Vol. 119 No. 3 March 2015
Relationship between physical factors and subjective image quality of cone-beam computed tomography images according to diagnostic task Jin-Woo Choi, DDS, MS,a Sam-Sun Lee, DDS, MSD, PhD,b Soon-Chul Choi, DDS, MSD, PhD,b Min-Suk Heo, DDS, MSD, PhD,b Kyung-Heo Huh, DDS, MSD, PhD,b Won-Jin Yi, PhD,b Se-Ryong Kang, BS,c Dong-Hun Han, DDS, MSD, PhD,d and Eun-Kyung Kim, DDS, MSD, PhDa Objective. This study was designed to investigate the relationship between physical factors and the subjective quality of cone beam computed tomography (CBCT) images used for different diagnostic tasks. Study Design. CBCT images of a real skull phantom and a SedentexCT IQ phantom were acquired under different exposure conditions (one Dinnova3 CBCT scanner, 60-110 kV and 4-10 mA). Radiologists evaluated subjective image quality of real skull phantom images for each diagnostic task. On the basis of the evaluation results, the images were classified into two groups: acceptable and unacceptable. The modulation transfer function (MTF), contrast-to-noise ratio (CNR), and image uniformity were measured using the SedentexCT IQ phantom images. The differences in physical factors were evaluated. Results. MTF and CNR values showed statistical differences in image quality in two groups with regard to all diagnostic tasks. In the maxilla, MTF and CNR values showed no significant differences between periapical diagnosis and implant planning in the acceptable groups. Higher MTF and CNR values were required in the periapical diagnosis compared with the implant planning of the mandible. Conclusions. This study proved that MTF and CNR values have a significant association with subjective image quality. The diagnostic task should be considered in evaluation of CBCT image quality. (Oral Surg Oral Med Oral Pathol Oral Radiol 2015;119:357-365)
Cone beam computed tomography (CBCT), which was introduced in 1998, involves lower radiation exposure and monetary costs compared with multi-detector computed tomography (MDCT).1,2 Although the radiation doses of CBCT are known to be lower than those of MDCT, they are still higher than those of conventional dental radiography. Therefore, the optimization of CBCT is essential for protection against radiation.2 Optimization can be achieved by reducing exposure while maintaining the image quality at an acceptable clinical level.2 However, many types of CBCT devices are currently available on the market, and the equipment features, radiation doses, and image quality of these units vary.1-3 For the optimization of these different devices, it is essential to assess the image quality by using a standardized method.4
a
Department of Oral and Maxillofacial Radiology, College of Dentistry, Dankook University, Seoul, Korea. b Department of Oral and Maxillofacial Radiology and Dental Research Institute, School of Dentistry, Seoul National University, Seoul, Korea. c Department of Biomedical Radiation Sciences, Graduate School of Convergence Science and Technology, Seoul National University, Seoul, Korea. d Department of Preventive and Social Dentistry, School of Dentistry, Seoul National University, Seoul, Korea. Received for publication Aug 1, 2014; returned for revision Nov 12, 2014; accepted for publication Nov 17, 2014. Ó 2015 Elsevier Inc. All rights reserved. 2212-4403/$ - see front matter http://dx.doi.org/10.1016/j.oooo.2014.11.010
Image quality can be assessed by using a subjective evaluation method or quantitative measurement of physical factors using test phantoms. CBCT images are used for many clinical purposes, such as implant planning, impacted teeth, trauma, and periapical diagnosis. Therefore, the evaluation of image quality should be based on the accomplishment of these various diagnostic tasks.5-7 Subjective evaluation is being used as the gold standard to assess the image quality in a task-based approach, but the possibility of standardizing this method is limiting due to its inherent subjectivity.3,5,6,8 Therefore, it is important to investigate the relationship between subjective evaluation and the physical qualities of CBCT images for the development of a standardized method. There is a general consensus that the physical factors of spatial resolution, contrast resolution, and image noise are related to clinical image quality and are used for quality control programs in multi-detector CT devices.9,10 Standardized phantoms are required to measure the physical
Statement of Clinical Relevance This study investigated the relationship between physical factors and subjective image quality, using a SedentexCT IQ phantom according to diagnostic tasks. Measurement of physical image quality can serve as a foothold for future studies on the quality control of various cone bean computed tomography (CBCT) devices. 357
ORAL AND MAXILLOFACIAL RADIOLOGY 358 Choi et al.
OOOO March 2015
UK). The phantom consists of a head-sized cylindrical phantom housing (diameter: 160 mm; height: 162 mm) made of polymethyl-methacrylate (PMMA) and 42 cylindrical inserts (diameter: 35 mm; height: 20 mm) (Figure 1). To obtain CBCT images with different image qualities, 24 combinations of six different tube voltages and four different tube currents were used for a skull phantom and a SedentexCT IQ phantom (60, 70, 80, 90, 100, and 110 kV; and 4, 6, 8, and 10 mA).
Fig. 1. SedentexCT IQ phantom and inserts.
image factors, and recently, the quality control phantom for CBCT was developed as part of the SedentexCT project.4 In the process of optimization, the acceptable clinical level of the image quality and the radiation dose may differ according to diagnostic tasks.2 Previous studies have reported that a reduction of the radiation dose can be achieved for implant planning with CBCT images.5,11,12 However, these studies did not measure the physical factors of an image, and research assessing the relationship between subjective evaluation and the physical image quality in CBCT devices is lacking. This study is designed to investigate the relationship between physical factors and the subjective evaluation of the quality of CBCT images in different diagnostic tasks.
MATERIAL AND METHODS CBCT system CBCT images were obtained by using a Dinnova3 CBCT scanner (HDXwill Inc., Seoul, Korea). Dinnova3 has an amorphous silicon flat panel detector. Further, the voxel size of 0.3 mm 0.3 mm 0.3 mm was used. A pulsed x-ray beam was rotated 360 degrees around the phantom, and the exposure time was 12 seconds (scan time: 24 seconds). A total filtration of 2.8-mm aluminum was used. The computed tomography dose index value was 3.183 mGy (120 kV, 120 mAs, field of view [FoV]: 200 mm 190 mm). The FoV of 200 mm 190 mm was used to obtain the complete image of the SedentexCT IQ phantom (SedentexCT IQ, Leeds Test Objects Ltd., Boroughbridge,
Measurement of physical image quality To measure the physical factors of the images, CBCT images of a SedentexCT IQ phantom were acquired under 24 exposure combinations. The images were saved in the digital imaging and communications in medicine (DICOM) format, and the MATLAB program (R2010 b, Mathworks Inc., Natick, MA) was used to analyze the modulation transfer function (MTF), contrast-to-noise ratio (CNR), and image uniformity. Under each exposure condition, a total of 640 slices of an axial image were loaded in the MATLAB program, and suitable inserts for each physical measurement were selected manually. All image factors were measured repeatedly for reliability: 10 times in the MTF measurement and three times in the other measurements. Modulation transfer function The point spread function (PSF) insert in row 4 of the phantom was used to calculate the MTF. The PSF insert is made of a stainless steel wire (diameter: 0.25 mm) suspended in air. MTF 50 and MTF 10 values were calculated using the fast Fourier transform method from the one-dimensional PSF. The wire area was magnified, and the image contrast was increased to find the exact center of the wire. When the window level and width were adjusted until the center of the wire was seen within 5 pixels and the brightest pixel was set as the center, the MTF values of four directions (vertical, horizontal, and two diagonal) were calculated automatically (Figure 2). In all, 10 slices of PSF insert images were used to obtain a reliable result. Contrast-to-noise ratio The pixel intensity value inserts in row 3 of the phantom were used to calculate the CNR. The inserts were made of five material disks (diameter: 25 mm) suspended in PMMA. Among them, aluminum (AL) and polyoxymethylene (POM) inserts were used to obtain image factors representing high- and low-contrast resolutions. The density values of AL, POM, and PMMA were 2.70, 1.42, and 1.20 g/cm3, respectively.
