Quantitative performance characterization of image quality and radiation dose for dental CBCT machine (CS9300) Elham Abouei

Transcription

1 Quantitative performance characterization of image quality and radiation dose for dental CBCT machine (CS9300) by Elham Abouei B.Sc., Sharif University of Technology, 2011 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in The Faculty of Graduate and Postdoctoral Studies (Physics) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) August 2014 Elham Abouei, 2014

2 Abstract Purpose: To characterize the performance of cone beam CT (CBCT) used in dentistry, investigating quantitatively the image quality and radiation dose during dental CBCT over different settings for partial rotation of the x-ray tube. Methods: Image quality and dose measurements were done on a variable field of view (FOV) dental CBCT (Carestream 9300). X-ray parameters for clinical settings were adjustable for 2-10 ma, kvp, and two optional voxel size values, with fixed time for each protocol and FOV. The phantoms were positioned in the FOV to imitate clinical positioning. Image quality was assessed by scanning a cylindrical poly-methyl methacrylate (PMMA) image quality phantom (SEDENTEXCT IQ), and the images were analyzed using ImageJ to calculate image quality parameters such as noise, uniformity, contrast to noise ratio (CNR), and spatial resolution. A protocol proposed by SEDENTEXCT, dose index 1 (DI1), was applied to dose measurements obtained using a thimble ionization chamber and cylindrical PMMA dose index phantom (SEDENTEXCT DI). Dose distributions were obtained using Gafchromic film. Results: The image noise was % which, when normalized to the difference of mean voxel value of PMMA and air, was comparable between different FOVs. Uniformity was % across the images. CNR was , , , , and for LDPE, POM, PTFE, air, and aluminum, respectively. The measured FWHM and spatial resolution were larger than the voxel size. FWHM were mm; spatial resolution was Dose distributions were symmetric about the rotation angle s bisector. For large and medium FOVs at 4 ma, kvp, and μm, DI1 values were in the range of mgy. DI1 values were between mgy for small FOV (5x5 cm 2 ) at 4-5 ma,75-84 kvp, and 200 μm. Conclusion: Noise and spatial resolution decreased and the CNR increased by increasing kvp; the geometric distortion, AAV, FWHM were very similar or the same when increasing the kvp. When FOV size increased, image noise increased and CNR decreased. FWHM and spatial resolution have no correlation with the voxel size. DI1 values were increased by increasing tube current (ma), tube voltage (kvp), and/or FOV. ii

3 Preface This research is conducted by the author under supervision of Dr. Nancy L. Ford. This thesis is written by the author and she has benefitted from the comments made by Dr. Ford. The images of the SEDENTEXCT IQ dental phantom for image quality purpose were obtained and analyzed by the author. Sierra Lee helped the author in image acquiring and analysis for reproducibility measurement. Pierre Deman helped the author in MATLAB coding for PSF measurement for image quality parameter. The dosimetry measurements were done by the author. iii

4 Table of Contents Abstract... ii Preface... iii Table of Contents... iv List of Tables... vi List of Figures... viii Acknowledgments... xi 1. INTRODUCTION BACKGROUND CBCT in dentistry: advantages and limitations Cone beam CT image production Acquisition configuration Image detection Image reconstruction Image display Image quality parameters Noise and uniformity Artifact Contrast resolution Spatial resolution Radiation risk from CBCT Radiation Dosimetry Absorbed Dose Equivalent Dose Effective Dose Dose measurement Dental CBCT research SEDENTEXCT IQ dental CBCT phantom SEDENTEXCT DI dental CBCT phantom MATERIALS iv

5 3.1. CS 9300 machine SEDENTEXCT IQ dental CBCT phantom Noise and uniformity Geometric distortion Beam hardening artefacts Contrast resolution Pixel intensity Spatial resolution Blank PMMA insert SEDENTEXCT DI dental CBCT phantom Dosimetery equipment ImageJ software METHOD Image quality measurement Positioning and imaging protocol Reproducibility Image quality parameters and image analysis in ImageJ Dose measurement RESULT Image quality performance Image quality parameters Reproducibility of image quality parameters Dose measurement DISCUSSION CONCLUSION AND FUTURE WORK Conclusion Future work References Appendix v

6 List of Tables Table 2.1. Major benefits and limitation of CBCT imaging in OMF region...4 Table 2.2. ICRP recommended tissue weighting factors Table 3.1. (a) Different protocols on CS 9300 CBCT machine suggested by the manufacturer...20 Table 3.1.(b) Different protocols on CS 9300 CBCT machine suggested by the manufacturer. There are two low and high resolution modes; each cell under the resolution mode columns includes time (s) and voxel size (µm) respectively Table 5.1. LP/mm along XY plane and Z axis observed form the LP/mm insert are reported for different protocols on CS 9300 CBCT..45 Table 5.2. Reproducibility of measurements of three different images obtained with the same machine settings; values are the standard deviation of three measurements. Intraobserver variability of repeated measurements of the same image performed by a single observer; values are the standard deviation of three measurements obtained 14 days apart Table 5.3. Interobserver variability obtained from measured values of the same image performed by two observers. The numbers in the table above represent percentage (%) Table A-1. Noise, uniformity, geometric distortion, AAV, FWHM, spatial resolution, and LP/mm obtained from the PSF insert are reported for different protocols on CS 9300 CBCT 61 Table A-2. Noise, uniformity, geometric distortion, AAV, FWHM, spatial resolution, and LP/mm obtained from the PSF insert are reported for different positioning of 5x5 cm2 FOV protocols on CS 9300 CBCT. 62 Table A-3. CNR for different material are reported for different protocols on CS 9300 CBCT...63 Table A-4. CNR for different material are reported for different positioning of 5x5 cm2 FOV protocols on CS 9300 CBCT.64 Table A-5. High and low contrast, LP/mm along XY plane and Z axis observed form the LP/mm insert, and LP/mm obtained from the PSF insert are reported for different protocols on CS 9300 CBCT.65 Table A-6. High and low contrast, LP/mm along XY plane and Z axis observed form the LP/mm insert, and LP/mm obtained from the PSF insert are reported for different positioning of 5x5 cm2 protocols on CS 9300 CBCT vi

7 Table A-7. Measured DI1 and reported DAP by CS 9300 for different protocols on CS 9300 CBCT vii

8 List of Figures Fig X-ray beam projection scheme comparing fan-beam CT (a) and cone-beam CT (b) geometry.5 Fig Novel method of extending FOV using a flat panel detector. (a) Conventional geometric arrangement whereby the central ray of the x-ray beam from the focal source is directed through the middle of the object to the center of the flat panel detector. (b) Alternate method of shifting the location of the flat panel imager and collimating..6 Fig (a) Diagram of base part, seven columns and caps, and a lid on the top of the phantom. (b) blank PMMA phantom with four rails to fit a column, (c) white horizontal and vertical lines engraved on the outer surface of IQ phantom the x-ray beam laterally to extend the FOV object...21 Fig (a) The artifact insert, (b) the axial view of two inserts the IQ phantom in ImageJ...22 Fig Contrast resolution inserts in row three Fig (a) The five disks inside pixel intensity insert, (b) sagittal view of the insert in ImageJ (from bottom to top: Air, LDPE, POM, PTFE, and Al)...23 Fig.3.5. (a) LSF insert, (b) sagittal view of the insert in ImageJ Fig (a) PSF insert, (b) top view in ImageJ Fig LP/mm insert in z-axis, (b) sagittal view of the insert in ImageJ Fig DI SEDENTEXCT phantom Fig (a) ionization chamber plate, (b) filler and adapter Fig (a) noise/uniformity ROI for 17x11,17x6, and 5x5 cm 2 FOV, (b) noise/uniformity ROIs for 8x8, 10x10, and 10x Fig (a) two perpendicular lines to measure line profiles for geometric distortion, (b) Plot the line profiles in MATLAB and calculate the periodicity Fig (a) Circular ROIs on each column for artifact analysis (b) rectangular ROIs around the artifact rods Fig a 20 mm diameter ROI at the Al disk for CNR calculation viii

9 Fig (a) The ROI on the central axial slice of the PSF insert image. (b) Surface plot of the PSF insert over the ROI. (c) an integrated 1D profile of the PSF, (d) MTF obtained from the FFT of the 1D PSF Fig The dose measurement points for DI1 using DI phantom. Number 1 is located at interior/right side of the phantom Fig The columns used to measure the point dose Fig Noise values for all protocols on CS9300 machine Fig 5.2. The uniformity graphs including the noise in the image, which is not representing the actual noise in the image due since it was smoothed. (a) Uniformity profile at center of image representing symmetrical uniformity sinking at middle as an example for 17x11, 17x6 (b) Uniformity profile at center of image is showing unsymmetrical sinking uniformity at middle of image for 5x5, 8x8, 10x10, and 10x5 FOVs Fig (a) Noise values were similar for different positioning for each setting. Also considering the error of 1.14, the noise values agreed within each setting. (b) There were more variations in uniformity over the image in different position of the 5x5 FOV which was the same by considering the error of Fig The mean gray value (MGV) and SD value at seven insert in artifact induction layer.(a) MGV and SD for 8x8, 10x10, 10x5, and 5x5 FOVs. (b) MGV and SD for 17x11 and 17x6 FOV Fig The CNR values vs. material densities are plotted for each FOV. Each graphs shows four curves, which correspond to child (blue), small adult (red), average adult (green), and large adult (purple) Fig CNR values for different material densities are plotted for settings of 5x5 FOV; Each graph shows CNR values of the same settings at five different positioning Fig The average of FWHM for each FOV is plotted vs. its voxel size Fig (a) The MTF graph for child setting for six FOVs, (b) The zoomed graph at 10% of the MTF...47 Fig The FWHM and resolution for different positioning of the 5x Fig Axial dose distribution for different FOV. The dose distribution was obtained just for average adult and child setting. Since all scanning parameters for 10x5 and 10x10 FOVs are the ix

