 5. Mean NCT across the multislice detector array for slice thicknesses of (a) 0.625 mm, (b) 1.25 mm, (c) 2.5 mm, (d) 5.0 mm, (e) 10.0 mm, and (f) 3.5 and 7.5 mm, and their associated detector configurations.

1673

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

1673

Table IV. The minimum (min), maximum (max), and the range (maximum–minimum) of mean NCT and for each slice thickness and its associated detector configurations. Slice thickness (mm) 0.625

1.25

2.5 3.75

5.0

7.5 10

Images 2 16 1 8 8 16 4 4 8 8 4 1 2 2 4 4 2 1 1 2 2

Detector configuration (mm) 2 ⇥ 0.625 16 ⇥ 0.625 2 ⇥ 0.625 8 ⇥ 1.25 16 ⇥ 0.625 16 ⇥ 1.25 8 ⇥ 1.25 16 ⇥ 0.625 8 ⇥ 2.5 16 ⇥ 1.25 4 ⇥ 3.75 4 ⇥ 1.25 8 ⇥ 1.25 16 ⇥ 0.625 8 ⇥ 2.5 16 ⇥ 1.25 4 ⇥ 3.75 8 ⇥ 1.25 16 ⇥ 0.625 8 ⇥ 2.5 16 ⇥ 1.25

3.D. Effects of increasing tube load

The consistency in NCT over the volume of the 1 L cylinder is illustrated in Fig. 6 for images acquired using [(a) and (b)] axial and [(c) and (d)] 4D CT mode of acquisition. Figures 6(a) and 6(c) show the mean NCT for images processed using sliceby-slice background subtraction with a central array, while for Figs. 6(b) and 6(d), slice-by-slice background subtraction was performed using an additional volume of background images. Error bars represent the standard deviation of NCT at the center of the 1 L cylinder. Mean NCT agree within uncertainty for all acquisition modes and background subtraction methods. Figure 7 shows the linear fit parameters computed for the arrays of each acquisition mode and background subtraction technique of Fig. 6. The slope of the linear fit for each array [shown in Fig. 7(a)] falls within ±0.1 HU and there is less than 0.5 HU variability in the computed intercepts [Fig. 7(b)] across the arrays for all acquisition modes and background subtraction methods. Together, these findings reveal increasing x-ray tube load does not significantly influence NCT when imaging large volumes with the multislice scanner. The results of Figs. 6 and 7 indicate volume scanning can be performed with the multislice system using either axial or 4D CT mode of acquisition. However, based on the tube cooling requirements of the scanner, it is clear that axial mode provides a more stable method of imaging. During each scanning session, CT images were collected over the volume of the 1 L cylinder twice. The scanner was then instructed to perform the same acquisition a third time to qualitatively assess the x-ray tube load by determining how Medical Physics, Vol. 42, No. 4, April 2015

NCT (HU) Min 1.07 0.22 — 0.28 0.40 0.19 0.08 0.12 0.16 0.14 0.15 — 0.19 0.17 0.12 0.08 0.15 — — 0.10 0.17

Max 1.04 0.26 — 0.02 0.01 0.29 0.04 0.01 0.11 0.10 0.11 — 0.30 0.10 0.06 0.12 0.16 — — 0.04 0.29

NCT

NCT

(HU)

Range

Min

Max

Range

0.03 0.48 — 0.30 0.41 0.48 0.04 0.13 0.27 0.24 0.26 — 0.11 0.07 0.18 0.20 0.01 — — 0.06 0.12

9.39 7.92 — 5.85 5.95 5.45 4.29 4.26 3.99 4.05 3.41 — 3.06 3.07 2.99 2.87 2.57 — — 2.19 2.20

9.18 8.78 — 6.22 6.29 6.37 4.40 4.52 4.70 4.51 3.64 — 3.18 3.21 3.26 3.05 2.62 — — 2.24 2.23

0.21 0.86 — 0.37 0.34 0.92 0.11 0.26 0.71 0.46 0.23 — 0.12 0.14 0.27 0.18 0.05 — — 0.05 0.03

soon tube cooling would be required after acquisition of two volumes. While no tube cooling was required at any point during scanning of a third volume using axial mode, the scanner indicated a cooling period would be required after

Fig. 6. Tube load for images collected using axial [(a) and (b)] and 4D CT [(c) and (d)] mode of acquisition. Slice-by-slice background subtraction was performed using a central array of test images in (a) and (c), and a separate volume of background images in (b) and (d).