OOOO Volume 119, Number 3
ORIGINAL ARTICLE Choi et al. 359
Fig. 2. Steps of modulation transfer function (MTF) measurement. A, Setting a region of interest. B, Magnification of region of interest. C, Image adjustment for setting a wire center. D, Automated calculation of MTF values in four directions.
CNR ¼
mobj mbg sobj þ sbg 2
(1)
where mobj and mbg denote the average attenuation coefficients of the contrast material and the background, and sobj and sbg represent the standard deviation of the contrast material and the background. The region of interest was selected from the central area of each insert, and the marginal area was not included.
Fig. 3. Regions of interest for image uniformity (red circles).
Image uniformity The uniform PMMA layer in the base part of the phantom was used to calculate image uniformity. Image uniformity was calculated from the average pixel value difference between the center and the four peripheral regions of interest (upper, lower, right, and left regions) (Figure 3).
The CNR was calculated from the mean difference in the pixel value between the contrast material and the background (PMMA) divided by the average standard deviation within the material and the background, by using formula (1).
Subjective evaluation CBCT images of a real skull phantom with a soft-tissue replica (X-ray phantom, head; product number 7280, Erler Zimmer Co., Lauf, Germany) were acquired under
ORAL AND MAXILLOFACIAL RADIOLOGY 360 Choi et al.
OOOO March 2015
Fig. 4. Examples of images displayed for subjective evaluation of diagnostic acceptability. A-C, Axial, coronal, and sagittal images of the acceptable quality group (110 kV, 8 mA). D-F, Axial, coronal, and sagittal images of the unacceptable quality group (60 kV, 8 mA).
the same 24 combinations of tube voltages and currents. Images were saved in the DICOM format. All 24 sets of images were reconstructed into three planes (axial, coronal, and sagittal) with a slice thickness of 0.3 mm. All sectional images were presented to five radiologists for a subjective evaluation of image quality. In all, three 20.8-inch monochrome monitors (ME315 L, Totoku Electric Co., Tokyo, Japan) with a resolution of 2048 1536 pixels were used, and images of each plane were displayed on a different monitor (Figure 4). All observers had a trial session before the evaluation, and the evaluation was performed individually in a random, irreversible order. The observers were blinded to the exposure conditions, and they were allowed to adjust the brightness and the contrast of the images. Each observer evaluated the left maxillary first molar area first and then the right mandibular first molar area. To ensure reliability of the evaluation, observers were asked about the visibility of three anatomic structures before evaluating the image quality for the diagnostic tasks of periapical diagnosis and implant planning (Table I). The following six-grade scale was used to answer five statements: strongly agree (6), agree (5), slightly agree (4), slightly disagree (3), disagree (2), and strongly disagree (1). The evaluation was repeated after an interval of 2 weeks to calculate intraobserver reliability. We classified image quality into two groupsd acceptable and unacceptabledby using the two agreement criteria of the average score and the consensus. The average score criterion used an average observer score of 3.5 as the threshold for both the acceptable and unacceptable groups. In the case of the consensus criterion, only images that obtained a score of more than 4 from all the observers were defined as the acceptable group.
Table I. Evaluation statement of subjective image quality Maxilla: #26 Mesiobuccal root area
Mandible: #47 Mesial root area
1. Clear inferior border of maxillary sinus 2. Clear lamina dura and periodontal ligament space 3. Clear trabecular bone pattern 4. Image quality sufficient for periapical diagnosis 5. Image quality sufficient for implant planning
1. Clear border of mandibular canal 2. Clear lamina dura and periodontal ligament space 3. Clear trabecular bone pattern 4. Image quality sufficient for periapical diagnosis 5. Image quality sufficient for implant planning
Statistical analysis Intra- and interobserver reliabilities of the subjective evaluations were calculated as weighted k-value by using an online calculator (http://vassarstats.net/index. html). On the basis of the distribution of the samples, an independent t test or a ManneWhitney U test was used to evaluate the differences in the physical image factors (MTF 50, MTF 10, CNR AL, CNR POM, and uniformity) between the acceptable and the unacceptable image quality groups by using IBM SPSS Statistics version 21 (IBM Corp., Armonk, NY). In the case of physical image factors with significant differences, the receiver operating characteristic (ROC) curve was used to determine the cutoff value by using IBM SPSS Statistics version 21. An independent t test or a ManneWhitney U test was used to investigate the differences in physical factors according to the type of jaw or diagnostic task. A statistical significance level of P < .05 was used.
OOOO Volume 119, Number 3
ORIGINAL ARTICLE Choi et al. 361
Fig. 5. Modulation transfer function (MTF) values of each diagnostic task for both criteria. A, MTF 50 differences in the case of the average score criterion. B, MTF 50 differences in the case of the consensus criterion. C, MTF 10 differences in the case of the average score criterion. D, MTF 10 differences in the case of the consensus criterion. Ac, acceptable group; Uc, unacceptable group; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible.
RESULTS In the subjective evaluation, the average weighted kvalue was 0.59 (0.48-0.78) for intraobserver reliability and 0.41 (0.29-0.59) for interobserver reliability, corresponding to moderate agreement. Figure 4 shows examples of CBCT images from the acceptable and unacceptable groups. The physical measurement results for the different diagnostic tasks are shown in Figures 5, 6, and 7. There were no statistical differences in image uniformity between the acceptable group and the unacceptable group in most diagnostic tasks (see Figure 7). Image uniformity did not seem to have a considerable influence in the subjective evaluation. The MTF 50, MTF 10, CNR AL, and CNR POM showed statistical differences between the acceptable group and the unacceptable group in all diagnostic tasks irrespective of the criteria (P < .05) (see Figures 5 and 6).
An ROC curve was used to determine the cutoff value of MTFs and CNRs between the acceptable group and the unacceptable group. The area under the curve (AUC) and the cutoff values of MTFs and CNRs are given in Table II. In most cases, AUC was higher than 0.9 and the significance level was calculated as 0.000, which implies that the cutoff value is useful in the distinction between the acceptable and the unacceptable groups. The differences in MTF and CNR values of the acceptable group according to the type of jaw are presented in Table III. There were no significant differences in the MTF and CNR values between the maxilla and the mandible for the same tasks. The required physical image quality difference depending on the type of jaw was considered unnecessary. In the case of the maxilla, there were no statistical differences in the MTF and CNR values between the
ORAL AND MAXILLOFACIAL RADIOLOGY 362 Choi et al.
OOOO March 2015
Fig. 6. Contrast-to-noise ratio (CNR) values of each diagnostic task for both criteria. A, CNR AL differences in the case of the average score criterion. B, CNR AL differences in the case of the consensus criterion. C, CNR POM differences in the case of the average score criterion. D, CNR POM differences in the case of the consensus criterion. AL, aluminum; POM, polyoxymethylene; Ac, acceptable group; Uc, unacceptable group; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible.