10 same, the dose distribution for both of them are the same, and dose distribution for 10x10 is not represented here. The red line represented the measuring diameter for DI1. The bottom and left side the dose distribution correspond to anterior and right side of the patient. The scale bar shows the distribution of dose as percentage of maximum dose in each image Fig The average of dose reading at each point along the measuring diameter is plotted. Points 1 to 5 were placed anterior to posterior uniformly Fig DI1 for different FOVs and different settings related to child, small, average, and large adult x

11 Acknowledgments First and foremost, I would like to express my appreciation to my supervisor Dr. Nancy L. Ford for her guidance, ongoing support and never-ending patience throughout the course of my graduate studies. I consider myself fortunate and honored to have had the opportunity to work under her supervision. I would thank Sierra Lee to help me in data acquiring and analysis in some part of this research. My appreciation is to Nadine Bunting and Dr. Andrea Esteves for the permission to use CS9300 machine in UBC dental clinic. To all my wonderful friends, I would like to express my gratitude; In particular, I am very grateful to Pierre Deman for all his help in my research. Last but not the least, I owe special thanks to my family for their endless love and support in my life. I am indebted to my parents for their determination, encouragements, perpetual love and being there for me; they are my best mentor in every steps of my life. I also owe special thanks to my sisters and brother who been a constant source of inspiration and love. I am grateful for their positive energy, calm and kind advice. It is to them that this thesis is dedicated. xi

12 1. INTRODUCTION X-ray computed tomography (CT) has developed over the past decades as an important medical imaging modality, which is today a well-established tool with widespread application in diagnosis, surgical guidance, and monitoring. New applications of CT are still coming into practice in areas that were hardly expected. One example of such novel use of CT was the introduction of cone beam computed tomography (CBCT) designed for imaging the maxillofacial region. Although CBCT has been used for some time in medicine, for example in cardiac imaging and image guided radiotherapy, the first dental CBCT machine was described in The first dental CBCT machine came into the United States market in 2000 after approval by the US Food and Drug Administration (FDA).[1],[2],[3] Dental radiography provides essential information to dentists during diagnosis and subsequent treatment planning in patients for whom a thorough patient history and examination has been performed. The introduction of CBCT specifically dedicated to imaging the maxillofacial region indicates a shift from two dimensional (2D) to three dimensional (3D) data acquisition and image reconstruction. This resulted in broad acceptance of CBCT in dentistry in the last decade. The unprecedented interest in CBCT from all fields of dentistry is because it has created a revolution in maxillofacial imaging, facilitating the transition of dental diagnosis from 2D to 3D images, and expanding the role of imaging from diagnosis to image guidance for operative and surgical procedures using a third-party applications software. The proliferation of new applications of CBCT (each involving new system geometries, scan orbits, radiation dose protocols, and image quality characteristics) needs quantitative technical assessment by scientific methodology and imaging physics to carefully characterize the performance of such systems, ensure they are appropriately installed, understand their performance capabilities with respect to specific imaging task, and make knowledgeable selection of technique protocols. On the other hand, radiology involves exposure of patients as well as clinical staff to x-rays. Dentists use radiology to a greater extent on children and young adults, so the need for careful use is important. So the use of radiation by dentists provides a responsibility to ensure appropriate protection.[1],[4],[2] The purpose of this study is to investigate the image quality and radiation dose associated with the Carestream 9300 (CS 9300) CBCT scanner using different scanning protocols. An 1

13 understanding of the effect of field of view (FOV), tube current, and voltage peak on image quality and radiation dose will aid in determining the optimum protocol for each patient and clinical question. This study will report technical assessment of the image quality and radiation dose associated with manufacturer-specified protocols to characterize the CS 9300 used in the UBC dental clinic. 2

14 2. BACKGROUND 2.1. CBCT in dentistry: advantages and limitations CBCT is valuable for imaging of the maxillofacial region in dentistry since in many important aspects the images are superior to the conventional 2D projection images. CBCT gives clear images of highly contrasted structures; it is extremely appropriate for evaluating bone and teeth. Although limitations currently exist in the use of this technology for soft tissue imaging, efforts are being directed toward the development of techniques and software algorithms to improve signal to noise ratio and increase contrast. In addition, the ability to make cross-sectional slices at any arbitrary angle greatly facilitates the identification and localization of disease. Accurate measurements, a particularly important feature for placing implants and orthodontic treatment planning can be performed on CBCT images. Currently, CBCT is used most commonly in the assessment of bony and dental pathologic conditions including fracture, structural maxillofacial deformity and fracture recognition, preoperative assessment of impacted teeth, temporomandibular joint imaging, and in the analysis of available bone for implant placement. In orthodontics, CBCT imaging is now being directed toward 3D cephalometry. Also, application of CBCT for endodontic purposes appears to be the most promising use of CBCT, in many instances instead of 2D images.[4],[5],[6],[7] The major advantages of CBCT imaging are its accessibility, easy handling, and 3D imaging based on a single scan with a low radiation dose. Patel et al. stated that perhaps the most clinical benefit of CBCT imaging is that the highly sophisticated software allows breaking down the huge volume of collected data and reconstructing them into a format that closely resembles that produced by other imaging modalities such as multi-slice CT (MSCT). In contrast to the conventional CT scanners which are expensive and large to purchase and maintain, CBCT is suited for use in clinical dental practice where cost and space consideration are important; also, scanning requirements are limited to the head and dose consideration are important. The advantages of CBCT in the oral and maxillofacial (OMF) region are summarized in table 2.1.[8],[9],[6] Although there are some important limitations and concerns, CBCT imaging will be the imaging modality of choice in OMF region. The most important disadvantage of CBCT imaging is the 3

15 low contrast resolution and limited soft tissue contrast. Another limitation is that CBCT cannot be used for the estimation of Hounsfield Unit (HU) because the voxel value of an organ depends on the position in the image volume. The limitations of CBCT in OMF region are summarized in table 2.1. [10],[9] Table 2.1. Major benefits and limitation of CBCT imaging in OMF region [9] Benefits - 3D dataset - real-size data - potential for generating all 2D images (e.g. orthopantomogram, lateral cephalogram, TMJ) - potential for vertical scanning in a natural seated position - isotropic voxel size - high resolution (e.g. bone trabeculae, Periodontal ligament (PDL), root formation) - lower radiation dose than MSCT - less disturbance from metal artifacts - reduced costs compared with MSCT - easy accessibility - in-office imaging - easy handling - small footprint - Digital Imaging and Communications in Medicine (DICOM) compatible - user-friendly post-processing and viewing software - energy saving compared with MSCT Limitations - low contrast resolution (dependent on the type of X-ray detector) - limited detector size causes limited field of view and limited scanned volume - limited inner soft tissue information - increased noise from scatter radiation and concomitant loss of contrast resolution - movement artifacts affecting the whole dataset - truncation artifacts (caused by the fact that projections acquired with region of interest selection do not contain the entire object) - cannot be used for estimation of Hounsfield units (HU) 4

16 2.2. Cone beam CT image production CBCT units are distinguished from CT units in part by their imaging geometry. The x-ray beam in a cone beam unit diverges as a cone to the patient rather than being collimated into a fan beam like in a CT unit (Fig. 2.1). Then, the region of interest (ROI) can be imaged with one rotation of the x-ray source, or less, around the patient s head. The four components of CBCT image production are (1) acquisition configuration, (2) image detection, (3) image reconstruction, and (4) image display which are discussed below:[4],[5] a. b. Fig X-ray beam projection scheme comparing fanbeam CT (a) and cone-beam CT (b) geometry.[11] Acquisition configuration The geometric and acquisition configuration for the cone beam technique are theoretically simple. An x-ray source does a single partial or full rotational scan around a fixed fulcrum within the patient s head while a reciprocating 2D area detector moves simultaneously with the x-ray source. During the scan rotation, each projection image is made by sequential, single-image capture of attenuated x-ray beams by the detector. Technically, exposing the patient using a continuous beam of radiation during the rotation is the easiest method, and allows the detector to sample the attenuated beam in its trajectory. However, continuous radiation emission does not totally contribute to the formation of the image and results in greater radiation dose to the patient. There is also another method that the x-ray beam be pulsed to coincide with the detector sampling. This method contributes to reduced patient radiation dose due to the actual exposure time being significantlyless than scanning time.[4] 5

17 The dimensions of the FOV or scan volume covered for imaging depend mainly on the detector size and shape, the beam projection geometry, and the beam collimation; the shape of the scan volume can be either cylindrical or spherical. Collimation of the primary x-ray beam limits exposure to the ROI. Field size limitation ensures that an optimal FOV can be selected for each patient, based on disease existence and the region designated to be imaged. CBCT systems can be classified according to the FOV s height as follows:[4] Localized region: about 5 cm or less (e.g. dentoalveolar, temporomandibular joint) Single arch: 5 cm to 7 cm (e.g. maxilla or mandible) Inter arch: 7 cm to 10 cm (e.g. mandible and superiorly to include the inferior concha) Maxillofacial: 10 cm to 15 cm (e.g. mandible and extending to Nasion) Craniofacial: greater than 15 cm (the lower border of the mandible to the head s vertex) Extended scanning field size related to the craniofacial region is difficult to incorporate into cone beam design due to the high cost of large area detectors. One unit (e.g. icat Extended Field of View model) has large FOV by the software addition of two rotational scans to produce a single volume with a 22 cm height. Another novel method for increasing the width of the FOV using a smaller area detector (to reduce manufacturing costs) is to offset the position of the detector, collimate the beam asymmetrically, and scan only half the patient (Fig. 2.2). a. b. Fig Novel method of extending FOV using a flat panel detector. (a) Conventional geometric arrangement whereby the central ray of the x-ray beam from the focal source is directed through the middle of the object to the center of the flat panel detector. (b) Alternate method of shifting the location of the flat panel imager and collimating the x-ray beam laterally to extend the FOV object. [4] 6

