CHAPTER 2 COMMISSIONING OF KILO-VOLTAGE CONE BEAM COMPUTED TOMOGRAPHY FOR IMAGE-GUIDED RADIOTHERAPY

14 CHAPTER 2 COMMISSIONING OF KILO-VOLTAGE CONE BEAM COMPUTED TOMOGRAPHY FOR IMAGE-GUIDED RADIOTHERAPY 2.1 INTRODUCTION kv-cbct integrated with linear accelerators as a tool for IGRT, was developed to acquire online volumetric and anatomical images (Xing 2006). The successful implementation of CBCT requires a quality assurance (QA) program that not only involves robust and intuitive quality control checks of the equipment, but also thoughtful image-guidance processes based on measurement of positional uncertainties and considering the practicalities of the patient throughput. This chapter discusses about the commissioning of kv-cbct system for the clinical use. It addresses a series of tests to ensure the CBCT system is safe both mechanically and dosimetrically for the patient use. 2.2 METHODS AND MATERIALS 2.2.1 On-Board Imaging System The on-board imager (OBI) integrated in a Clinac 2100 C/D medical linear accelerator (Varian Medical Systems, Palo Alto, CA) is used in this work (Figure 2.1). The OBI system consists of a kv X-ray source and an amorphous silicon (asi1000) digital imaging detector mounted on the linear accelerator using robotic arms (Exact ), which are orthogonal to the

15 electronic portal imaging device (asi1000, PortalVision ). OBI system provides has capability of two-dimensional (2D) radiographic acquisition, fluoroscopic image acquisition, and CBCT acquisition. Figure 2.1 OBI hardware mounted on a Clinac 2100 C/D linear accelerator A 150 kv x-ray tube with maximum 32 ms pulse length for continuous irradiation and maximum 320 ms pulse length for single pulse is designed for generating high resolution images from a moving gantry. The focal spot of the tube is located at 90 to the mega-voltage (MV) source and 100 cm from the radiation axis of the accelerator. A 39.7 cm 29.8 cm asi1000 image detector acquires digital images with a pixel matrix of 2048 1536.

16 Using the OBI system, the CBCT data can be acquired in two modes: a full-fan mode and half-fan mode. In the full-fan mode, the beam central axis passes through the detector centre and a full projection of the scanned patient is acquired for each acquirement angle. A total of 650 to 700 projections are taken during the whole 364 gantry rotation with a maximum field of view (FOV) about 25 cm in diameter and 17 cm in length. The data acquisition time is about 60 s and the reconstruction time for 340 slices of 512 512 CBCT images with a voxel size of 0.5 mm 3 is also about 1 min. The half-fan mode is designed to obtain a larger FOV. In this mode, the detector is shifted laterally by 14.8 cm to take only half of the projection of the scanned patient for each acquirement angle. In half-fan mode, a FOV of 45 cm in the axial plane and 15 cm in the longitudinal direction can be achieved. The effects of X-ray scatter and artifacts are larger in CBCT images than in CT images. A bowtie filter mounted to the X-ray tube improved image quality because it reduces intensity variations across the detector and charge trapped in the detector. 2.2.2 Commissioning Procedure The test categories, which are introduced in detail in the subsections that follow, were i. System mechanical safety ii. iii. iv. Geometric accuracy (agreement of MV and kv beam isocenter) CBCT image quality Registration and couch shift accuracy, and v. Dose to the patient