Fig. 7. Uniformity values of each diagnostic task for both criteria. A, Standard deviation in the case of the average score criterion. B, Standard deviation in the case of the consensus criterion. Ac, acceptable group; Uc, unacceptable group; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible.
acceptable image groups of periapical diagnosis and implant planning. However, higher MTF and CNR values were required in the periapical diagnosis of the mandible than in the implant planning of the mandible (Table IV).
DISCUSSION This study investigated the relationship between physical factors and subjective image quality and attempted to assess the differences in the required physical image quality according to diagnostic tasks. The evaluation
OOOO Volume 119, Number 3
ORIGINAL ARTICLE Choi et al. 363
Table II. Cutoff value of MTF and CNR in the acceptable and unacceptable groups according to diagnostic task By average score criteria Task PD of Mx Cutoff value AUC P value IP of Mx Cutoff value AUC P value PD of Mn Cutoff value AUC P value IP of Mn Cutoff value AUC P value
By consensus criteria
MTF 50
MTF 10
CNR AL
CNR POM
MTF 50
MTF 10
CNR AL
CNR POM
0.37 0.980 .000
0.67 0.967 .000
41.64 0.989 .000
5.59 0.978 .000
0.42 0.958 .000
0.76 0.956 .000
53.25 0.964 .000
6.90 0.952 .000
0.36 0.982 .000
0.70 0.935 .000
34.47 0.988 .000
3.92 0.992 .000
0.37 0.965 .000
0.67 0.941 .000
38.77 0.981 .000
4.77 0.969 .000
0.39 0.988 .000
0.70 0.984 .000
46.27 0.994 .000
6.90 0.982 .000
0.43 0.855 .000
0.72 0.843 .000
56.94 0.897 .000
7.35 0.904 .000
0.36 0.982 .000
0.70 0.935 .000
34.47 0.988 .000
4.41 0.992 .000
0.36 1.000 .000
0.70 0.964 .000
38.77 0.981 .000
4.77 0.988 .000
MTF, modulation transfer function; CNR, contrast-to-noise ratio; AL, aluminum; POM, polyoxymethylene; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible; AUC, area under receiver operating characteristic (ROC) curve.
Table III. Differences in physical factors of acceptable groups according to the type of jaw By average score criteria
By consensus criteria
Task
Jaw
MTF 50
MTF 10
CNR AL
CNR POM
MTF 50
MTF 10
CNR AL
CNR POM
PD
Mx Mn Mx Mn
0.592*
0.592*
0.617y
0.610y
0.908*
0.993*
0.494*
0.316*
1.000*
1.000*
1.000y
1.000y
0.486*
0.511*
0.413*
0.469*
IP
MTF, modulation transfer function; CNR, contrast-to-noise ratio; AL, aluminum; POM, polyoxymethylene; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible. *Obtained using ManneWhitney U test. y Obtained using independent t test.
was performed using a single CBCT device, SedentexCT IQ phantom, and real skull phantom. There were significant differences in the MTF values between clinically acceptable and unacceptable images in all tasks. However, the usefulness of the MTF is not clearly proven, since the cutoff value of MTF 10 by ROC was low (0.67-0.76), and we are not sure whether the accomplishment of a diagnostic task depends on these small differences of spatial resolution or these differences in MTF values were just the result of the image quality classification by other factors. In addition, the spatial resolution can be influenced by various factors, such as voxel size, current, noise, FoV, and geometric accuracy.4,13 Spatial resolution can be measured by using a line pair chart or an MTF calculation. However, the use of a line pair insert of the SedentexCT IQ phantom will not yield a detailed result.13 In previous research, MTF values were calculated by measuring the PSF of one14 or two15,16 directions in CBCT images with the use of a wire-suspended phantom. In this study, we measured the MTF in four directions because in the pilot
measurement, the MTF values of the diagonal directions differed considerably from the vertical and horizontal values. This corresponds to the result obtained by Ozaki et al, which shows a difference between the MTF of the radial direction and that of the azimuthal direction.17 Further studies are required to develop a standardized MTF measurement of different CBCT images. Limited literature is available with regard to the MTF and diagnostic acceptability, but there are earlier studies on the voxel sizes and the diagnostic accuracy of CBCT images.8,18-20 Özer reported that there were no significant differences in accuracy in the diagnosis of a root fracture among CBCT images of different voxel sizes.8 In contrast, Wenzel et al reported higher sensitivity in high-resolution CBCT images than in low-resolution images.20 Liedke et al reported that the voxel size did not influence the diagnostic accuracy of external root resorption,18 whereas another study reported that high-resolution CBCT images performed similarly or better than low-resolution images in the detection of simulated internal resorption.19 Although there has
ORAL AND MAXILLOFACIAL RADIOLOGY 364 Choi et al.
OOOO March 2015
Table IV. Physical factors of acceptable groups according to diagnostic task By average score criteria Jaw Mx Mn
Task PD IP PD IP
MTF 50 *
MTF 10 *
CNR AL y
By consensus criteria CNR POM y
MTF 50 *
MTF 10 *
CNR AL *
CNR POM
0.97
0.110
0.760
0.570
0.052
0.054
0.053
0.075*
0.034*
0.037*
0.026y
0.015y
0.024*
0.035*
0.005*
0.004*
MTF, modulation transfer function; CNR, contrast-to-noise ratio; AL, aluminum; POM, polyoxymethylene; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible. *Obtained using ManneWhitney U test. y Obtained using independent t-test.
been disagreement over the voxel sizes and the diagnostic accuracy of CBCT images, spatial resolution is considered an important image quality factor in dental radiography, and further studies on the spatial resolution of CBCT images are required.2,18-21 It is unquestionable that the CNR can be a useful standard factor for evaluation of image quality, and previous studies have measured the CNR to evaluate the technical image quality of CBCT images.4,22-24 In this study, AL and POM inserts were used to obtain CNR values representing high- and low-contrast resolutions. Publication 93 of the International Commission on Radiologic Protection states that low-contrast resolution has a closer relationship to clinical image quality compared with high contrast resolution because high contrast resolution is not considerably influenced by an increase in noise level.9 Although CBCT in dentistry is mainly used for high-contrast anatomic structures, such as tooth and bone lesions, the periodontal ligament space and lamina dura, which have a low contrast compared with trabecular bone, are important anatomic structures and should be clearly depicted. Earlier studies have already revealed that the periodontal ligament space and lamina dura are less visible compared with other anatomic structures in several protocols with different CBCT devices.3,24 Therefore, low-contrast resolution can be considered an important image factor in the observation of the periodontal ligament space and lamina dura in CBCT devices. Bamba et al reported that a low-contrast-resolution phantom made of lowdensity polyethylene, POM, and polytetrafluoroethylene is useful in distinguishing between the performances of the devices’ contrast resolutions.14 Average CNR values of POM in the acceptable group ranged from 7.55 to 9.27, and the cutoff values were 3.92 to 7.35. Information from further research that uses different CBCT devices will be helpful in identifying the relationship between CNR values and the subjective image quality. However, image noise is influenced by the voxel size, and the CNR values from different devices should be compared with great care. Pauwels et al reported that CNR values are highly
device dependent due to the differences in hardware and software.4,24 The relationship between CNR values and the visibility of various-sized rod inserts are worth investigating. Because the measurement of CNR values using Image J or MATLAB software is difficult and time consuming, evaluating the image contrast by using the visibility of the rod inserts would be more convenient in the use of CBCT devices. Of interest, accreditation programs of MDCT devices now include a rod visibility test.10 There was no distinction in the type of jaw for the assessment of image quality on the same task, and there were no statistical differences in MTF and CNR values between the periapical diagnosis and the implant planning of the maxilla (see Table IV). These results do not correspond to a previous task-based study, which showed different decision levels between the upper and lower jaw and between implant planning and periapical diagnosis.5 It is speculated that the reason for these differences in our results is the use of different criteria in each study. Although the same six-grade scale was used, Lofthag-Hansen et al used separation criteria between “totally agree” and “agree,” whereas we used the criteria “acceptable” and “unacceptable” images. In the case of the mandible, higher MTF and CNR values were required to perform the task of periapical diagnosis compared with implant planning (see Table IV). This result corresponds to the result obtained by Lofthag-Hansen et al.5 The use of a single CBCT device with a large FoV is the limitation of this study. Therefore, this result or the cutoff values cannot be compared directly with those obtained in the case of other CBCT devices. Future work will focus on other CBCT devices and different exposure conditions.