18 During the scan, single exposures are made at defined degree intervals, providing individual 2D projection images, known as basis, frame, or raw images. These images are similar to lateral and posterior-anterior radiographic images, each slightly offset from one another. The complete series of images is referred to as the projection data. The number of images comprising the projection data throughout the scan is determined by the frame rate (number of images acquired per second), the speed of the rotation, and the completeness of the trajectory arc. The number of projection scans comprising a single scan may be fixed or variable. More projection data provide more information to reconstruct the image, allowing for greater spatial and contrast resolution, increasing the signal to noise ratio, producing smoother images, and reducing metallic artifacts. However, more projection data usually necessitates a longer scan time, a higher patient dose, and longer primary reconstruction time. In accordance with the as low as reasonably achievable (ALARA) principle, the number of basis images should be minimized to produce an image of diagnostic quality.[4] Image detection After passing the beam through the patient, the remnant beam is captured on a 2D area detector, usually an amorphous silicon or complementary metal-oxide semiconductor (CMOS) flat panel or image intensifier tube/charge coupled device (IIT/CCD) detector. CBCT units can be divided into two groups, based on detector type as an IIT/CCD combination or a flat panel imager. Some have found flat panel detectors offer higher spatial resolution and higher signal to noise ratios than image intensifiers in cone beam machines.[4] The resolution and detail of CBCT images are determined by the individual voxels produced from the volumetric data set. In CBCT imaging, voxel dimensions primarily depend on the pixel size on the area detector; the voxels are isotropic, equal in all three dimensions. The resolution of the area detector is generally submillimeter in CBCT units. [4] Image reconstruction Once the raw projection frames have been acquired, they must be processed to create the volumetric data set. This process is image reconstruction. The number of raw projection frames may be from 100 to more than 600, each with more than one million pixels, with 12 to 16 bits of 7

19 data assigned to each pixel. Then, data reconstruction is computationally a complex process. In CBCT data reconstruction is performed by personal computer unlike conventional CT which data are usually acquired by one computer and transferred by way of an Ethernet connection to a processing computer to facilitate data handling. Reconstruction times vary, depending on the acquisition parameters (voxel size, FOV, number of projections), software (reconstruction algorithms), and hardware (processing speed) used; it should be accomplished in an appropriate time (less than 3 minutes for standard resolution scans) to complement patient flow. The reconstruction process includes two stages, acquisition stage and reconstruction stage. Acquisition stage is the first stage in reconstruction process. Due to the spatial variation of physical properties of the photodiodes and the switching elements in the flat panel, and also because of variations in the x-ray sensitivity of the scintillator layer, raw images from CBCT detectors show spatial variations of dark image offset and pixel gain. Even high-quality detectors exhibit inherent pixel imperfections or a certain amount of defect pixels in addition to offset and gain variations. As a result, raw images require systematic offset and gain calibration and a correction of defect pixels; the sequence of the required calibration steps is called detector preprocessing and the calibration requires the acquisition of additional image sequences including dark field and bright field images. The next stage is Reconstruction stage. Image reconstruction is a mathematical operation that converts attenuation values per voxel from x-ray projection data to gray-scale images. After images are corrected, they need to be related to each other and assembled. One method involves constructing a sinogram, and then processing the corrected sinograms. To reconstruct the image from the sinogram, the most common reconstruction algorithm is Filtered Back Projection (FBP) that is applied to the sinogram and converts it into a 2D CT slice. FBP consists of two steps, filtering of projection data followed by backprojection. The filtered projection data are then backprojected along the original rays. Once all the slices have been reconstructed, they can be recombined into a single volume for visualization Image display The CBCT technology provides the dental clinician with a great choice of image display formats. The volumetric data set, for most CBCT devices, is presented to the clinician on screen as secondary reconstructed images in three orthogonal planes (axial, sagittal, and coronal); the slice 8

20 thickness is usually at a thickness defaulted to the native resolution. Optimum visualization of orthogonal reconstructed images depends on the adjustment of window level and window width to favor bone and the application of specific filters.[4] 2.3. Image quality parameters Image quality refers to the capability of a radiograph modality to reproduce structures and tissues accurately. The quality of the final radiographic image is dependent on many different variables in the radiographic technique. The goal of optimization of all aspects of the radiographic technique is to create a final image which is useful in diagnosis and treatment plan procedure. In order to quantify image quality, image quality phantoms are used. They include a set of test objects to objectively assess image quality. The parameters that are used to quantify image quality are discussed below Noise and uniformity Noise introduces a random or stochastic component into the image; noise adds or subtracts to a measurement value, then the recorded measurement differs from the actual value. Noise represents itself as inconsistent gray values in the projection images, which means large standard deviations in areas where a constant attenuation should be present. There are several sources of noise in an image. One is additive noise stemming from round-off errors or electrical noise; it may originate from the scintillator structures or detector electronics. Another source is quantum noise that should be expected to follow a Poisson distribution; the fluctuation in x-ray photon detection is an intrinsic property to the x-ray imaging and cannot be prevented with current technology. Also, scattered radiation seems to affect the image quality; the additional share of scattered X-rays results in increased measured intensities that cause incorrect gray values and introduce non-uniformities in the reconstructed image. The scattered radiation is affected by kvp, FOV size, and patient thickness. For dose reduction reasons, dental CBCT machines are operated at ma settings approximately one order of magnitude below those of medical CT machines. Thus, a higher noise level is to be expected in CBCT images than in CT. Moreover, the image-degrading effect of scattered radiation will affect CBCT machines more than classical 9

21 CT due to the geometry of area detectors; the larger the detector, the higher probability that scattered photons detected. [12],[13],[14] Image accuracy CBCT imaging produces images with sub-millimeter isotropic voxel resolution ranging from 0.4 mm to as low as mm. Due to the isotropic voxels constituting the volumetric dataset, image data can be sectioned non-orthogonally. Most software provides multiplanar reformation (MPR) including oblique, curved planar reformation to provide simulated distortion free panoramic images and serial transplanar reformation to provide cross-sections that can be used to highlight specific anatomic regions. Zoom magnification, window/level adjustments, and text or arrow annotation can be applied on the image. Clinician benefits from the cursor-driven measurement algorithms with an interactive capability for real-time dimensional assessment. On screen measurements are free from distortion and magnification.[4],[15] Artifact An image artifact may be defined as a visualized structure in the image that is not present in the object under investigation. They are induced by discrepancies between the mathematical modelling and the actual physical imaging process. Artifacts represent themselves very often by streaks, line structures and shadows oriented along the projection lines. Since artifacts may interfere with the diagnostic process performed on CBCT data sets, every user should be aware of their presence. In the scientific literature, the following artifacts are reported: [13] Extinction artifact is often termed missing value artifacts. If the scanned object contains highly absorbing materials such as prosthetic gold restorations, then the recorded signal in the detector pixels behind that material may be close to zero or actually zero. So, no absorption can be computed and severe artifacts are induced as these zero entries are backprojected into the volume.[13] Beam hardening artifact is one of the most prominent sources of artifacts. The lower energy rays of the polychromatic x-ray spectrum may be substantially absorbed when passing through the scanned object. So, the beam is hardened, and a non-linear error induced in the 3D reconstruction 10

22 data which is backprojected into the volume, resulting in dark streaks. It is shown that even light metal (e.g. titanium) causes massive beam hardening for the typical kvp in a CBCT unit.[13] Exponential edge gradient effect appears at sharp edges with high contrast to neighbouring structures. It is caused by averaging the measured intensity over a finite beam width (and finite focal spot width), while the reconstruction algorithm assumes zero width. The width is determined by the focal spot and detector pixel size in combination with the imaging geometry of the machine. The effect will always reduce the computed density value, and represent as streaks tangent to long straight edges in the projection direction. As sharp edges of high contrast may commonly occur in the oral cavity, e.g. at metallic crown borders, this artifact also has to be considered in dental CBCT.[13] Aliasing artifact is caused when the Nyquist sampling theorem is violated. It requires that the sampling frequency must be higher than twice the highest frequency in the signal. The size of the detector elements owing to under-sampling and the divergence of the cone beam in CBCT are typical factors causing aliasing.[13] Ring artifacts are visible as concentric rings centred around the location of the axis of rotation. They are most prominent when homogeneous media are imaged. Apparently they are caused by defect or uncalibrated detector elements. [13] Motion artifacts and misalignment artifacts are two sources of error closely related in that a misalignment of any of the three components (source, object and detector) causes inconsistencies in the backprojection process. [13] Contrast resolution Contrast resolution is the ability of an imaging system to distinguish between objects of similar subject contrast; it refers to the difference in gray value in the image. Contrast resolution decreases with higher kvp as higher energy photons are more likely to pass through tissues than be attenuated by them. Therefore, more x-rays reach the detector creating a final image that has smaller intensity differences between similar tissues. Scattered radiation also increases when kv increases resulting in a noisier image. Contrast resolution is evaluated by two different methods: 11

23 a) High and Low Contrast Resolution is used to assess the minimum size of various density objects that may be visualized when they are embedded in a PMMA matrix. The more similar the density of the object is to the background matrix, the more difficult it will be to visualize at smaller dimensions. b) Pixel Intensity Values (PIV), also known as CT numbers or Hounsfield Units in CT imaging, are used to assess the known density of materials. PIVs depend on the kvp of the CBCT system. The output measurement of contrast resolution is referred to as contrast to noise ratio (CNR) in digital imaging which is the ratio of difference between PIVs of two materials to the average noise in these materials Spatial resolution Spatial resolution describes the capability of an imaging system to display high contrast detail separately. It can be measured visually using a bar phantom in which small line pairs are positioned closer and closer together; Line pairs/mm, involves visual analysis of the number of line pairs made up of one black and one white line to create a pair. The smallest distance between the black and white lines that may be clearly distinguished defines the spatial resolution of the imaging system. Another way to evaluate the spatial resolution of an imaging system is by measuring the modulation transfer function (MTF) of the system. The MTF can be calculated by measuring the Point Spread Function (PSF) or the Line Spread Function (LSF). Then, the MTF can be calculated as the modulus of the Fourier transform of the PSF or the LSF.[2] The values reported are the frequencies at which the modulation falls to 50% or 10% of its initial value since the eye is relatively insensitive to detail at spatial frequencies where MTF is lower than 10%. The PSF[16] can be measured directly by imaging a wire test object. The wire is embedded in a suitable medium and placed perpendicular to the scan plane. The PSF is obtained by surface plotting the pixel values across the image of the wire. Orthogonal slices through the 2D MTF are calculated by computing the Radon transform of the PSF (integrating along either the x or y axis), and then calculating the 1D Fourier transform, and dividing by the spatial frequency spectrum described by the wire. Also, resolution can be measured directly from the PSF by measuring the full width at half maximum (FWHM).[14] 12