1674

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

1674

Fig. 7. The (a) slope and (b) intercept computed for each individual array for the data shown in Fig. 6.

collecting seven images in 4D CT mode. Based on this simple test and the fact that the tube will su↵er wear over its lifetime that may induce earlier tube cooling for a given protocol, it is recommended that CT scanning of large volumes be performed using axial mode of acquisition. The results found here also show that slice-by-slice background subtraction can be e↵ectively applied using a single central array of test images or an additional volume of background images. Background subtraction using a central array is preferred as it allows both test and background CT images to be acquired from the same gel dosimeter which can significantly reduce scan time. However, as both calibration and test dose distributions are delivered to the same cylinder, the central region of the gel is exposed to a low dose of scattered radiation that can vary significantly across the central region. This does not pose a problem when using a single slice scanner as one background slice is subtracted from each test slice and the scattered dose influences each slice in the same way. When using the multislice scanner, variations in scatter dose across background slices could cause di↵erences in NCT that are large enough to produce inaccurate dose measurements. For example, di↵erences in mean NCT of up to 20% were found between the first and last slices in an 8-slice array acquisition for a representative gel dosimeter irradiated with both calibration and test irradiations. As such, it is recommended that slice-by-slice background subtraction be performed using a separate set of images acquired over the volume of a blank gel. Based on the recommendations given above, a comparison was made between the tube load characteristics of the single and multislice machines. The imaging time associated with the multislice scanner is significantly lower compared to the single slice system. Over 56 min are required to image 6 cm of the 1 L cylinder with 25 images acquired at each slice position when using the single slice machine (scanning was terminated after 6 cm due to lengthy tube cooling periods). However, imaging the entire 12 cm length of the cylinder twice using the multislice system requires only 12 min. As such, an approximately 4-fold increase in scanning efficiency is achieved in using the multislice scanner. Moreover, CT imaging of polymer gels using the multislice system is 1.5–20 Medical Physics, Vol. 42, No. 4, April 2015

times faster than scanning times reported for optical-CT1,38 and up to 50 times faster than acquiring images using MRI.1 This finding clearly demonstrates the exceptional efficiency of CT for gel imaging compared to other readout techniques. The variation in NCT observed as the number of images collected increases was also compared for the single and multislice systems. Figure 8 shows mean NCT plotted as a function of sequential slice number for the (a) single and (b) multislice scanners. For the single slice machine, the mean NCT varies as a function of slice number according to Eq. (1), while for the multislice machine, Eq. (2) describes the change in NCT as the number of averaged slices increases. single

NCT

= (0.158 ± 0.005) HU · slice + (0.23 ± 0.06) HU, (1)

multi NCT = (0.009 ± 0.002) HU · slice + (0.05 ± 0.04) HU. (2)

It is clear from the slope values of Eqs. (1) and (2) that variations in NCT due to increasing x-ray tube load are significantly reduced for the multislice scanner and will have a negligible impact on dose measurements compared to the single slice machine. The improvement in NCT constancy associated with the multislice scanner is likely due to the increase in anode heat capacity (see Sec. 2.A) and reduction in wear of its x-ray tube compared to the single slice scanner. This finding, together with the dramatic increase in scanning efficiency, indicates multislice scanning is the

Fig. 8. The tube load associated with (a) the single slice scanner and (b) the multislice system when images are acquired using the reference protocol and 25 image averages.