CONCLUSIONS This study proves that the MTF and the CNR, among physical image factors, have a significant association with subjective image quality. In the case of the maxilla, the diagnostic task may be of little importance for the subjective evaluation of diagnostic acceptability.
OOOO Volume 119, Number 3
However, in the case of the mandible, the periapical diagnosis required better physical image quality compared with implant planning. This study was supported by grant no 04-2014-0076 from the SNUDH Research Fund.
ORIGINAL ARTICLE Choi et al. 365
14.
15.
16.
REFERENCES 1. Hatcher DC. Operational principles for cone-beam computed tomography. J Am Dent Assoc. 2010;141:3S-6S. 2. The SEDENTEXTCT Project. Radiation protection: cone beam CT for dental and maxillofacial radiology. Evidence based guidelines. Geneva, Switzerland: European Commission: 2011. Available at: http://www.sedentexct.eu/files/radiation_protection_ 172.pdf: Accessed December 23, 2014. 3. Liang X, Jacobs R, Hassan B, et al. A comparative evaluation of cone beam computed tomography (CBCT) and multi-slice CT (MSCT) Part I. On subjective image quality. Eur J Radiol. 2010;75:265-269. 4. Pauwels R, Stamatakis H, Manousaridis G, et al. Development and applicability of a quality control phantom for dental conebeam CT. J Appl Clin Med Phys. 2011;12:245-260. 5. Lofthag-Hansen S, Thilander-Klang A, Gröndahl K. Evaluation of subjective image quality in relation to diagnostic task for cone beam computed tomography with different fields of view. Eur J Radiol. 2011;80:483-488. 6. Tapiovaara M. Relationships between Physical Measurements and User Evaluation of Image Quality in Medical Radiology: A Review. Helsinki, Finland; Radiation and Nuclear Safety Authority; 2006. 7. Kalra MK, Maher MM, D’Souza RV, et al. Detection of urinary tract stones at low-radiation-dose CT with z-axis automatic tube current modulation: phantom and clinical studies. Radiology. 2005;235:523-529. 8. Özer SY. Detection of vertical root fractures by using cone beam computed tomography with variable voxel sizes in an in vitro model. J Endod. 2011;37:75-79. 9. International Commission on Radiological Protection. Managing Patient Dose in Digital Radiology. Ottawa, ON, Canada; ICRP Publication 93. Ann.; 2004. 10. McCollough CH, Bruesewitz MR, McNitt-Gray MF, et al. The phantom portion of the American College of Radiology (ACR) computed tomography (CT) accreditation program: practical tips, artifact examples, and pitfalls to avoid. Med Phys. 2004;31: 2423-2442. 11. Dawood A, Brown J, Sauret-Jackson V, Purkayastha S. Optimization of cone beam CT exposure for pre-surgical evaluation of the implant site. Dentomaxillofac Radiol. 2012;41:70-74. 12. Sur J, Seki K, Koizumi H, et al. Effects of tube current on conebeam computerized tomography image quality for presurgical implant planning in vitro. Oral Surg Oral Med Oral Pathol Oral Radiol. 2010;110:e29-e33. 13. Pauwels R, Beinsberger J, Stamatakis H, et al; SEDENTEXCT Project Consortium. Comparison of spatial and contrast resolution
17.
18.
19.
20.
21.
22.
23.
24.
for cone-beam computed tomography scanners. Oral Surg Oral Med Oral Pathol Oral Radiol. 2012;114:127-135. Bamba J, Araki K, Endo A, Okano T. Image quality assessment of three cone beam CT machines using the SEDENTEXCT CT phantom. Dentomaxillofac Radiol. 2013;42:20120445. Watanabe H, Honda E, Kurabayashi T. Modulation transfer function evaluation of cone beam computed tomography for dental use with the oversampling method. Dentomaxillofac Radiol. 2010;39:28-32. Xu J, Reh DD, Carey JP, et al. Technical assessment of a conebeam CT scanner for otolaryngology imaging: image quality, dose, and technique protocols. Med Phys. 2012;39:4932-4942. Ozaki Y, Watanabe H, Nomura Y, et al. Location dependency of the spatial resolution of cone beam computed tomography for dental use. Oral Surg Oral Med Oral Pathol Oral Radiol. 2013;116:648-655. Liedke GS, da Silveira HE, da Silveira HL, et al. Influence of voxel size in the diagnostic ability of cone beam tomography to evaluate simulated external root resorption. J Endod. 2009;35: 233-235. Kamburoglu K, Kursun S. A comparison of the diagnostic accuracy of CBCT images of different voxel resolutions used to detect simulated small internal resorption cavities. Int Endod J. 2010;43:798-807. Wenzel A, Haiter-Neto F, Frydenberg M, Kirkevang LL. Variable-resolution cone-beam computerized tomography with enhancement filtration compared with intraoral photostimulable phosphor radiography in detection of transverse root fractures in an in vitro model. Oral Surg Oral Med Oral Pathol Oral Radiol. 2009;108:939-945. Choi DH, Choi BR, Choi JW, et al. Reference line-pair values of panoramic radiographs using an arch-form phantom stand to assess clinical image quality. Imaging Sci Dent. 2013;43:7-15. Suomalainen A, Kiljunen T, Käser Y, et al. Dosimetry and image quality of four dental cone beam computed tomography scanners compared with multislice computed tomography scanners. Dentomaxillofac Radiol. 2009;38:367-378. Pauwels R, Silkosessak O, Jacobs R, et al. A pragmatic approach to determine the optimal kVp in cone beam CT: balancing contrast-to-noise ratio and radiation dose. Dentomaxillofac Radiol. 2014;43:20140059. Pauwels R, Seynaeve L, Bosmans H, et al. Leuven/BE: technical versus diagnostic image quality in dental CBCT imaging. Available at: http://posterng.netkey.at/esr/viewing/index.php? module¼viewing_poster&task¼&pi¼118196. Accessed December 23, 2014.