24 The LSF can be used to assess the spatial resolution of the system similar to the PSF. The LSF can be obtained by differentiating the Edge Spread Function (ESF) which is measured by imaging an edge of a block of material embedded in a suitable material with the face of the block perpendicular to the scanned plane. The ESF is obtained by plotting the pixel values across the image.[2] 2.4. Radiation risk from CBCT In humans, the biological effects of radiation are categorized as stochastic or nonstochastic effects. Nonstochastic effects are also called acute or deterministic effects, and are characterized by a threshold below which effects are not observed; when the threshold is exceeded, the effects are seen and the magnitude increases with increased dose. These effects have a clear association with radiation exposure. Examples include erythema, loss of hair, cataracts, nausea, vomiting and depression of bone marrow cell division. Stochastic effects are also referred to as late effects since they usually occur years after exposure. These effects are probabilistic in nature, and may or may not present in a given individual. The probability of stochastic effects increases with radiation dose but not necessarily the magnitude of the effect. Unlike the deterministic effects, a threshold may not exist and no clear association between exposure and effect has been seen. Examples of stochastic effects include cancer and hereditary effects on the offspring of exposed individuals.[17] In oral and maxillofacial diagnostic imaging the main concern is the risk for stochastic effects, namely radiation induced cancer since the doses given are all well below the thresholds for deterministic effects. Prescription of dental and medical imaging must be done with diligence, and therefore recognizing the risk associated with radiation exposure is necessary; dosimetry studies estimates the biological risk associated with radiation in imaging.[17] 2.5. Radiation Dosimetry Interaction of radiation with matter results in a deposit of energy in tissue. Dosimetry is the determination of dose which is described in terms of the energy absorbed per unit mass at a site of interest. In the clinic, dosimetry provides an estimation of the biological effects of radiation 13

25 from which appropriate therapeutic and diagnostic uses can be determined. There are different methods to measure the quantitative effects of ionizing radiation with matter.[17] Absorbed Dose The absorbed dose describes the energy absorbed from any type of ionizing radiation per unit mass of any type of matter. The SI unit for absorbed dose is Gray (Gy) replacing the traditional unit of Rad (radiation absorbed dose) which is equal to 1cGy. Then the absorbed dose is determined by the following equation:[17] Equivalent Dose Absorbed dose measures the physical energy absorbed, however it does not account for effect of different types of radiation which have a different potential in producing biological damage. High linear energy transfer (LET) radiations (e.g. high-energy protons) are more damaging to tissue than lower LET radiations (e.g. x-rays). The radiation weighting factor ( ) accounts for this relative biologic effect in human tissue and results in calculating the equivalent dose ( ). Equivalent dose has the same unit as absorbed dose, but it is described in the special unit of Sieverts (Sv). For the x-rays used in dental CBCT, the radiation weighting factor is 1.0; hence the values of Dose (in Gray) matches the values of Equivalent Dose (in Sv). Finally the equivalent dose is formulated as following:[17] Effective Dose Effective dose (E) estimates the biologic risk in humans exposed to radiation. It allows for comparison of risk between partial exposures by representing such exposure as a full-body dose of equivalent detriment. The effective dose is also described in units of Sv and obtained through the sum of the products of average equivalent dose to each tissue and the associated tissue weighting factor ( ) as shown in table 2.2. The effective dose is summarized as follows:[17] 14

26 Table 2.2. ICRP recommended tissue weighting factors 2007.[18] Tissue Tissue weighting factor ( ) Bone-marrow (red), Colon, Lung, Stomach, Breast, Remainder tissues* Gonads Bladder, Oesophagus, Liver, Thyroid Bone surface, Brain, Salivary glands, Skin Total 1.00 * Remainder tissues: Adrenals, Extrathoracic (ET) region, Gall bladder, Heart, Kidneys, Lymphatic nodes, Muscle, Oral mucosa, Pancreas, Prostate (male), Small intestine, Spleen, Thymus, Uterus/cervix (female) Dose measurement The quality assurance (QA) programme in radiography aims to ensure doses are kept as low as reasonably achievable (ALARA). Clinicians who make the decision regarding the justification of the exposure need to know about the patient dose during radiography, and it is important to ensure doses are optimised and within any national and international guidelines. In this case, patient doses are needed to be monitored on a regular basis and compared to agreed standards. The international commission on radiological protection (ICRP) has recommended use of diagnostic reference levels (DRLs) as standard dose levels in patients undergoing medical imaging and nuclear medicine procedures.[2],[19] DRLs refer to an easily measurable dose quantity. The effective dose indicates radiation risk and can be compared to doses from other radiation sources; it is usually considered as a best overall indicator of patient dose to be compared between systems. On the other hand, effective dose cannot be measured easily, and must be derived from more easily measurable dose quantities. So, other dose quantities are evaluated for the purpose of QA. Ideally, the dose quantity used should give a good correlation to the effective dose and hence overall patient risk; dose quantities used for the regular assessment of patient dose must be relatively easy to measure in a clinic. Entrance surface dose (ESD) and dose area product (DAP) are quantities that are routinely used in conventional radiology. In the field of CT, the computed tomography dose index (CTDI) 15

27 and dose length product (DLP) are routinely used. CTDI is a measurement of the dose integrated across the dose profile along the patient s length. Such a dose index cannot be used in dental CBCT units because of the greater beam size and asymmetry of the dose distribution. The Health Protection Agency, in the UK, has proposed the use of DAP. This dose quantity is promising since it provides one reading per exposure and gives an indication of both the dose level in the beam and the area irradiated. Some CBCT devices provide this information after each exposure, automatically. If this became universal like CT scanners that all display DLP, it would greatly facilitate patient dose audit. The accuracy of these readings should be checked by the medical physics expert during routine testing.[2],[20],[21] AAPM TG-111 introduces the concepts of integral dose ( ) and planar average equilibrium dose ( ) which are related to each other by this equation ; where R is the radius of the volume, L is the scan length and is the mass density of the phantom. The integral dose, serves as a simplified indicator of patient risk: the presumption is that cancer risk increases with the larger the dose and irradiation volume containing radiosensitive tissue. [22] Recognising that when D(r) has a parabolic form, then measurement of Deq at two points such as the centre and periphery of the phantom as conventionally measured for CTDI, would be reasonable and practical. Recognising that D(r) does not always have a parabolic form, it would require more detailed measurement or the use of Monte Carlo modelling.[22],[23] The SEDENTEXCT team have proposed two other alternatives for dose quantities based on point measurement within PMMA phantoms for dental CBCT units. However, further work is required to establish whether these indices are appropriate for the setting of DRLs. Two CBCT dose indices are currently suggested. Dose index 1 requires point measurements along a diameter of the phantom and is calculated as the mean of the readings. Dose index 2 involves measurements at the centre of the phantom and at points around the periphery. The dose measurements are points across the X-Y plane in the centre of the Z axis. Measurements can be performed using a thimble ionization chamber or TLDs, within a suitable PMMA phantom (diameter of 16 cm is recommended). Index 1 allows the measurement of dose quantity for onaxis and off-axis exposures, and full and partial dose distributions simply by rotating the phantom in such a way that the isocentre of the x-ray beam lies on the measuring diameter. Index 2 is only suitable for symmetrical dose distributions.[2] 16

28 2.7. Dental CBCT research Cone beam computed tomography (CBCT) has been widely accepted and used in dental practice for only about ten years. It provides 3D imaging of a patient maxillofacial region at doses significantly lower than medical CT scanners, but at higher radiation dose than conventional dental radiography.[24] The establishment of the clinical role and the optimisation routines of CBCT are behind its growth due to the rapid increasing use of CBCT in dental practice. Therefore, the foremost aim of Safety and Efficacy of a New and Emerging Dental X-ray Modality (SEDENTEXCT), a project of the seven partners of the European Atomic Energy Community including University of Manchester (United Kingdom), National and Kapodistrian University of Athens (Greece), Iuliu Hatieganu University of Medicine and Pharmacy in Cluj- Napoca (Romania), Leeds Test Objects Ltd. (United Kingdom), Katholieke Universiteit Leuven (Belgium), Malmo University (Sweden), and Vilnius University (Lithuania) was to acquire key information necessary for scientifically based clinical use of CBCT in dentistry.[2] SEDENTEXCT IQ dental CBCT phantom The development of phantom (test object) goes along with the formation of QA protocols. A quality control (QC) procedure, as part of the QA process, design necessary tests to ensure all parameters during the examination are in accordance with the standard operating protocol, which result in images with diagnostic value without exposing the patient to unnecessary dose. During these activities, the diagnostic image quality always needs to be assessed, implying that dose measurements are to be reported in terms of diagnostic needs. Moreover, technical image quality assessments need to be evaluated for their diagnostic relevance. This is particularly the case for dental imaging since it involves a large variety of diagnostic indications requiring different imaging approaches. Then, it requires suitable test tools and protocols which are usually performed using a test phantom containing various inserts. The phantom constructed of materials with known characteristics to simulate a standard patient is allowing repeated x-ray exposure and measurement of image quality parameters. [6],[24],[25] Preliminary tests on a dental CBCT unit revealed that using a phantom designed for QA on MSCT scanner results in images with worse contrast resolution than the MSCT scan; 17