1675

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

superior imaging tool for CT PGD compared to single slice imaging in terms of x-ray tube load. As such, multislice CT scanning is the recommended imaging technique for CT PGD when performing dosimetry using large gel volumes. 3.E. Effects of imaging an active gel

Figure 9 illustrates the consistency in (a) NCT, (b) image uniformity computed using grid ROI analysis, and (c) uniformity determined using ring ROIs over the length of the active gel. Slices acquired near I60 correspond to the top of the 1 L cylinder (i.e., the region closest to the lid) while slices collected near S60 are located near the bottom of the cylinder. It is clear from Fig. 9(a) that the consistency in NCT is excellent over most of the active gel as mean NCT agree within uncertainty for all image slices between I56.25 and S56.25. However, beyond S40, mean NCT systematically decrease as the image slices approach the bottom of the cylinder and mean NCT for the last two image slices no longer agree with the slices more inferior in the gel. Similar behavior is observed in the image uniformity computed using both grid and ring ROIs [Figs. 9(b) and 9(c), respectively]. In both cases, Nmean CT for the corresponding ROIs vary by less than 0.2 HU over the length of the gel, but image uniformity changes as the image slices approach the bottom of the cylinder. In the case of the grid ROI analysis, there is consistent variability in Nmean CT computed between I56.25 and S23.75, but beyond S23.75, the standard deviations become relatively constant. For the ring ROI analysis, a systematic decrease in the Nmean CT is observed over the entire length of the cylinder, with a more rapid decrease seen beyond S40, similar to the mean NCT values. These findings indicate NCT and image uniformity for an active, unirradiated gel are consistent across all image slices collected inferior of S40, or equivalently, ⇠2 cm from the bottom of the cylinder. This is in contrast to the results found for a water-filled cylinder [Fig. 6(d)], where mean NCT are consistent over the entire phantom, indicating this e↵ect is not due to increasing x-ray tube load. Moreover, any potential change in NCT due to CT readout exposure is negligible for scans completed in less than 15 min.15 Since the entire gel was imaged in under 6 min, recording of CT dose as the scan proceeded is also not the source of the deviation

1675

in NCT toward the bottom of the cylinder. However, it is possible that the observed e↵ect is due to the settling of gelatin molecules as the gel solidifies while standing upright in the refrigerator. As such, the e↵ect may be mitigated by adjusting the fabrication procedure to include agitation of the gel during or immediately before cooling. Additional details on gel manufacture are provided in previous work.16,17 Adjustments to the fabrication technique are the subject of current study. Further investigation was performed to evaluate the image uniformity of the active gel using profiles extracted through the diameter of the 1 L cylinder. Figure 10 shows NCT maps of the gel with corresponding row and column profiles for slices located at [(a) and (b)] I48.75, [(c) and (d)] I13.75, [(e) and (f)] S21.25, and [(g) and (h)] S56.25 along the length of the 1 L cylinder. Error bars represent the standard deviation of NCT computed over seven pixels in the perpendicular direction of each profile, and mean NCT values plotted at 50 mm correspond to the left and top of the 1 L cylinder. Each NCT map shown on the left illustrates NCT decrease radially from the center of the cylinder for all image slices. This is also illustrated in the row and column profiles shown on the right with di↵erences in NCT between the center and periphery of the cylinder of up to 2 HU and 7 HU for the row and column profiles, respectively. However, the e↵ect is much less pronounced in the central region of the image collected at the bottom of the cylinder (S56.25). The di↵erence in the image characteristics of the active gel toward the bottom of the cylinder is consistent with the results shown in Fig. 9. The cause of the variability in NCT over the diameter of the gel is presently unknown. The observed variability has the appearance of the familiar cupping artifact and multislice configurations, with increased scatter, may produce such artifacts. However, the excellent image uniformity illustrated in Fig. 4 of the paper for a water phantom demonstrates cupping artifacts are e↵ectively removed using the sliceby-slice image background subtraction technique. Moreover, similar radial variability is observed for an irradiated NIPAM gel imaged using MRI (Ref. 39) but is absent in a study examining small field dosimetry using NIPAM gel and CT readout.40 This suggests the variability in NCT found in this work is not the result of cupping artifacts but may be due to interaction of gel components during fabrication and storage.

Fig. 9. The consistency in (a) mean NCT and uniformity determined using (b) grid ROI analysis and (c) ring ROIs across the volume of the active blank gel. Medical Physics, Vol. 42, No. 4, April 2015

1676

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

1676

4. CONCLUSIONS

Fig. 10. Maps of NCT and row and column profiles through the diameter of the active blank gel for images acquired at [(a) and (b)] I48.75, [(c) and (d)] I13.75, [(e) and (f)] S21.25, and [(g) and (h)] S56.25.