17 2.2.2.1 System mechanical safety The safety QA test checks the safety features built into the OBI system. As per manufacturer s recommendation, tube warm-up is performed before patient treatment to prevent premature failure of the x-ray tube. While the x-ray is on for the tube warm-up, the door interlock, warning lights, and audible warning can be checked. After 20 s of the warm-up, the arms can be retracted using the hand pendant in the room. While retracting the arms, the motion-enable bars of the hand pendant to be released to verify that arms stop moving. The collision detection covers and paddles can also be checked, one at a time, while retracting the arms. 2.2.2.2 Geometric accuracy (agreement of MV and kv beam isocenter) Alignment of the isocenters for the kv imaging and the MV treatment system is crucial for accurate patient positioning, because the kv imaging system is used to position the patient with respect to the MV treatment system. The OBI isocenter accuracy quality assurance test evaluates whether the digital graticule generated by the OBI application coincides with the treatment (MV) isocenter. This check is performed using marker block with one fiducial marker at the center and four markers at known locations inside the block. A test patient was created with a plan containing two kv setup fields. The setup field had a digitally reconstructed radiograph (DRR) associated with it so that the entire patient reposition process could be performed. The phantom was placed on the couch using field light cross-hair and the wall lasers so that the center of the phantom was aligned with the treatment isocenter. First, agreement between the center fiducial marker and the isocenter of the MV beam is established by image-supported adjustment of the position of the marker block. Eight MV portal images are taken at the

18 four cardinal gantry angles and at collimator settings -90 degrees and 90 degrees. Based on image analysis by system software, the deviation of the location of the center fiducial marker from the location of the MV beam isocenter is calculated. The position of the marker is adjusted, and the images are retaken. The process is repeated until the deviation in any directions is below 0.25 mm, the threshold suggested by the manufacturer. Then, the deviation of the fiducial marker from the center of the kv imaging system is determined. Four images are taken with the kv system at four cardinal gantry angles. On each image, pixel locations are used to determine the deviation between the center of the image (automatically marked with digital crosshairs by the software) and the center of the fiducial marker in the image. The horizontal and vertical differences between the observed center of the fiducial marker and the crosshairs are calculated for the image. These measurements and subtractions are performed for each of the four images. Based on other tolerances in the QA process of linear accelerators, the deviation limit for this test was set to 1.04 mm in each direction, which corresponds to 4 pixels. The kv image pixel size corresponds to 0.259 mm at the isocenter for this high resolution acquisition mode. 2.2.2.3 CBCT image quality The image quality achievable with the CBCT imaging system was tested for maximal achievable resolution and ability to display low-contrast objects. A Catphan 504 phantom (The Phantom Laboratory, Salem, NY) was used for image quality measurements. The phantom consists of various cylindrical sections (modules), each of which is designed for a specific test. The Catphan phantom was hung over the end of the couch with the aid of its wooden case and leveled. The phantom was centered so that its physical center was at the treatment isocenter and one full-fan and half-fan CBCT scan of the phantom with scan parameters 125 kvp, 80 ma, 25 ms, 150 cm source

19 to imager distance (SID) with a bowtie filter was acquired. The reconstruction matrix was 512 512 and the slice thickness was 2.5 mm. From the scans, all image quality evaluations were made. (a) Hounsfield Units (HU) linearity The CTP 404 insert contains seven tissue substitute materials of different densities. The materials and the expected HU values are listed in the Catphan phantom manual. A region of interest was selected using the Area profile tool to measure the mean HU value for each material in the CBCT application. The region of interest (ROI) size was adjusted to cover approximately 7 mm 7 mm, which covered about 14 14 pixels for the full-fan mode and 8 8 pixels for the half-fan mode. The measured HU value for each material should be within ± 40 HU of the value indicated in the Catphan manual. (b) Low contrast resolution In the CTP 515 insert, there are three groups of nine supra-slice disks with diameters ranging from 2-15 mm and subject contrasts of 1%, 0.5%, and 0.3%. The image is visually inspected and the lowest contrast and lowest diameter supra-slice disk that was determined. The recommendation is that the 1%, 7 mm disk should be visible [Varian CBCT Customer Acceptance Procedure]. (c) Spatial resolution (high contrast resolution) The CTP 528 insert contains a spatial resolution rule with bar patterns between 1 to 21 line pairs/cm (lp/cm). The least discernable bar pattern in CBCT images should be selected and compared to the baseline obtained at the time of commissioning and recommendation is that 6 lp/cm [Varian CBCT Customer Acceptance Procedure], which is group 6 should be visible.