29 commercial QC phantoms have been designed for conventional CT, but these are not applicable for dental CBCT due to the difference in performance for certain image quality aspects. CT phantoms use soft tissue-equivalent materials for analysis of gray value; however, dental CBCT units are optimized for imaging hard tissue, i.e. bone and teeth. This is also due to the low dose delivered in comparison to medical CT. In addition, dental imaging needs a high spatial resolution and a limitation of metal artifacts, both of which are not evaluated by conventional CT phantoms. They also show the need for a standardized QC phantom suited for use on all dental CBCT devices, and which provides results relevant to dental imaging and can be compared between systems. [25],[24] Finally, one of the main objectives of the SEDENTEXCT project was to develop a universal QC phantom particularly designed for dental CBCT, with size and densities resembling dental interest, with special software tools for the interpretation of the results and evaluation of image quality. The phantom can be applied on all CBCT devices currently on the market SEDENTEXCT DI dental CBCT phantom A major output from the CBCT dosimetry working party 2 (WP2) was the development of standardised dose indices for characterising dose distribution for CBCT devices. For the measurement of DI1 and DI2, a new customized cylindrical PMMA dosimetry phantom was created by partner Leeds Test Objects (LTO), which they are now marketing ("SEDENTEXCT DI dental", The proposed dose indices were validated on a range of CBCT machines for a large number of exposure protocols, by variation of all possible exposure factors (field of view size, kvp, mas, and rotation arc). [26] 18

30 3. MATERIALS In this study, the image quality and radiation dose of a dental CBCT machine, CS 9300, were evaluated using image quality (IQ) and dose index (DI) phantoms developed by Leeds Test Objects Ltd. Image quality assessment of CS9300 was obtained using images of the IQ phantom scanned with the machine and analyzed using ImageJ software. To do radiation dose evaluation, the DI phantom was scanned with the machine while using the XR-QA2 Gafchromic TM films to measure dose distribution and using a thimble ionization chamber to measure the exposure CS 9300 machine The CS 9300 (Carestream Health, Rochester, NY) installed in the UBC dentistry clinic provides dental CBCT imaging for different settings. Six different fields of view (FOV) are available on this machine. The tube voltage and tube current can be selected over a range of kv and 2-15 ma, respectively. Although voxel size is between μm, there are only two options for each FOV to be selected. The exposure time is fixed for each protocol and cannot be changed. Different protocols suggested by the manufacturer for dental CS 9300 are shown in table 3.1 (a) and (b). In this study the larger voxel size was selected for 5x5 cm 2 FOV, and smaller voxel size were selected for other FOVs to have the most consistent voxel sizes ( µm); other parameters were the same as table 3.1 There is one positioning for each FOV, except for the 5x5 and 8x8 (TMJ) FOVs. The 8x8 (TMJ) FOV can be positioned on right or left side symmetrically which are assumed to have the same results. The 5x5 FOV can be positioned centrally on front, right middle, right back, left middle, and left back teeth. The x-ray tube and detector rotate partially, greater than 180, around the patient to provide a number of projections sequentially. 19

31 Table 3.1.(a) Different protocols on CS 9300 CBCT machine suggested by the manufacturer. Patient type KV ma Time (s) Voxel size(µm) FOV 17x11 Child Adult- small Adult- average Adult- large FOV 17x6 FOV 8x8 FOV 10x10 FOV 10x5 Child Adult- small Adult- average Adult- large Child Adult- small Adult- average Adult- large Child Adult- small Adult- average Adult- large Child Adult- small Adult- average Adult- large Table 3.1.(b) Different protocols on CS 9300 CBCT machine suggested by the manufacturer. There are two low and high resolution modes; each cell under the resolution mode columns includes time (s) and voxel size (µm) respectively. FOV 5x5 Patient type KV ma Low resolution mode High resolution mode Child ,200 20, 90 Adult- small , , 90 Adult- average , , 90 Adult- large , , SEDENTEXCT IQ dental CBCT phantom The phantom was designed by Leeds Test Objects Ltd (Bouroughbridge, UK) for assessment of image quality for dental CBCT machines. The phantom is head-sized cylindrical poly-methyl methacrylate (PMMA) with a density of to simulate the human tissue. 20

32 PMMA is chosen since it has the closest density to water, the major component of human body. The phantom is 160 mm in diameter and 176 mm in height. The base part of the phantom (45 mm height) consists of uniform PMMA, and the upper part consists of seven cylindrical columns (35 mm diameter and 131 mm height) for accommodation of different test inserts which are 35 mm in diameter and 20 mm in height. One column is at the center of the phantom and six peripheral columns are at the vertices of a regular hexagon. Four rails, placed into each column, allow reproducible orientation of the test inserts as well as preventing the inserts from rotating inside the round columns. (Fig. 3.1) [27] There are six horizontal engraved lines each related to the center of the rows of the inserts and seven vertical engraved lines, six related to the six peripheral columns and one for recognizing the order of columns, on the outer surface of the phantom used for exact alignment and positioning of the phantom in a FOV. Seven caps with a letter from A-G are placed at the top of each column to ensure inserts remain secure in the columns and for recognizing columns. (Fig. 3.1) A lid is put on the top of the caps with a depression on it for spirit level placement to aid in phantom leveling. A hole is threaded at the bottom of the phantom to facilitate easy and secure positioning of the phantom on a tripod or the phantom may be placed on a flat surface. a. b. c. Fig (a) diagram of base part, seven columns and caps, and a lid on the top of the phantom[24]. (b) blank PMMA phantom with four rails to fit a column[27], (c) white horizontal and vertical lines engraved on the outer surface of IQ phantom[24] Noise and uniformity The noise and uniformity layer is the lower section of the phantom, found just above the screw thread. It is a uniform PMMA layer; the height of this layer is about 30 mm. It is used to measure the noise and uniformity of image. 21

33 Geometric distortion The geometric distortion layer is placed above the noise/uniformity layer. It contains an array of 2.0 mm diameter and 3.0 mm deep air gaps uniformly pitched at 10.0 mm intervals through one slice of the cylinder Beam hardening artefacts There are two inserts including a line of three 5.0 mm diameter rods of Titanium suspended in PMMA. These two inserts are positioned in the peripheral columns in row one so that one line of rods are perpendicular to the edge of the phantom and the other one parallel to the phantom edge and separated by two peripheral columns; other columns in the row are filled with blank PMMA inserts (Fig. 3.2). These inserts measure the effect of metal objects in the path of the x-ray beam. a. b Contrast resolution Fig (a) the artifact insert[27], (b) the axial view of two inserts the IQ phantom in ImageJ There are five contrast resolution inserts at peripheral columns in row three. Each insert contains 1.0, 2.0, 3.0, 4.0, and 5.0 mm diameter rods suspended in PMMA; each one includes different materials including Aluminium (Al), PTFE, POM, LDPE, and Air. (Fig. 3.3) The placement of these inserts is not crucial, as long as the central hole is not used; the central column is filled with blank PMMA insert. 22

34 Column B (Al) Column C (PTFE) Column E (POM) Column F (LDPE) Column G (Air) Contrast resolution insert [27] Fig Contrast resolution inserts in row three Pixel intensity An insert accommodating stack of five disks, each 25 mm in diameter and 2 mm in height, made of different materials (Al, PTFE, POM, LDPE, and air) is used for the pixel intensity measurement. This insert is placed in the remaining peripheral column in row three (column A). The disks are in the order of Air, LDPE, POM, PTFE, and Al from bottom to top of the insert. (Fig. 3.4) a. b Spatial resolution Fig (a) The five disks inside pixel intensity insert[24], (b) sagittal view of the insert in ImageJ (from bottom to top: Air, LDPE, POM, PTFE, and Al) There are four different inserts for measurement of spatial resolution. Line spread function (LSF) and point spread function (PSF) inserts can be used to calculate the modulation transfer function 23

35 (MTF) for quantitative spatial resolution measurement; line pair per mm (LP/mm) inserts are used for semi-quantitative measurement of spatial resolution. I. Line spread function The LSF insert is made of two PMMA and two PTFE quarters bonded together in an alternating pattern to create PMMA/PTFE interface. (Fig 3.5) It is placed in peripheral column A in row four. a. b. Fig.3.5. (a) LSF insert [27], (b) sagittal view of the insert in ImageJ II. Point spread function The PSF insert is a 0.25 mm diameter stainless steel wire suspended in air gap. It is placed in peripheral column B in row four. a. b. Fig (a) PSF insert [27], (b) top view in ImageJ III. Line pairs per mm There are two inserts for LP/mm with alternating discs of Aluminium and Polymer for the spatial resolution tests. The inserts are the same and the frequencies of the alternative disks (line pairs) 24

36 are 1.0, 1.7, 2.0, 2.5, 2.8, 4.0, and 5.0 line pairs per millimeter. The number of LP/mm is a measurement of capability of the CBCT machine to resolve line pairs in a millimeter of length. (Fig. 3.7) a. b. Fig (a) LP/mm insert in z-axis [27], (b) sagittal view of the insert in ImageJ These two inserts are oriented perpendicularly to each other with one insert in the peripheral column C and the other one in the central column D in row four. The LP/mm insert in column C is placed in xy-plane, and measures the spatial resolution in xy-plane; the other one in column D is placed along the z-axis and measures the spatial resolution in the z-direction Blank PMMA insert The inserts are homogenous PMMA, which are used to fill any remaining spaces in columns to make a solid phantom for imaging (Fig. 3.1.b) SEDENTEXCT DI dental CBCT phantom The SEDENTEXCT DI phantom comprises a stack of six PMMA plates (density 1.20 ± 0.01 g/cm 3 ) forming a cylinder in the size of an adult head (160 mm diameter and 176 mm height). A scale is engraved on outer surface of plates in 4 positions. The phantom is put on a PMMA base plate that has a spirit level for phantom leveling. (Fig. 3.8) There are three types of PMMA plates to provide capability of using different types of detector systems including ionization chamber, thermoluminescent detectors (TLDs), and gafchromic film. [28] 25