A recent study by Vandecasteele and De Deene,5 reported di↵erences of more than 10% across an unirradiated normoxic polyacrylamide gel exposed to a temperature gradient of ⇠2.8 /cm as it solidified. The observed variation was attributed to an interaction between gelatin and the antioxidant in the gel system. It is possible that a temperature gradient present during storage introduced a radial variation in gel density across the active gel examined in this work. As such, it may be possible to mitigate radial variations in NCT by reducing temperature gradients during storage through cooling the gel more slowly in a warmer environment. Nevertheless, further study is required to determine the exact mechanism responsible for the variability in NCT observed across the diameter of the gel. Medical Physics, Vol. 42, No. 4, April 2015

This study presents work undertaken to evaluate multislice CT imaging for CT PGD. A new slice-by-slice background subtraction technique was developed to remove image artifacts, which involves subtracting the background image at a given slice from the test image at the corresponding slice in the detector array. The new technique was found to e↵ectively remove variations in NCT due to the anode heel e↵ect and is the recommended method for performing background subtraction when images are acquired using a multislice scanner. Further study revealed the relationships between image noise and the x-ray tube voltage and current, gantry rotation time and slice thickness are consistent with those found for single slice machines with the exception of tube voltage which has less influence on noise for the multislice system. This finding indicates the recommendations for optimizing image quality for CT PGD using a single slice machine also apply to multislice CT scanners. Image noise was higher for images collected using the multislice system compared to the single slice machine but remains acceptable for reliable CT PGD dose measurements. Image uniformity was consistent between the two scanners. In addition, NCT and image noise were found to be consistent for image slices acquired simultaneously using the multislice detector array for all slice thicknesses and their associated detector configurations, indicating multiple images can be reliably collected at the same time. It was also found that the detector configurations for slice thicknesses between 1.25 and 10 mm can be used interchangeably, allowing some flexibility in the number of images acquired simultaneously for a given experiment. Additional work was performed to examine the e↵ects of x-ray tube load which was determined to have a negligible influence on NCT when imaging large gel volumes. Due to the tube load associated with 4D CT scanning and the scattered radiation dose delivered to the center of a gel in a typical CT PGD experiment, it is recommended that volume imaging be performed using axial mode of acquisition and slice-by-slice background subtraction with a separate set of background images. Using these recommendations, imaging a large gel volume with the multislice system reduces the scanning time by up to 4 times when compared to imaging using the single slice machine. Finally, images of an active gel that was not irradiated indicate deviations in NCT occur at the bottom of the 1 L cylinder that may be due to the settling of gelatin molecules and NCT can vary across the diameter of the cylinder by up to 7 HU. ACKNOWLEDGMENTS The authors would like to acknowledge financial support from NSERC. The authors would also like to acknowledge Dave Yardley and Steven Gray for help with phantom design and construction as well as Warren Campbell and Derek Wells for insightful discussions.

1677

Johnston, Hilts, and Jirasek: Multislice imaging in x-ray CT polymer gel dosimetry