Reprint requests: Sam-Sun Lee Department of Oral and Maxillofacial Radiology School of Dentistry, Seoul National University Chang-kyeong-gung-ro 62-1 Jongno-gu Seoul 110-768 Korea [email protected]
Relationship between physical factors and subjective image quality of cone-beam computed tomography images according to diagnostic task Jin-Woo Choi, DDS, MS,a Sam-Sun Lee, DDS, MSD, PhD,b Soon-Chul Choi, DDS, MSD, PhD,b Min-Suk Heo, DDS, MSD, PhD,b Kyung-Heo Huh, DDS, MSD, PhD,b Won-Jin Yi, PhD,b Se-Ryong Kang, BS,c Dong-Hun Han, DDS, MSD, PhD,d and Eun-Kyung Kim, DDS, MSD, PhDa Objective. This study was designed to investigate the relationship between physical factors and the subjective quality of cone beam computed tomography (CBCT) images used for different diagnostic tasks. Study Design. CBCT images of a real skull phantom and a SedentexCT IQ phantom were acquired under different exposure conditions (one Dinnova3 CBCT scanner, 60-110 kV and 4-10 mA). Radiologists evaluated subjective image quality of real skull phantom images for each diagnostic task. On the basis of the evaluation results, the images were classified into two groups: acceptable and unacceptable. The modulation transfer function (MTF), contrast-to-noise ratio (CNR), and image uniformity were measured using the SedentexCT IQ phantom images. The differences in physical factors were evaluated. Results. MTF and CNR values showed statistical differences in image quality in two groups with regard to all diagnostic tasks. In the maxilla, MTF and CNR values showed no significant differences between periapical diagnosis and implant planning in the acceptable groups. Higher MTF and CNR values were required in the periapical diagnosis compared with the implant planning of the mandible. Conclusions. This study proved that MTF and CNR values have a significant association with subjective image quality. The diagnostic task should be considered in evaluation of CBCT image quality. (Oral Surg Oral Med Oral Pathol Oral Radiol 2015;119:357-365)
Cone beam computed tomography (CBCT), which was introduced in 1998, involves lower radiation exposure and monetary costs compared with multi-detector computed tomography (MDCT).1,2 Although the radiation doses of CBCT are known to be lower than those of MDCT, they are still higher than those of conventional dental radiography. Therefore, the optimization of CBCT is essential for protection against radiation.2 Optimization can be achieved by reducing exposure while maintaining the image quality at an acceptable clinical level.2 However, many types of CBCT devices are currently available on the market, and the equipment features, radiation doses, and image quality of these units vary.1-3 For the optimization of these different devices, it is essential to assess the image quality by using a standardized method.4
a
Department of Oral and Maxillofacial Radiology, College of Dentistry, Dankook University, Seoul, Korea. b Department of Oral and Maxillofacial Radiology and Dental Research Institute, School of Dentistry, Seoul National University, Seoul, Korea. c Department of Biomedical Radiation Sciences, Graduate School of Convergence Science and Technology, Seoul National University, Seoul, Korea. d Department of Preventive and Social Dentistry, School of Dentistry, Seoul National University, Seoul, Korea. Received for publication Aug 1, 2014; returned for revision Nov 12, 2014; accepted for publication Nov 17, 2014. Ó 2015 Elsevier Inc. All rights reserved. 2212-4403/$ - see front matter http://dx.doi.org/10.1016/j.oooo.2014.11.010
Image quality can be assessed by using a subjective evaluation method or quantitative measurement of physical factors using test phantoms. CBCT images are used for many clinical purposes, such as implant planning, impacted teeth, trauma, and periapical diagnosis. Therefore, the evaluation of image quality should be based on the accomplishment of these various diagnostic tasks.5-7 Subjective evaluation is being used as the gold standard to assess the image quality in a task-based approach, but the possibility of standardizing this method is limiting due to its inherent subjectivity.3,5,6,8 Therefore, it is important to investigate the relationship between subjective evaluation and the physical qualities of CBCT images for the development of a standardized method. There is a general consensus that the physical factors of spatial resolution, contrast resolution, and image noise are related to clinical image quality and are used for quality control programs in multi-detector CT devices.9,10 Standardized phantoms are required to measure the physical
Statement of Clinical Relevance This study investigated the relationship between physical factors and subjective image quality, using a SedentexCT IQ phantom according to diagnostic tasks. Measurement of physical image quality can serve as a foothold for future studies on the quality control of various cone bean computed tomography (CBCT) devices. 357
ORAL AND MAXILLOFACIAL RADIOLOGY 358 Choi et al.
OOOO March 2015
UK). The phantom consists of a head-sized cylindrical phantom housing (diameter: 160 mm; height: 162 mm) made of polymethyl-methacrylate (PMMA) and 42 cylindrical inserts (diameter: 35 mm; height: 20 mm) (Figure 1). To obtain CBCT images with different image qualities, 24 combinations of six different tube voltages and four different tube currents were used for a skull phantom and a SedentexCT IQ phantom (60, 70, 80, 90, 100, and 110 kV; and 4, 6, 8, and 10 mA).
Fig. 1. SedentexCT IQ phantom and inserts.
image factors, and recently, the quality control phantom for CBCT was developed as part of the SedentexCT project.4 In the process of optimization, the acceptable clinical level of the image quality and the radiation dose may differ according to diagnostic tasks.2 Previous studies have reported that a reduction of the radiation dose can be achieved for implant planning with CBCT images.5,11,12 However, these studies did not measure the physical factors of an image, and research assessing the relationship between subjective evaluation and the physical image quality in CBCT devices is lacking. This study is designed to investigate the relationship between physical factors and the subjective evaluation of the quality of CBCT images in different diagnostic tasks.
MATERIAL AND METHODS CBCT system CBCT images were obtained by using a Dinnova3 CBCT scanner (HDXwill Inc., Seoul, Korea). Dinnova3 has an amorphous silicon flat panel detector. Further, the voxel size of 0.3 mm 0.3 mm 0.3 mm was used. A pulsed x-ray beam was rotated 360 degrees around the phantom, and the exposure time was 12 seconds (scan time: 24 seconds). A total filtration of 2.8-mm aluminum was used. The computed tomography dose index value was 3.183 mGy (120 kV, 120 mAs, field of view [FoV]: 200 mm 190 mm). The FoV of 200 mm 190 mm was used to obtain the complete image of the SedentexCT IQ phantom (SedentexCT IQ, Leeds Test Objects Ltd., Boroughbridge,
Measurement of physical image quality To measure the physical factors of the images, CBCT images of a SedentexCT IQ phantom were acquired under 24 exposure combinations. The images were saved in the digital imaging and communications in medicine (DICOM) format, and the MATLAB program (R2010 b, Mathworks Inc., Natick, MA) was used to analyze the modulation transfer function (MTF), contrast-to-noise ratio (CNR), and image uniformity. Under each exposure condition, a total of 640 slices of an axial image were loaded in the MATLAB program, and suitable inserts for each physical measurement were selected manually. All image factors were measured repeatedly for reliability: 10 times in the MTF measurement and three times in the other measurements. Modulation transfer function The point spread function (PSF) insert in row 4 of the phantom was used to calculate the MTF. The PSF insert is made of a stainless steel wire (diameter: 0.25 mm) suspended in air. MTF 50 and MTF 10 values were calculated using the fast Fourier transform method from the one-dimensional PSF. The wire area was magnified, and the image contrast was increased to find the exact center of the wire. When the window level and width were adjusted until the center of the wire was seen within 5 pixels and the brightest pixel was set as the center, the MTF values of four directions (vertical, horizontal, and two diagonal) were calculated automatically (Figure 2). In all, 10 slices of PSF insert images were used to obtain a reliable result. Contrast-to-noise ratio The pixel intensity value inserts in row 3 of the phantom were used to calculate the CNR. The inserts were made of five material disks (diameter: 25 mm) suspended in PMMA. Among them, aluminum (AL) and polyoxymethylene (POM) inserts were used to obtain image factors representing high- and low-contrast resolutions. The density values of AL, POM, and PMMA were 2.70, 1.42, and 1.20 g/cm3, respectively.