37 a. b. Fig 3.8. SEDENTEXCT DI phantom [29] There are four ionization chamber plates: two of them are 22 mm thick and the other two are 44 mm thick with five 26 mm diameter through holes for the insertion of ionization chambers. One TLD plate is a 22 mm thick PMMA plate with an array of 37 holes (4 mm squares, 1 mm corner radii) uniformly pitched for the insertion of TLDs. One film plate is a 22 mm thick PMMA plate with a 0.5 mm recess for the placement of Gafchromic film. There are 3 PMMA adapters with 26 mm diameter and heights of 22 mm, 44 mm and 66 mm to reduce the ionization chamber through hole s diameter from 26 mm to 13 mm. A range of blank PMMA fillers can be inserted into any of the unused ionization chamber through holes. (Fig. 3.9) a. b. Fig (a) ionization chamber plate, (b) filler and adapter. [29] 3.4. Dosimetery equipment A calibrated thimble ionization chamber system, a 0.6 cm 3 Farmer ionization chamber (10x6 0.6CT, Radcal Corp., Monrovia, CA) in conjunction with a Radcal electrometer (AccuDose, 26

38 Radcal Corp., Monrovia, CA), was used to measure the dose imparted from all protocols listed in table 3.1. XR-QA2 Gafchromic TM (ISP technologies INC., Wayne, NJ) films were used to visualize axial dose distributions ImageJ software ImageJ 1.46r (National Institute of Health, Bethesda, MD, USA) was used to measure the image quality parameters. Different parameters including area and pixel value statistics of user-defined region of interest (ROI) can be calculated in ImageJ. ImageJ can create density histogram and line profile plots. It supports standard image processing functions such as contrast manipulation, sharpening, smoothing, edge detection, and median filtering. The image can be zoomed up and down to 32 times. All analysis and processing functions are available at any magnification factor. Any number of windows (images) can be displayed simultaneously, limited only by available memory. 27

39 4. METHOD 4.1. Image quality measurement Image quality assessment of CS 9300 was performed for all protocols listed in table 3.1 by scanning the SEDENTEXCT IQ phantom, and analyzing images in ImageJ and MATLAB. Scout scans were performed before each imaging protocol to confirm the IQ phantom was positioned within a FOV correctly. The scanning protocol of the phantom in the FOV was varied for different sized FOVs Positioning and imaging protocol The IQ Phantom was mounted on the tripod, levelled using the spirit level on the lid and correctly positioned in all three planes of space using the white vertical and horizontal lines. The 17x11cm 2 FOV was large enough to capture the desired height of IQ phantom in one scan; the width of the FOV was also able to accommodate the entire diameter of the IQ phantom. So the IQ phantom was scanned once to take the image of all image quality inserts. The phantom was positioned so that the isocenter of the x-ray beam was at the center of the phantom and the height of the FOV covered the noise/uniformity layer up to a little above the row 4, and the phantom was scanned centrally in the FOV. For 17x6 cm 2 FOV, the width of FOV was wide enough to image the entire diameter of the phantom but two different vertical positions were required to image the related phantom height. The phantom was positioned so that the isocenter of the x-ray beam was at the center of the phantom, and the phantom was scanned centrally in the FOV. The first scan contained the noise/uniformity layer up to row 2. The CS 9300 was then moved higher without changing its position in any other plane of space until its vertical laser light included rows 3 and 4. A second scan was performed at this phantom position. The 8x8, 10x10, 10x5, 5x5 cm 2 FOVs were small which could not scan the total diameter of the phantom. Each column was scanned separately, and two different vertical positions were required to image the related IQ phantom height. The phantom was positioned so that the isocenter of the x-ray beam was at the center of corresponding column to be scanned; other 28

40 columns were scanned by rotating/moving the phantom that put its center at the isocenter of the x-ray beam. Inserts in the same column were imaged by simply lowering/raising the CS 9300, until the vertical positioning laser at least covered the noise/uniformity layer up to row 1 for the lower scan, and then covered row 3 and 4 for the upper scan Reproducibility Prior to initiating the scanning protocols, intraobserver variability, interobserver variability, and interscan reproducibility of the images obtained from the CS 9300 machine were determined for 17x11 cm 2 and 5x5 cm 2 FOVs with average adult settings to assess the consistency and reproducibility of the measurements. For interscan reproducibility, three phantom images were collected and analyzed by a single observer; the phantom was set up and taken down for each of three phantom images. For intraobserver variability, a single observer analyzed the images on three different dates, fourteen days apart. For interobserver variability, a second observer analyzed the same data. The standard deviation of measurements was calculated for the interscan and intraobserver variability consistency, and the difference of measurements was recorded for the interobserver variability Image quality parameters and image analysis in ImageJ To assess the image quality, we measured image quality metrics, including noise, uniformity, geometric distortion, artifacts induction tolerance, contrast resolution, and spatial resolution for the images of IQ phantom. These measurements were performed on the images of the corresponding inserts of the IQ phantom; so images of the noise/uniformity layer, geometric distortion layer, beam hardening inserts in row one, contrast resolution and pixel intensity inserts in row three, and spatial resolution inserts in row four were obtained. The images were transferred using a hard drive, as DICOM file from the CBCT computer to the office computer for more analysis. Each stack of images were imported to ImageJ 1.46r, and different tools in ImageJ were employed in order to select appropriate lines or ROIs for analysis of various regions in the IQ phantom. Then, the selected line or ROI were analyzed in ImageJ to measure the necessary data; mean gray value, and standard deviations (SD) were measured for selected ROI, 29

41 and line profile was measured for selected line. The data obtained from ImageJ was then imported into Microsoft Office Excel 2010 for manipulation and further analysis. For all measurements except geometric distortion and point spread function (PSF), the measurement was performed on five consecutive central axial slices to obtain a sufficient sample size. Measured parameters of a ROI (mean gray value, SD) were averaged over these slices for estimation of different image quality parameters. The top and bottom of the layer/insert were avoided because of possible interference by adjacent inserts. The center slice was visually determined in the orthogonal view, and then two upper and lower slices were considered Noise & Uniformity Image noise and image uniformity were measured on the images obtained by scanning the lower part of the phantom. For the 17x11, 17x6, and 5x5 cm 2 FOVs, five circular ROIs were placed at the FOV center (ROI 5) and periphery (ROIs 1, 2, 3, 4) on the center slice of noise/uniformity layer. For other FOVs, five circular ROIs were placed at the phantom image s center and periphery on the center slice of noise/uniformity layer. The ROI diameter for 17x11 and 17x6 cm 2 FOVs was 40 mm; however for other FOVs, it was 20% of the FOV diameter. (Fig. 4.1) a. b. Fig (a) noise/uniformity ROI for 17x11,17x6, and 5x5 cm 2 FOV, (b) noise/uniformity ROIs for 8x8, 10x10, and 10x5. Noise was defined as average SD of gray values of five ROIs. It was normalized to the difference of mean gray value (MGV) between PMMA and air regions to be comparable through different protocols. [30] 30

42 eq. 4.1 Uniformity was defined as the difference between mean gray values of central and peripheral ROIs. [30] eq. 4.2 where MGV peripheral was the average of the mean gray value of four peripheral ROIs. Uniformity was also represented by vertical and horizontal line profiles through the center of the central slice of noise/homogeneity layer Geometric distortion The distance between each void is nominally 10 mm in the IQ phantom. Periodicity of voids in the geometric distortion layer was defined as the geometric distortion parameter. The centers of the voids at the phantom image s center on the center slice of geometric distortion layer were connected horizontally and vertically. Line profile across these two perpendicular lines measured, and then line profiles entered into MATLAB to find the periodicity and compare two lines perpendicular to each other (Fig. 4.2.b). a. b. Fig (a) two perpendicular lines to measure line profiles for geometric distortion, (b) Plot the line profiles in MATLAB and calculate the periodicity 31

43 Artifact induction tolerance Artifact induction tolerance was analyzed qualitatively and quantitatively. For qualitative measurement, the average of mean gray value and SD of circular ROI (30 mm in diameter) for five central slices were measured for seven inserts in row 1 and compared (Fig. 4.3.a). a. b. Fig (a) Circular ROIs on each column for artifact analysis (b) rectangular ROIs around the artifact rods. For the quantification of metal artifacts, two metal artifact inserts were used, and then the artifact areas were analysed by measuring the extent of metal artifacts. The extent of the metal artifact was measured by subtraction of the PMMA s mean gray value of noise/uniformity layer from the average of mean gray value of central five slices of artifact insert over each rectangular ROIs (25x5 mm 2 ) surrounding the titanium rods (Fig 4.3.b). It was called artifact added value (AAV) which normalized to the difference of MGV between PMMA and air regions. eq. 4.3 The AAV for each protocol was the average of four AAVs for the four ROIs. Its value was mainly related to the amount of artifact from metal object on its vicinity area in image. 32

44 Contrast resolution Images of the contrast resolution inserts were viewed in ImageJ with 100% magnification and default contrast and brightness settings which ImageJ uploaded. The five different high and low contrast resolution inserts were analyzed by visually determining the diameter of the smallest visible rod when the centre slice for this insert was viewed. The minimum observable rod diameter was recorded Contrast to noise ratio Contrast to noise ratio (CNR) was defined as the ratio of the difference between mean gray value of PMMA and a material to the average of SD of both PMMA and the material, eq. 4.4 where MGV PMMA and SD PMMA are measured from the noise/uniformity layer; MGV material and SD material of a circular ROI are measured for each material s disk in the pixel intensity insert (figure 4.4). The pixel intensity insert provides MGV and SD values for Al, PTFE, POM, LDPE, and Air material. The centres of each of the five horizontal discs were visually determined using the orthogonal XZ views. 20 mm diameter circular ROI was used on five central slices of each disk concentric with the disk to measure the MGV material and SD material. Fig a 20 mm diameter ROI at the Al disk for CNR calculation 33