a)Author

to whom correspondence should be addressed. Electronic mail: [email protected]; Present address: Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, TX. b)Present address: Medical Physics, BC Cancer Agency, Sindi Ahluwalia Hawkins Centre for the Southern Interior, Kelowna, BC, Canada. 1C. Baldock, Y. De Deene, S. Doran, G. Ibbott, A. Jirasek, M. Lepage, K. B. McAuley, M. Oldham, and L. J. Schreiner, “Polymer gel dosimetry,” Phys. Med. Biol. 55, R1–R63 (2010). 2M. J. Maryanski, J. C. Gore, R. P. Kennan, and R. J. Schulz, “NMR relaxation enhancement in gels polymerized and cross-linked by ionizing radiation: A new approach to 3D dosimetry by MRI,” Magn. Reson. Imaging 11, 253–258 (1993). 3Y. De Deene, “Review of quantitative MRI principles for gel dosimetry,” J. Phys.: Conf. Ser. 164, 012033 (2009). 4J. Vandecasteele and Y. De Deene, “On the validity of 3D polymer gel dosimetry: I. Reproducibility study,” Phys. Med. Biol. 58, 19–42 (2013). 5J. Vandecasteele and Y. De Deene, “On the validity of 3D polymer gel dosimetry: II. Physico-chemical e↵ects,” Phys. Med. Biol. 58, 43–61 (2013). 6J. Vandecasteele and Y. De Deene, “On the validity of 3D polymer gel dosimetry: III. MRI-related error sources,” Phys. Med. Biol. 58, 63–85 (2013). 7J. C. Gore, M. Ranade, M. J. Maryanski, and R. J. Schulz, “Radiation dose distributions in three dimensions from tomographic optical density scanning of polymer gels: I. Development of an optical scanner,” Phys. Med. Biol. 41, 2695–2704 (1996). 8S. J. Doran, “The history and principles of optical computed tomography for scanning 3-D radiation dosimeters: 2008 update,” J. Phys.: Conf. Ser. 164, 012020 (2009). 9L. Rankine and M. Oldham, “On the feasibility of optical-CT imaging in media of di↵erent refractive index,” Med. Phys. 40, 051701 (8pp.) (2013). 10W. G. Campbell, D. A. Rudko, N. A. Braam, D. M. Wells, and A. Jirasek, “A prototype fan-beam optical CT scanner for 3D dosimetry,” Med. Phys. 40, 061712 (12pp.) (2013). 11M. Hilts, C. Audet, C. Duzenli, and A. Jirasek, “Polymer gel dosimetry using x-ray computed tomography: A feasibility study,” Phys. Med. Biol. 45, 2559–2571 (2000). 12B. Hill, A. J. Venning, and C. Baldock, “A preliminary study of the novel application of normoxic polymer gel dosimeters for the measurement of CTDI on diagnostic x-ray CT scanners,” Med. Phys. 32, 1589–1597 (2005). 13J. Hindmarsh, R. Fulton, L. Oliver, and C. Baldock, “Polymer gel dosimetry using x-ray computed tomography: Investigation of the e↵ect of reconstruction technique,” J. Phys.: Conf. Ser. 250, 012073 (2010). 14M. B. Kakakhel, T. Kairn, J. Kenny, and J. V. Trapp, “Improved image quality for x-ray CT imaging of gel dosimeters,” Med. Phys. 38, 5130–5135 (2011). 15H. Johnston, M. Hilts, J. Carrick, and A. Jirasek, “An x-ray CT polymer gel dosimetry prototype: II. Gel characterization and clinical application,” Phys. Med. Biol. 57, 3155–3175 (2012). 16J. N. M. Chain, A. Jirasek, L. J. Schreiner, and K. B. McAuley, “Cosolventfree polymer gel dosimeters with improved dose sensitivity and resolution for x-ray CT dose response,” Phys. Med. Biol. 56, 2091–2102 (2011). 17A. Jirasek, J. Carrick, and M. Hilts, “An x-ray CT polymer gel dosimetry prototype: I. Remnant artefact removal,” Phys. Med. Biol. 57, 3137–3153 (2012). 18A. Jirasek and M. Hilts, “Dose calibration optimization and error propagation in polymer gel dosimetry,” Phys. Med. Biol. 59, 597–614 (2014). 19B. Hill, A. J. Venning, and C. Baldock, “The dose response of normoxic polymer gel dosimeters measured using X-ray CT,” Br. J. Radiol. 78, 623–630 (2005). 20B. Hill, A. J. Venning, and C. Baldock, “Polymer gel dosimetry on a multislice computed tomography scanner: E↵ect of changing parameters on CTDI,” Phys. Med. 24, 149–158 (2008).