OOOO Volume 119, Number 3
ORIGINAL ARTICLE Choi et al. 359
Fig. 2. Steps of modulation transfer function (MTF) measurement. A, Setting a region of interest. B, Magnification of region of interest. C, Image adjustment for setting a wire center. D, Automated calculation of MTF values in four directions.
CNR ¼
mobj mbg sobj þ sbg 2
(1)
where mobj and mbg denote the average attenuation coefficients of the contrast material and the background, and sobj and sbg represent the standard deviation of the contrast material and the background. The region of interest was selected from the central area of each insert, and the marginal area was not included.
Fig. 3. Regions of interest for image uniformity (red circles).
Image uniformity The uniform PMMA layer in the base part of the phantom was used to calculate image uniformity. Image uniformity was calculated from the average pixel value difference between the center and the four peripheral regions of interest (upper, lower, right, and left regions) (Figure 3).
The CNR was calculated from the mean difference in the pixel value between the contrast material and the background (PMMA) divided by the average standard deviation within the material and the background, by using formula (1).
Subjective evaluation CBCT images of a real skull phantom with a soft-tissue replica (X-ray phantom, head; product number 7280, Erler Zimmer Co., Lauf, Germany) were acquired under
ORAL AND MAXILLOFACIAL RADIOLOGY 360 Choi et al.
OOOO March 2015
Fig. 4. Examples of images displayed for subjective evaluation of diagnostic acceptability. A-C, Axial, coronal, and sagittal images of the acceptable quality group (110 kV, 8 mA). D-F, Axial, coronal, and sagittal images of the unacceptable quality group (60 kV, 8 mA).
the same 24 combinations of tube voltages and currents. Images were saved in the DICOM format. All 24 sets of images were reconstructed into three planes (axial, coronal, and sagittal) with a slice thickness of 0.3 mm. All sectional images were presented to five radiologists for a subjective evaluation of image quality. In all, three 20.8-inch monochrome monitors (ME315 L, Totoku Electric Co., Tokyo, Japan) with a resolution of 2048 1536 pixels were used, and images of each plane were displayed on a different monitor (Figure 4). All observers had a trial session before the evaluation, and the evaluation was performed individually in a random, irreversible order. The observers were blinded to the exposure conditions, and they were allowed to adjust the brightness and the contrast of the images. Each observer evaluated the left maxillary first molar area first and then the right mandibular first molar area. To ensure reliability of the evaluation, observers were asked about the visibility of three anatomic structures before evaluating the image quality for the diagnostic tasks of periapical diagnosis and implant planning (Table I). The following six-grade scale was used to answer five statements: strongly agree (6), agree (5), slightly agree (4), slightly disagree (3), disagree (2), and strongly disagree (1). The evaluation was repeated after an interval of 2 weeks to calculate intraobserver reliability. We classified image quality into two groupsd acceptable and unacceptabledby using the two agreement criteria of the average score and the consensus. The average score criterion used an average observer score of 3.5 as the threshold for both the acceptable and unacceptable groups. In the case of the consensus criterion, only images that obtained a score of more than 4 from all the observers were defined as the acceptable group.
Table I. Evaluation statement of subjective image quality Maxilla: #26 Mesiobuccal root area
Mandible: #47 Mesial root area
1. Clear inferior border of maxillary sinus 2. Clear lamina dura and periodontal ligament space 3. Clear trabecular bone pattern 4. Image quality sufficient for periapical diagnosis 5. Image quality sufficient for implant planning
1. Clear border of mandibular canal 2. Clear lamina dura and periodontal ligament space 3. Clear trabecular bone pattern 4. Image quality sufficient for periapical diagnosis 5. Image quality sufficient for implant planning
Statistical analysis Intra- and interobserver reliabilities of the subjective evaluations were calculated as weighted k-value by using an online calculator (http://vassarstats.net/index. html). On the basis of the distribution of the samples, an independent t test or a ManneWhitney U test was used to evaluate the differences in the physical image factors (MTF 50, MTF 10, CNR AL, CNR POM, and uniformity) between the acceptable and the unacceptable image quality groups by using IBM SPSS Statistics version 21 (IBM Corp., Armonk, NY). In the case of physical image factors with significant differences, the receiver operating characteristic (ROC) curve was used to determine the cutoff value by using IBM SPSS Statistics version 21. An independent t test or a ManneWhitney U test was used to investigate the differences in physical factors according to the type of jaw or diagnostic task. A statistical significance level of P < .05 was used.
OOOO Volume 119, Number 3
ORIGINAL ARTICLE Choi et al. 361
Fig. 5. Modulation transfer function (MTF) values of each diagnostic task for both criteria. A, MTF 50 differences in the case of the average score criterion. B, MTF 50 differences in the case of the consensus criterion. C, MTF 10 differences in the case of the average score criterion. D, MTF 10 differences in the case of the consensus criterion. Ac, acceptable group; Uc, unacceptable group; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible.
RESULTS In the subjective evaluation, the average weighted kvalue was 0.59 (0.48-0.78) for intraobserver reliability and 0.41 (0.29-0.59) for interobserver reliability, corresponding to moderate agreement. Figure 4 shows examples of CBCT images from the acceptable and unacceptable groups. The physical measurement results for the different diagnostic tasks are shown in Figures 5, 6, and 7. There were no statistical differences in image uniformity between the acceptable group and the unacceptable group in most diagnostic tasks (see Figure 7). Image uniformity did not seem to have a considerable influence in the subjective evaluation. The MTF 50, MTF 10, CNR AL, and CNR POM showed statistical differences between the acceptable group and the unacceptable group in all diagnostic tasks irrespective of the criteria (P < .05) (see Figures 5 and 6).
An ROC curve was used to determine the cutoff value of MTFs and CNRs between the acceptable group and the unacceptable group. The area under the curve (AUC) and the cutoff values of MTFs and CNRs are given in Table II. In most cases, AUC was higher than 0.9 and the significance level was calculated as 0.000, which implies that the cutoff value is useful in the distinction between the acceptable and the unacceptable groups. The differences in MTF and CNR values of the acceptable group according to the type of jaw are presented in Table III. There were no significant differences in the MTF and CNR values between the maxilla and the mandible for the same tasks. The required physical image quality difference depending on the type of jaw was considered unnecessary. In the case of the maxilla, there were no statistical differences in the MTF and CNR values between the
ORAL AND MAXILLOFACIAL RADIOLOGY 362 Choi et al.