45 Spatial resolution I. Line pair charts The LP/mm inserts were used to determine the spatial resolution of the image semiquantitatively. The centres of both the axial (XY Plane) and vertical (Z plane) LP/mm inserts were determined by viewing the inserts in ImageJ to evaluate the axial (xy) and coronal (z) planes. The observation for this parameter is obtained by visually counting the smallest lp/mm for both xy and z plane inserts that are clearly visualized showing distinct black and white lines for each scanning protocol. II. PSF and MTF The spatial resolution was calculated quantitatively using the PSF insert. The PSF insert s image was imported into MATLAB. The centre of a rectangular ROI, which was 30x30 pixels, was placed at the wire. Ten adjacent central axial slices were averaged to generate a low noise PSF. Pixel values over this ROI were integrated in the longitudinal direction (integrating along either the x or y axis) to yield a one-dimensional PSF. The resulting distribution was plotted, and FWHM was calculated after normalizing the graph s tail to zero. The modulation transfer function (MTF) was calculated from one dimensional PSF using the fast Fourier transform method, and dividing by the spatial frequency spectrum described by the steel wire. [31],[14] The frequency at 10% of the MTF was used to calculate the spatial resolution. 34

46 a. b. c. d Dose measurement Fig (a) The ROI on the central axial slice of the PSF insert image. (b) Surface plot of the PSF insert over the ROI. (c) an integrated 1D profile of the PSF, (d) MTF obtained from the FFT of the 1D PSF. Dose measurements were performed with a method adapted from the SEDENTEXCT project, which suggested that CBCT dose can be estimated by DI1. Using this method, measurements were performed using a thimble ionization chamber along a diameter of the SEDENTEXCT DI phantom mounted on a tripod and positioned in the FOV for scanning. The measurement of DI1 for on-axis/off-axis exposures and full/partial dose distributions simply is executed by rotating the phantom in such a way that the isocentre of the x-ray beam lies on the measuring diameter as shown in Figure 4.6. Using this index, it is perfectly possible to use off-axis scanning (i.e. exposing the phantom as one would expose a patient) as well as partial scanning, as long as leftright symmetry of the measuring diameter is kept. 35

47 a. b. Fig The dose measurement points for DI1 using DI phantom. Number 1 is located at interior/right side of the phantom. The measuring diameter was determined from a dose distribution map for each protocol listed in table 3.1. Dose distribution maps were generated using Gafchromic TM film placed in the center slice of the phantom. The phantom was scanned in the same configuration as a patient would be scanned; center of the phantom was at the isocenter of 17x6 cm 2 FOV, then the isocenter of the x-ray was moved to the corresponding clinical position for each protocol. The center slice of phantom was nominally at the level of the central axial slice of the image volume. I performed five scans on the film at the same place to accumulate sufficient dose to be in the sensitive range of the film. Films were scanned in a HP scanner (HP LJ 1536dnf MFP), and then the dose distribution map was obtained using MATLAB. The dose maps provide visualization of heterogeneous dose distribution about the anterior, posterior and lateral aspects of the phantom. (17x6 and 8x8 cm 2 measuring diameter was perpendicular to the measuring diameter for all other FOV, which were in posterior-anterior direction). By determining the measuring diameter, dose measurements at five points along this line were performed, and then dose index was defined as the average dose along this line: The phantom was placed in the FOV the same as above. The thimble ionization chamber was placed into the DI phantom columns at the phantom s center slice. Measurements were nominally performed at the level of the central axial slice of the image volume. Two scans at 36

48 each point were taken for each protocol listed in table 4.1 to assess variation without overheating the x-ray tube and the data recorded for analysis. These two dose values reported in milligray were averaged and that value was used to report radiation dose at each point. Dose Area Product (DAP) values, as reported by the CS 9300 machine for each of the protocols, was also recorded and compared with the radiation doses obtained. Fig The columns used to measure the point dose. 37

49 Noise 5. RESULT In this section the results of image quality measurements and dosimetery analysis of CS 9300 are reported. The detailed results for different parameters can be found in appendix A. The reproducibility measurements were done only for image quality parameters. For dosimetery measurements, two readings of the central axial radiation dose at each of five points were obtained to consider the variation in ionization chamber readings Image quality performance Image quality parameters Noise and uniformity Noise was defined as the SD of gray values in PMMA layer divided by the mean gray value to be comparable between different protocols since the mean values for each protocol were totally different. As the results in figure 5.1 shows, 17x11 FOV had the greatest noise against the 8x8 FOV with the lowest noise value. The 5x5 FOV had noise values less than 17x11 FOV and larger than other FOVs. The noise values for 17x6 were larger than 10x10 which was larger than 10x5 FOV. For 17x11 and 17x6, the noise values were obtained from one image; the noise values for other FOVs were averaged of noise values of image of A, B, C, E, F, and G columns. Since column D is at center of the phantom, it had different mean gray values and SD, and then it was not included for noise calculation. It was the same for uniformity calculation x11 17x6 8x8 10x10 10x5 5x5 0 Large Average Small Child adult adult adult Fig Noise values for all protocols on CS9300 machine. 38

50 The uniformity of the images (raw images) was greater than 93.5% for all protocols. Table 5.2 shows two example of uniformity of the images; the uniformity graph is not representing the actual noise in the image since it was smoothed to be displayed. The uniformity was symmetric with a sagging in the middle of the individual graphs for 17x11 and 17x6 FOVs. For other FOVs, the uniformity graphs for peripheral column s images were not symmetric although it was symmetric for central column s image of the IQ phantom. For 5x5 it was due to the fact that x- ray passed through thicker medium on the inward side of the image rather than the peripheral side of the phantom image. It is also true for 8x8, 10x10, and 10x5 FOVs, in addition to the fact that the uniformity graph includes only about half of the image. Sagging was noted in the middle of the individual plots for each protocol which may actually represent a cupping artifact due to beam hardening which occurs when there is more beam attenuation in the center of an object than around the edges. a. b. Fig 5.2. The uniformity graphs including the noise in the image, which is not representing the actual noise in the image since it was smoothed for visualization. (a) Uniformity profile at center of image representing symmetrical uniformity sinking at middle as an example for 17x11, 17x6 (b) Uniformity profile at center of image is showing unsymmetrical sinking uniformity at middle of image for 5x5, 8x8, 10x10, and 10x5 FOVs. Noise was similar for different positioning of 5x5 FOV as seen in figure 5.3; considering the error bar of 1.14, the noise values for each protocol were in agreement with each other. The uniformity values within each setting agreed with each other when considering the error of

51 a. 17 Noise Large Average Small Child 7 front left left right right middle back middle back b Uniformity front left left right right middle back middle back Large Average Small Child Fig. 5.3 (a) Noise values were similar for different positioning for each setting. Also considering the error of 1.14, the noise values agreed within each setting. (b) There were more variations in uniformity over the image in different position of the 5x5 FOV which was the same by considering the error of Geometric distortion Geometric distortion was defined as the periodicity of the voids in the geometric distortion layer. The average periodicity of the voids for two perpendicular lines is shown for each protocol in table 1 in appendix A. The average periodicity was between mm for each setting that means ±0.05 mm geometric distortion. Geometric distortions for different protocols regarding different positioning of 5x5 FOV were consistent. All of them showed void periodicity of 10 except small adult settings for front, right back, and left back teeth show 10.1 mm periodicity. 40

52 Artifact induction tolerance The qualitative analysis of artifact induction tolerance for the two columns containing the artifacts induction tolerance inserts as well as the five other columns containing blank PMMA inserts provided an indication of the adjacent and surrounding effects of metal objects on the image quality. Figure 5.4 shows that two columns containing the three titanium rods (columns C and E) had the highest MGV and largest SD from these inserts. The column containing the blank PMMA insert that was located between the two artifacts induction inserts (column D) showed lowest MGV and a little bit higher SD than four other peripheral columns A, B, F, and G. There was no difference between the rod direction using 8x8, 10x10, 10x5, and 5x5 FOVs due to the same MGV for C and E columns and similar MGV for columns A, B, F, and G. However, the mean gray values of columns C and E were different as well as the MGV of A and B were not similar to the MGV of F and G for 17x11 and 17x6 the MGV. a. 200 MGV 1500 SD 0 A B C D E F G A B C D E F G b MGV 1500 SD A B C D E F G A B C D E F G Fig The mean gray value (MGV) and SD value at seven insert in artifact induction layer.(a) MGV and SD for 8x8, 10x10, 10x5, and 5x5 FOVs. (b) MGV and SD for 17x11 and 17x6 FOV. The artifact added value (AAV) parameter was defined to quantitatively evaluate the artifact induction tolerance. Its value represents the average of AAV obtained from four rectangular ROIs surrounding the titanium rods in columns C and E. The AAVs were between for all 41

53 protocols; except one protocol had the value of 0.2. No clear difference was seen between large/small FOVs or high/low dose protocols Contrast resolution The Contrast resolution of different high and low contrast resolution inserts were the same for all protocols. 1 mm rod was visible for all Al, Air, LDPE, and PTFE contrast resolution inserts, and 2 mm rod was visible for POM insert in all protocols and FOVs. The contrast resolutions were the same for all five positioning of the 5x5 FOV Contrast to noise ratio The material s MGV were averaged over five central slices MGV obtained for each material s disk from pixel intensity insert for each scan protocols. Aluminium (Al) had the greatest MGV value against air with the lowest value. MGV value of PTFE, POM, and LDPE were respectively lower than Al. The standard deviations were greatest for the Al inserts for all protocols which indicate the greatest amount of noise for this density material. The standard deviations were generally similar for the PTFE, POM, LDPE, and air materials. CNR ratios were calculated for the five different materials and the various scanning protocols using equation 4.4. The CNRs were fairly consistent for each material for all the scanning protocols with the same FOV except for the small FOV, 5x5 cm 2, which had the greatest range of CNR values for child to large adult settings. Figure 5.5 shows the CNR value results for different protocols within each FOV. The numeric values are also reported in table 3 in the appendix. Al and air had the greatest and most similar CNR values. There was a small difference between the CNR values of LDPE and POM values; LDPE had a little greater value than POM for all protocols as it was expected (due to the density difference between PMMA-LDPE is greater than the PMMA-POM). 42