Medical Physics, Vol. 42, No. 4, April 2015 View publication stats

1677

21A. Jirasek and M. Hilts, “An overview of polymer gel dosimetry using x-ray

CT,” J. Phys.: Conf. Ser. 164, 012038 (2009). V. Trapp, S. A. J. Back, M. Lepage, G. Michael, and C. Baldock, “An experimental study of the dose response of polymer gel dosimeters imaged with x-ray computed tomography,” Phys. Med. Biol. 46, 2939–2951 (2001). 23M. Hilts, A. Jirasek, and C. Duzenli, “E↵ects of gel composition on the radiation induced density change in PAG polymer gel dosimeters: A model and experimental investigations,” Phys. Med. Biol. 49, 2477–2490 (2004). 24S. Brindha, A. J. Venning, B. Hill, and C. Baldock, “Experimental study of attenuation properties of normoxic polymer gel dosimeters,” Phys. Med. Biol. 49, N353–N361 (2004). 25R. J. Senden, P. De Jean, K. B. McAuley, and L. J. Schreiner, “Polymer gel dosimeters with reduced toxicity: A preliminary investigation of the NMR and optical dose-response using di↵erent monomers,” Phys. Med. Biol. 51, 3301–3314 (2006). 26V. I. Koeva, E. S. Csaszar, R. J. Senden, K. B. McAuley, and L. J. Schreiner, “Polymer gel dosimeters with increased solubility: A preliminary investigation of the NMR and optical dose-response using di↵erent crosslinkers and co-solvents,” Macromol. Symp. 261, 157–166 (2008). 27V. I. Koeva, T. Olding, A. Jirasek, L. J. Schreiner, and K. B. McAuley, “Preliminary investigation of the NMR, optical and x-ray CT dose–response of polymer gel dosimeters incorporating cosolvents to improve dose sensitivity,” Phys. Med. Biol. 54, 2779–2790 (2009). 28A. Jirasek, M. Hilts, A. Berman, and K. B. McAuley, “E↵ects of glycerol co-solvent on the rate and form of polymer gel dose response,” Phys. Med. Biol. 54, 907–918 (2009). 29A. Jirasek, M. Hilts, and K. B. McAuley, “Polymer gel dosimeters with enhanced sensitivity for use in x-ray CT polymer gel dosimetry,” Phys. Med. Biol. 55, 5269–5281 (2010). 30M. Hilts, A. Jirasek, and C. Duzenli, “Technical considerations for implementation of x-ray CT polymer gel dosimetry,” Phys. Med. Biol. 50, 1727–1745 (2005). 31S. Mutic, J. R. Palta, E. K. Butker, I. J. Das, M. S. Huq, L.-N. D. Loo, B. J. Salter, C. H. McCollough, and J. Van Dyk, “Quality assurance for computedtomography simulators and the computed-tomography-simulation process: Report of the AAPM Radiation Therapy Committee Task Group No. 66,” Med. Phys. 30, 2762–2792 (2003). 32M. de Denaro and P. Bregant, “Dosimetric evaluation of a 320 detector row CT scanner unit,” Radiol. Oncol. 45, 64–67 (2011). 33M. Endo, T. Tsunoo, N. Nakamori, and K. Yoshida, “E↵ect of scattered radiation on image noise in cone beam CT,” Med. Phys. 28, 469–474 (2001). 34M. Hilts, J. Carrick, H. Johnston, and A. Jirasek, “A CT polymer gel dosimetry system for end-to-end dosimetry,” Med. Phys. 38, 3507 (2011). 35M. Hilts and C. Duzenli, “Image filtering for improved dose resolution in CT polymer gel dosimetry,” Med. Phys. 31, 39–49 (2004). 36M. Hilts and A. Jirasek, “Adaptive mean filtering for noise reduction in CT polymer gel dosimetry,” Med. Phys. 35, 344–355 (2008). 37J. Hsieh, Computed Tomography Principles, Design, Artifacts, and Recent Advances, 2nd ed. (Society of Photo-Optical Instrumentation and Engineers, Bellingham, 2009). 38T. Olding, O. Holmes, and L. J. Schreiner, “Cone beam optical computed tomography for gel dosimetry I: Scanner characterization,” Phys. Med. Biol. 55, 2819–2840 (2010). 39Y.-L. Chen, B.-T. Hsieh, C.-M. Chiang, C.-T. Shih, K.-Y. Cheng, and L.-L. Hsieh, “Dose verification of a clinical intensity-modulated radiation therapy eye case by the magnetic resonance imaging of N-isopropylacrylamide gel dosimeters,” Radiat. Phys. Chem. 104, 188–191 (2014). 40C.-Y. Chiu, Y.-W. Tsang, and B.-T. Hsieh, “N-isopropylacrylamide gel dosimeter to evaluate clinical photon beam characteristics,” Appl. Radiat. Isot. 90, 245–250 (2014). 22J.