OOOO March 2015
Fig. 6. Contrast-to-noise ratio (CNR) values of each diagnostic task for both criteria. A, CNR AL differences in the case of the average score criterion. B, CNR AL differences in the case of the consensus criterion. C, CNR POM differences in the case of the average score criterion. D, CNR POM differences in the case of the consensus criterion. AL, aluminum; POM, polyoxymethylene; Ac, acceptable group; Uc, unacceptable group; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible.
Fig. 7. Uniformity values of each diagnostic task for both criteria. A, Standard deviation in the case of the average score criterion. B, Standard deviation in the case of the consensus criterion. Ac, acceptable group; Uc, unacceptable group; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible.
acceptable image groups of periapical diagnosis and implant planning. However, higher MTF and CNR values were required in the periapical diagnosis of the mandible than in the implant planning of the mandible (Table IV).
DISCUSSION This study investigated the relationship between physical factors and subjective image quality and attempted to assess the differences in the required physical image quality according to diagnostic tasks. The evaluation
OOOO Volume 119, Number 3
ORIGINAL ARTICLE Choi et al. 363
Table II. Cutoff value of MTF and CNR in the acceptable and unacceptable groups according to diagnostic task By average score criteria Task PD of Mx Cutoff value AUC P value IP of Mx Cutoff value AUC P value PD of Mn Cutoff value AUC P value IP of Mn Cutoff value AUC P value
By consensus criteria
MTF 50
MTF 10
CNR AL
CNR POM
MTF 50
MTF 10
CNR AL
CNR POM
0.37 0.980 .000
0.67 0.967 .000
41.64 0.989 .000
5.59 0.978 .000
0.42 0.958 .000
0.76 0.956 .000
53.25 0.964 .000
6.90 0.952 .000
0.36 0.982 .000
0.70 0.935 .000
34.47 0.988 .000
3.92 0.992 .000
0.37 0.965 .000
0.67 0.941 .000
38.77 0.981 .000
4.77 0.969 .000
0.39 0.988 .000
0.70 0.984 .000
46.27 0.994 .000
6.90 0.982 .000
0.43 0.855 .000
0.72 0.843 .000
56.94 0.897 .000
7.35 0.904 .000
0.36 0.982 .000
0.70 0.935 .000
34.47 0.988 .000
4.41 0.992 .000
0.36 1.000 .000
0.70 0.964 .000
38.77 0.981 .000
4.77 0.988 .000
MTF, modulation transfer function; CNR, contrast-to-noise ratio; AL, aluminum; POM, polyoxymethylene; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible; AUC, area under receiver operating characteristic (ROC) curve.
Table III. Differences in physical factors of acceptable groups according to the type of jaw By average score criteria
By consensus criteria
Task
Jaw
MTF 50
MTF 10
CNR AL
CNR POM
MTF 50
MTF 10
CNR AL
CNR POM
PD
Mx Mn Mx Mn
0.592*
0.592*
0.617y
0.610y
0.908*
0.993*
0.494*
0.316*
1.000*
1.000*
1.000y
1.000y
0.486*
0.511*
0.413*
0.469*
IP
MTF, modulation transfer function; CNR, contrast-to-noise ratio; AL, aluminum; POM, polyoxymethylene; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible. *Obtained using ManneWhitney U test. y Obtained using independent t test.
was performed using a single CBCT device, SedentexCT IQ phantom, and real skull phantom. There were significant differences in the MTF values between clinically acceptable and unacceptable images in all tasks. However, the usefulness of the MTF is not clearly proven, since the cutoff value of MTF 10 by ROC was low (0.67-0.76), and we are not sure whether the accomplishment of a diagnostic task depends on these small differences of spatial resolution or these differences in MTF values were just the result of the image quality classification by other factors. In addition, the spatial resolution can be influenced by various factors, such as voxel size, current, noise, FoV, and geometric accuracy.4,13 Spatial resolution can be measured by using a line pair chart or an MTF calculation. However, the use of a line pair insert of the SedentexCT IQ phantom will not yield a detailed result.13 In previous research, MTF values were calculated by measuring the PSF of one14 or two15,16 directions in CBCT images with the use of a wire-suspended phantom. In this study, we measured the MTF in four directions because in the pilot
measurement, the MTF values of the diagonal directions differed considerably from the vertical and horizontal values. This corresponds to the result obtained by Ozaki et al, which shows a difference between the MTF of the radial direction and that of the azimuthal direction.17 Further studies are required to develop a standardized MTF measurement of different CBCT images. Limited literature is available with regard to the MTF and diagnostic acceptability, but there are earlier studies on the voxel sizes and the diagnostic accuracy of CBCT images.8,18-20 Özer reported that there were no significant differences in accuracy in the diagnosis of a root fracture among CBCT images of different voxel sizes.8 In contrast, Wenzel et al reported higher sensitivity in high-resolution CBCT images than in low-resolution images.20 Liedke et al reported that the voxel size did not influence the diagnostic accuracy of external root resorption,18 whereas another study reported that high-resolution CBCT images performed similarly or better than low-resolution images in the detection of simulated internal resorption.19 Although there has
ORAL AND MAXILLOFACIAL RADIOLOGY 364 Choi et al.
OOOO March 2015
Table IV. Physical factors of acceptable groups according to diagnostic task By average score criteria Jaw Mx Mn
Task PD IP PD IP
MTF 50 *
MTF 10 *
CNR AL y
By consensus criteria CNR POM y
MTF 50 *
MTF 10 *
CNR AL *
CNR POM
0.97
0.110
0.760
0.570
0.052
0.054
0.053
0.075*
0.034*
0.037*
0.026y
0.015y
0.024*
0.035*
0.005*
0.004*
MTF, modulation transfer function; CNR, contrast-to-noise ratio; AL, aluminum; POM, polyoxymethylene; PD, periapical diagnosis; IP, implant planning; Mx, maxilla; Mn, mandible. *Obtained using ManneWhitney U test. y Obtained using independent t-test.