20 (d) HU uniformity The CTP 486 insert is a uniform disk of 20 cm diameter that has been used to access HU uniformity. In CTP 486, the image slice containing the white markers is selected for the measurement of HU uniformity. Five ROI s one in center and four symmetrically in the upper, lower, left and right peripheral areas in the phantom were selected. Using area profile tool, the HU values in 1 cm 1 cm ROI was measured. The HU values for the all regions should be within ± 40 HU of each other [Varian CBCT Customer Acceptance Procedure]. (e) In slice spatial linearity The geometric accuracy of the CBCT scans is determined by the magnification accuracy as that defines the divergence of the x-ray beam. If the actual divergences of the x-ray beam matches that assumed in the reconstruction, then the CBCT images will be geometrically consistent. The in-slice spatial linearity can be checked using the CTP 404 insert. There are four rods of 3 mm diameter at the corners of a 5 cm square in the CTP 404 module. The distance between pair of rods in the axial view using the distance measuring tool in the CBCT application was measured and averaged. The tolerance of less than 1% of difference (± 0.5 mm) is recommended [Varian CBCT Customer Acceptance Procedure]. 2.2.2.4 Registration and couch shift accuracy The CBCT system s ability to correctly register a localization geometry with a reference geometry was tested using anthropomorphic phantom (Rando phantom) that went through the entire patient planning chain. Several methods of performing such tests have been presented in the literature (Letourneau 2005, Oldham 2005 and Sharpe 2006).

21 In this test, the algorithm for repositioning the patient at the isocenter was tested to show that the system can precisely detect a preset deviation and determine the appropriate shift back to alignment. The head and neck part of anthropomorphic phantom was used for the test. With the CBCT system, alignment of the phantom on the treatment couch and its position relative to isocenter were first brought into agreement with a reference image (DRR) from a previous CT scan. The phantom was then moved by specific distances in one or more directions relative to the reference position. Each movement of the phantom away from the initial start position was measured by the couch digital readout (1 mm resolution). The movements were each 10 mm in magnitude. The phantom was first moved right, then additionally up and finally in. The ability of the OBI system to accurately describe these moves was analyzed for each step. At the end, the phantom was moved back to isocenter using the repositioning directions generated by the algorithm. Following the repositioning, another CBCT was taken and compared with the reference image. 2.2.2.5 Dose to the patient The dose to the patients from a CBCT scan depends significantly on the scanning parameters used. In addition, based on scanning technique (full-fan or half-fan), the dose across the patient may vary. Various authors has already been reported the patterns of dose distribution and the impacts of the various factors on the dose (Letourneau 2005, Islam 2006 and Sawyer 2009). As a part of quality assurance, the dose measurement for a 360 degree CBCT and measurement of the half-value layer of the kv beam was done. Doses were measured at the centre and periphery (1.2 cm from the edge) of a 32 cm diameter cylindrical perspex phantom with 9.5 cm thickness. It should be noted that the limited thickness of computed tomography dose index (CTDI) phantoms can significantly underestimate the contribution of scatter (Perisinakis 2007). A 0.6 cc Farmer type ionization chamber calibrated in the kv energy range was used for the dose measurements in conjunction

22 with a NE 2670 electrometer. The measurements were performed for the scanning parameters 125 kv, 80 ma and 25 ms per frame for a 360 degree scan. The half value layer of the kv beam was determined using a stationary beam with aluminum plates and standard procedures (Ma 2001). 2.3 RESULTS 2.3.1 System mechanical safety The beam termination at the control console and the door interlock were functional. A not-fully-extended kv source arm and an open door each inhibited engagement of the beam. All touch guards were found to work correctly, in that they inhibited motion when activated. 2.3.2 Geometric accuracy (agreement of MV and kv beam isocenter) The maximum deviation of the centre fiducial marker from the MV isocenter was below the threshold of 0.25 mm after two iterations of adjustment. Analysis of the subsequent kv images determined that the maximal deviation between the kv isocenter and the center fiducial marker was 1 mm or less in each orthogonal direction. Figure 2.2 shows an example of the fiducial marker in marker phantom aligned with the center of the kv images. Figure 2.2 Orthogonal kilovoltage images of the marker phantom