54 CNR CNR CNR CNR CNR CNR 20 17x x Al PTFE POM LDPE Air 0 Al PTFE POM LDPE Air 20 17x6 20 8x Al PTFE POM LDPE Air 10x Al PTFE POM LDPE Air 10x Al PTFE POM LDPE Air Fig The CNR values vs. material densities are plotted for each FOV. Each graphs shows four curves, which correspond to child (blue), small adult (red), average adult (green), and large adult (purple). For different positioning of the 5x5 FOV, The CNR values were the same for all protocols, except for large adult which had a shift between front teeth protocol and other four protocols. 0 Al PTFE POM LDPE Air 43

55 CNR CNR CNR CNR 20 Large Adult 20 Average Adult Al PTFE POM LDPE Air 0 Al PTFE POM LDPE Air Small Adult Child Al PTFE POM LDPE Air 2 0 Al PTFE POM LDPE Air Fig CNR values for different material densities are plotted for settings of 5x5 FOV; Each graph shows CNR values of the same settings at five different positioning Spatial resolution I. LP per mm The results of the two line pair inserts observation for each protocol are represented in table 5.1. Although the CBCT has isotropic voxels, the LP/mm was not the same in xy plane and z direction for most protocols. Most protocols showed line pairs per mm greater than 1 LP/mm but all protocols showed that it was less than 2.5 LP/mm. 44

56 Table 5.1. LP/mm along XY plane and Z axis observed form the LP/mm insert are reported for different protocols on CS 9300 CBCT. FOV cm 2 Voxel size (μm) 17x x x x x x5 200 Patient type LP/mm Z LP/mm XY Child 0 1 Adult- small 0 1 Adult- average 0 1 Adult- large 0 1 Child Adult- small Adult- average Adult- large Child Adult- small Adult- average Adult- large Child Adult- small Adult- average Adult- large 1 1 Child Adult- small Adult- average Adult- large Child 2 2 Adult- small 2 2 Adult- average Adult- large 2 2 The LP/mm observation of the xy-plane insert was the same for all positioning of the 5x5 FOV. However, the observations of z-axis insert were varied between LP/mm. II. PSF and MTF FWHM for each protocol are represented in table 1 in the appendix. The FWHM were almost the same within each FOV. The comparison of FWHM against the voxel size is shown in figure 5.7; average of FWHM for each FOV are plotted vs. the voxel size. FWHM did not correlate with the 45

57 FWHM (mm) voxel size. Although 17x6, 10x10, 10x5, and 8x8 FOVs have the same voxel size as well as other x-ray tube parameters, FWHM were not the same for these protocols. It was similar for8x8, 10x10, and 10x5 FOVs due to the similar FOV diameter; however the height of FOVs were different. It was also the same for 17x11 and 17x6 which had the same FOV diameter rather than the same voxel size x mm 17x6-0.2 mm 8x mm 5x5-0.2 mm 10x mm 10x mm voxel size (mm) Fig The average of FWHM for each FOV is plotted vs. its voxel size. The spatial resolution was calculated for each protocol from the LP/mm measured at the 10% of the MTF graph. The results for both the LP/mm and resolution are shown in table 1 in the appendix. There were no correlation between the resolutions and voxel sizes. A protocol with the smallest voxel size did not show the highest MTF response; figure 5.8 shows MTFs of child setting for six FOVs. 46

58 Fig (a) The MTF graph for child setting for six FOVs, (b) The zoomed graph at 10% of the MTF The variation of FWHM and resolution for different positions of 5x5 FOV were more notable than other image quality parameters as it is shown in figure FWHM front left left right right middle back middle back Large Average Small Child Resolution front left left right right middle back middle back Large Average Small Child Fig The FWHM and resolution for different positioning of the 5x5. 47

59 Reproducibility of image quality parameters The reproducibility of image quality parameters for CS 9300 assessed through interscan, intraobserver, and interobserver variabilities represents the reproducibility of the phantom, the measurement method, and the observer s analysis. Standard deviations of the measurements obtained for three repeated scans using identical exposure settings (interscan) as well as three repeated measurements on the same image (intraobserver) are shown in table 5.2. Percentage difference was reported as the interobserver measurement for deviation between two different observers measurements in table 5.3. The SD or percentage difference for CNR is the average SD or percentage difference of the CNRs of five materials. Table 5.2. Reproducibility of measurements of three different images obtained with the same machine settings; values are the standard deviation of three measurements. Intraobserver variability of repeated measurements of the same image performed by a single observer; values are the standard deviation of three measurements obtained 14 days apart. SD for interscan variability SD for Intraobserver variability Image quality parameter 5 x 5 FOV 17 x 11 FOV 5X5 FOV 17X11FOV Uniformity Noise Geometric Distortion Artifact added value CNR LP/mm For interscan reproducibility (Table 5.2), the SDs for all IQ parameter values in the 5x5 FOV were larger than the 17x11 FOV because more image acquisitions and more adjustments were required to image the entire phantom for the 5x5 FOV. For intraobserver variability (Table 5.2), the 5x5 FOV had less variation than the 17x11 FOV except for AAV parameter since the 5x5 FOV was acquired with higher resolution, which reduced the variability in the measurements performed by a single observer. For interobserver reproducibility, the variation was less for noise,cnr, and LP/mm for 5x5 compared to 17x11 FOV; however it was larger for uniformity, geometric distortion, and AAV for 5x5 compared to 17x11 FOV. 48

60 Table 5.3. Interobserver variability obtained from measured values of the same image performed by two observers. The numbers in the table above represent percentage (%). Image quality parameter % Difference for Interobserver variability 5X5 FOV 17X11 FOV Uniformity Noise Geometric Distortion Artifact added value CNR 1 2 LP/mm The highest reproducibility was seen for uniformity and noise; on the other hand, metal artifact had the most deviation due to the large difference between gray values in the ROI, and sensitivity of the rectangular ROI s positioning near the Titanium rods (figure 4.3.b) Dose measurement Dose distribution was obtained as a 2D isodose contour graph of Gafchromic film in MATLAB. Dose distributions for 10 protocols are shown in figure 5.10 as the percentage of the maximum dose; dose distribution of FOVs 10x5 and 10x10 were the same due to the same partial rotation of the x-ray tube and isocenter position. As shown on the graph, the bottom of each 2D distribution represents the front (anterior) side of the phantom (patient), whereas the left side of the graph corresponds to the right side of the phantom (patient) placed in the FOV. The dose distributions were not uniform for all measurements because the device does about half rotation instead of a full rotation around the patient s head. As seen on the figures, a gradient of dose which can be higher in the posterior region or vice versa for FOVs 17x11, 10x10, 10x5, and 5x5, due to either a partial rotation (starting from the left-anterior and ending at right-anterior) or offaxis positioning. Also, a peculiar left-right asymmetry can be seen; one factor could be that the CS 9300 uses half beam scanning, although this should lead to a dose peak in the overlapping region, rather than a left-right asymmetry. It can also partly be due to a small positional error (rotation of the phantom towards the right side). So the measuring diameter for these FOVs was a line from posterior to the anterior of the phantom. For 17x6 and 8x8 FOVs, the dose distribution was the same as previous except that it was rotated 90 due to the half rotation starting from the left-posterior and ending at left-anterior. In that order the measuring diameter was a line from left to the right side of the phantom. 49

61 The average dose in the mid axial slice is defined as DI1 which was obtained as the average of five point dose measurements for each protocol in table 3.1 along the measuring diameter determined above. The data for DI1 as measured on the CS 9300 CBCT machine is shown in appendix 1. In general, the DI1 was in the range mgy; the highest value was for large adult, 17x11 FOV protocol and lowest DI1 for the child, 5x5 FOV, as we expected. DI1 for each protocol is shown in figure 5.12 as column bars. The dose distributions at points 1-5 for average adult settings for all FOVs are shown in figure The order of points were shown in figure 4.6 with point 1 at the anterior and the point 5 at the posterior and three other points between them spaces uniformly. Different settings with the same FOV had very similar dose distribution on the axial slice, as well as these five points. The 5x5 shows a different trend than the others; this was likely due to measurements inside the FOV and outside the FOV which was more notable for small 5x5 FOV rather than 8x8, 10x5, and 10x10 FOVs. Those FOVs cover mostly more than half of the phantom. Dose distribution is almost uniform at the measuring diameter for 8x8, 10x10, and10x5 FOVs. However there are two hot spots at two sides of this line on the periphery of the phantom. For other FOVs, dose distribution is not uniform on the measuring diameter; but the similar distribution can be found over the mid axial slice. (Fig. 5.11) 50

62 Fig Axial dose distribution for different FOV. The dose distribution was obtained just for average adult and child setting. Since all scanning parameters for 10x5 and 10x10 FOVs are the same, the dose distribution for both of them are the same, and dose distribution for 10x10 is not represented here. The red line represented the measuring diameter for DI1. The bottom and left side of dose distribution correspond to anterior and right side of the patient. The scale bar shows the distribution of dose as percentage of maximum dose in each image which 0% (white) represents no dose and 100% (black) represents maximum dose. FOV Child Average adult 5x5 10x5 8x8 (left TMJ) 17x6 17x11 51

63 point dose (mgy) x11 17x6 8x8 10x10 10x5 5x point measurements Fig The average of dose reading at each point along the measuring diameter is plotted. Points 1 to 5 were placed anterior to posterior uniformly. In general, DI1 increased for larger FOVs, higher ma and kv, smaller voxel sizes and longer scan times. It may be noted from Figure 5.12 below that FOV size is not the only parameter that affects dose but that small FOVs taken with higher ma and kv, longer exposure time, smaller voxel size may produce significant radiation exposures as may be seen with the 5x5cm 2 FOV, 8 ma, 84 kv, 12 s, and 200 μm (large adult) in comparison with 17x11cm 2 FOV, 4 ma, 80 kv, 6.40 s, 250 μm (child); These imaging parameters affect DI1 altogether. Fig DI1 for different FOVs and different settings related to child, small, average, and large adult. 52