23 2.3.3 CBCT image quality (a) Hounsfield Units linearity In full-fan mode, the maximum differences in HU value compared with reference CT were less than 20 HU. For the half-fan mode, teflon material shows the maximum HU difference from CT of 40 HU. Figure 2.3 (a) shows the HU profile measured along the Catphan phantom. (b) Contrast resolution Figure 2.3 (b) shows CBCT images of CTP 515 insert, which measures low contrast resolution. In 1% supraslice group, 6 mm and 7 mm disks are clearly discernable in full-fan mode and half-fan mode respectively. The images acquired using a full-fan mode have better contrast resolution than those acquired using half-fan mode. This is because, the volume irradiated in full-fan mode receives a higher dose than the volume irradiated by a half-fan mode. (c) Spatial resolution (high contrast resolution) Figure 2.3 (c) shows a CBCT image of the CTP 528 insert of Catphan phantom. The images obtained using full-fan and half-fan mode scan could able to resolve the 7 lp/cm and 6 lp/cm spatial resolution respectively. As the pixel size in full-fan mode was smaller than that in the half-fan mode, the images acquired using full-fan mode had a better spatial resolution than half-fan mode images. (d) HU uniformity In CTP 486 insert of Catphan phantom, the mean HU values in five regions of interest (ROI: 1 cm 1 cm) were measured to determine HU uniformity. All HU values measured in five ROI for full-fan and half-fan mode acquisition were within the specification of ± 40 HU. The HU values in peripheral regions were found be lesser than that of middle of phantom. The

24 images acquired using full-fan mode had more homogeneous HU values than those acquired using half-fan mode. (e) In-slice spatial linearity Figure 2.4 (d) displays an image of the CTP 404 insert, which can be used for the spatial linearity test. The four rods are shown and the two lines generated using the measure distance tool indicate exactly 5 cm separation between the two upper rods and the two right rods. Figure 2.3 Sample images of Catphan 504 phantom imaged using kv-cbct in full-fan mode: (a) HU profile comparison (b) Contrast resolution test (c) Spatial resolution test (d) Spatial linearity test

25 2.3.4 Registration and couch shift accuracy The results of this test are summarized in Table 2.1. It shows the initial positional deviations of the head and neck part of Anthropomorphic phantom about the linear accelerator isocenter after alignment to the DRR of previous CT scan as Step 1. Steps 2 4 are the cumulative shifts of 10 mm each in the lateral direction (Step 2, patient s right), the vertical direction (Step 3, couch up), and the longitudinal direction (Step 5, couch in). The data in the corresponding columns on the right show that the algorithm detected translational shift for every move with an accuracy of better than 1 mm, which is the equivalent of the tolerance of the couch. The rotational error (which should be zero, because no rotations were introduced) stayed within 0.3 degree. To verify the reproducibility of the system, two more scans and image alignments were performed without moving the phantom (Steps 5 and 6). The fluctuation found in this test was within 0.1 mm and 0.1 degree for translation and rotation respectively. In the final step, Step 7, the phantom was moved back to isocenter using the repositioning directions generated by the algorithm. The data show that the phantom was brought within 0.4 mm and 0.4 degree of the starting point (Step 1). Figure 2.5 shows a transverse and sagittal view of the phantom after registration. Table 2.1 Step Position errors reported by the registration following defined moves of the phantom Defined phantom movement 1 Initial position 2 3 4 lat: -10 mm (1 cm patient s right) vrt: -10 mm (additional 1 cm up) lng: 10 mm (additional 1 cm in) Position error after 3D-3D match Translation Rotation vrt : 0.4 pitch : 0.2 lng : 0.3 roll : 0.2 lat : 0.3 rtn : 0.3 vrt : 0.3 lng : 0.2 lat : -9.9 vrt : -10.2 lng : 0.1 lat : -10.0 vrt : -10.2 lng : 9.9 lat : -9.9 pitch : 0.3 roll : 0.2 rtn : 0.2 pitch : 0.1 roll : 0.2 rtn : -0.1 pitch : 0.2 roll : 0.2 rtn : -0.1

26 Table 2.1 (Continued) Step 5 6 7 Defined phantom movement No movement (Reproducibility check) No movement (Reproducibility check) vrt: 10 mm, lng: -9.9 mm, lat: -10.1 Position error after 3D-3D match vrt : -10.1 lng : 9.9 lat : 10.0 vrt : -10.1 lng : 9.9 lat : 10.1 vrt : 0.1 lng : -0.1 lat : 0.2 pitch : 0.2 roll : 0.3 rtn : 0.0 pitch : 0.2 roll : 0.2 rtn : -0.1 pitch : 0.2 roll : 0.1 rtn : -0.1 Figure 2.4 A transverse and sagittal view of the phantom after registration. The selected display type is chess view, showing alternating panels of reconstructed and reference image. 2.3.5 Dose to the Patient For full-fan mode with the scanning parameters 125 kv, 80 ma, 40 ms per frame, for a 360 degree scan (~650 projections), the dose in the CTDI phantom was 23.4 mgy at the isocentre and 53.1 mgy at the periphery. This can be compared with a value of 54 mgy reported for the same procedure in

27 the literature (Song 2008). With bowtie filter, the half-value layer of the beam was found to be 6.1 mm Al. 2.4 DISCUSSION The clinical introduction of kv-cbct based IGRT systems necessitates formal commissioning of hardware and image-guided processes to be used. Because, positioning of the patient is based directly on data from the CBCT systems, their performance has a major effect on the outcome of treatment. The accuracy and reliability of CBCT systems therefore need to be tested and related to familiar tolerances of the treatment process. This study details about the commissioning of kv-cbct integrated into Clinac 2100 C/D linear accelerator. The mechanical safety of the OBI system was verified. The significant point of concern is the agreement of the kv imaging line isocenter and the MV treatment line isocenter. In the commissioning, this agreement was found to be within 1 mm. The importance of this agreement has already been reported by others (Sharpe 2006 and Oldham 2005). The test of CBCT image quality only used the Catphan phantom. However, the CBCT images of a large patient can result in more artifacts due to phenomenon such as x-ray scatter and charge trapping than would occur with the Catphan phantom. Therefore it is advisable to acquire the CBCT images of an anthropomorphic phantom on a routine basis to detect changes in image quality that might not be detected by the Catphan measurements. Regarding image registration and couch shift accuracy, we have performed a simple QA using Rando phantom where the image analysis and position correction algorithm was found to be within the resolution of the patient couch motion (1 mm and 1 degree). The dose to the patient from a

28 CBCT scan and its relation to image quality is an interesting field of research. If the primary use for the images is to enable reliable patient alignment before treatment, image quality need be sufficient only for that task. Optimal settings of parameters like kilovolt, milliampere, field size and other parameters specific to each anatomic area will be found through clinical use and protocol development. The impact of target motion in CBCT imaging, use of CBCT scans for dose recalculation based on the patient s anatomy on the day of treatment and CBCT based adaptive radiotherapy were studied in detail in later chapters. 2.5 CONCLUSION The development of volumetric imaging in the treatment room and the opportunity for off-line and on-line guidance of radiotherapy offer strategies to increase the precision and accuracy of dose delivery. The above described procedures allow for safe clinical implementation of CBCT systems in a modern radiotherapy department.