been disagreement over the voxel sizes and the diagnostic accuracy of CBCT images, spatial resolution is considered an important image quality factor in dental radiography, and further studies on the spatial resolution of CBCT images are required.2,18-21 It is unquestionable that the CNR can be a useful standard factor for evaluation of image quality, and previous studies have measured the CNR to evaluate the technical image quality of CBCT images.4,22-24 In this study, AL and POM inserts were used to obtain CNR values representing high- and low-contrast resolutions. Publication 93 of the International Commission on Radiologic Protection states that low-contrast resolution has a closer relationship to clinical image quality compared with high contrast resolution because high contrast resolution is not considerably influenced by an increase in noise level.9 Although CBCT in dentistry is mainly used for high-contrast anatomic structures, such as tooth and bone lesions, the periodontal ligament space and lamina dura, which have a low contrast compared with trabecular bone, are important anatomic structures and should be clearly depicted. Earlier studies have already revealed that the periodontal ligament space and lamina dura are less visible compared with other anatomic structures in several protocols with different CBCT devices.3,24 Therefore, low-contrast resolution can be considered an important image factor in the observation of the periodontal ligament space and lamina dura in CBCT devices. Bamba et al reported that a low-contrast-resolution phantom made of lowdensity polyethylene, POM, and polytetrafluoroethylene is useful in distinguishing between the performances of the devices’ contrast resolutions.14 Average CNR values of POM in the acceptable group ranged from 7.55 to 9.27, and the cutoff values were 3.92 to 7.35. Information from further research that uses different CBCT devices will be helpful in identifying the relationship between CNR values and the subjective image quality. However, image noise is influenced by the voxel size, and the CNR values from different devices should be compared with great care. Pauwels et al reported that CNR values are highly
device dependent due to the differences in hardware and software.4,24 The relationship between CNR values and the visibility of various-sized rod inserts are worth investigating. Because the measurement of CNR values using Image J or MATLAB software is difficult and time consuming, evaluating the image contrast by using the visibility of the rod inserts would be more convenient in the use of CBCT devices. Of interest, accreditation programs of MDCT devices now include a rod visibility test.10 There was no distinction in the type of jaw for the assessment of image quality on the same task, and there were no statistical differences in MTF and CNR values between the periapical diagnosis and the implant planning of the maxilla (see Table IV). These results do not correspond to a previous task-based study, which showed different decision levels between the upper and lower jaw and between implant planning and periapical diagnosis.5 It is speculated that the reason for these differences in our results is the use of different criteria in each study. Although the same six-grade scale was used, Lofthag-Hansen et al used separation criteria between “totally agree” and “agree,” whereas we used the criteria “acceptable” and “unacceptable” images. In the case of the mandible, higher MTF and CNR values were required to perform the task of periapical diagnosis compared with implant planning (see Table IV). This result corresponds to the result obtained by Lofthag-Hansen et al.5 The use of a single CBCT device with a large FoV is the limitation of this study. Therefore, this result or the cutoff values cannot be compared directly with those obtained in the case of other CBCT devices. Future work will focus on other CBCT devices and different exposure conditions.
CONCLUSIONS This study proves that the MTF and the CNR, among physical image factors, have a significant association with subjective image quality. In the case of the maxilla, the diagnostic task may be of little importance for the subjective evaluation of diagnostic acceptability.
OOOO Volume 119, Number 3
However, in the case of the mandible, the periapical diagnosis required better physical image quality compared with implant planning. This study was supported by grant no 04-2014-0076 from the SNUDH Research Fund.
ORIGINAL ARTICLE Choi et al. 365
14.
15.
16.
REFERENCES 1. Hatcher DC. Operational principles for cone-beam computed tomography. J Am Dent Assoc. 2010;141:3S-6S. 2. The SEDENTEXTCT Project. Radiation protection: cone beam CT for dental and maxillofacial radiology. Evidence based guidelines. Geneva, Switzerland: European Commission: 2011. Available at: http://www.sedentexct.eu/files/radiation_protection_ 172.pdf: Accessed December 23, 2014. 3. Liang X, Jacobs R, Hassan B, et al. A comparative evaluation of cone beam computed tomography (CBCT) and multi-slice CT (MSCT) Part I. On subjective image quality. Eur J Radiol. 2010;75:265-269. 4. Pauwels R, Stamatakis H, Manousaridis G, et al. Development and applicability of a quality control phantom for dental conebeam CT. J Appl Clin Med Phys. 2011;12:245-260. 5. Lofthag-Hansen S, Thilander-Klang A, Gröndahl K. Evaluation of subjective image quality in relation to diagnostic task for cone beam computed tomography with different fields of view. Eur J Radiol. 2011;80:483-488. 6. Tapiovaara M. Relationships between Physical Measurements and User Evaluation of Image Quality in Medical Radiology: A Review. Helsinki, Finland; Radiation and Nuclear Safety Authority; 2006. 7. Kalra MK, Maher MM, D’Souza RV, et al. Detection of urinary tract stones at low-radiation-dose CT with z-axis automatic tube current modulation: phantom and clinical studies. Radiology. 2005;235:523-529. 8. Özer SY. Detection of vertical root fractures by using cone beam computed tomography with variable voxel sizes in an in vitro model. J Endod. 2011;37:75-79. 9. International Commission on Radiological Protection. Managing Patient Dose in Digital Radiology. Ottawa, ON, Canada; ICRP Publication 93. Ann.; 2004. 10. McCollough CH, Bruesewitz MR, McNitt-Gray MF, et al. The phantom portion of the American College of Radiology (ACR) computed tomography (CT) accreditation program: practical tips, artifact examples, and pitfalls to avoid. Med Phys. 2004;31: 2423-2442. 11. Dawood A, Brown J, Sauret-Jackson V, Purkayastha S. Optimization of cone beam CT exposure for pre-surgical evaluation of the implant site. Dentomaxillofac Radiol. 2012;41:70-74. 12. Sur J, Seki K, Koizumi H, et al. Effects of tube current on conebeam computerized tomography image quality for presurgical implant planning in vitro. Oral Surg Oral Med Oral Pathol Oral Radiol. 2010;110:e29-e33. 13. Pauwels R, Beinsberger J, Stamatakis H, et al; SEDENTEXCT Project Consortium. Comparison of spatial and contrast resolution
17.
18.
19.
20.
21.
22.
23.
24.
for cone-beam computed tomography scanners. Oral Surg Oral Med Oral Pathol Oral Radiol. 2012;114:127-135. Bamba J, Araki K, Endo A, Okano T. Image quality assessment of three cone beam CT machines using the SEDENTEXCT CT phantom. Dentomaxillofac Radiol. 2013;42:20120445. Watanabe H, Honda E, Kurabayashi T. Modulation transfer function evaluation of cone beam computed tomography for dental use with the oversampling method. Dentomaxillofac Radiol. 2010;39:28-32. Xu J, Reh DD, Carey JP, et al. Technical assessment of a conebeam CT scanner for otolaryngology imaging: image quality, dose, and technique protocols. Med Phys. 2012;39:4932-4942. Ozaki Y, Watanabe H, Nomura Y, et al. Location dependency of the spatial resolution of cone beam computed tomography for dental use. Oral Surg Oral Med Oral Pathol Oral Radiol. 2013;116:648-655. Liedke GS, da Silveira HE, da Silveira HL, et al. Influence of voxel size in the diagnostic ability of cone beam tomography to evaluate simulated external root resorption. J Endod. 2009;35: 233-235. Kamburoglu K, Kursun S. A comparison of the diagnostic accuracy of CBCT images of different voxel resolutions used to detect simulated small internal resorption cavities. Int Endod J. 2010;43:798-807. Wenzel A, Haiter-Neto F, Frydenberg M, Kirkevang LL. Variable-resolution cone-beam computerized tomography with enhancement filtration compared with intraoral photostimulable phosphor radiography in detection of transverse root fractures in an in vitro model. Oral Surg Oral Med Oral Pathol Oral Radiol. 2009;108:939-945. Choi DH, Choi BR, Choi JW, et al. Reference line-pair values of panoramic radiographs using an arch-form phantom stand to assess clinical image quality. Imaging Sci Dent. 2013;43:7-15. Suomalainen A, Kiljunen T, Käser Y, et al. Dosimetry and image quality of four dental cone beam computed tomography scanners compared with multislice computed tomography scanners. Dentomaxillofac Radiol. 2009;38:367-378. Pauwels R, Silkosessak O, Jacobs R, et al. A pragmatic approach to determine the optimal kVp in cone beam CT: balancing contrast-to-noise ratio and radiation dose. Dentomaxillofac Radiol. 2014;43:20140059. Pauwels R, Seynaeve L, Bosmans H, et al. Leuven/BE: technical versus diagnostic image quality in dental CBCT imaging. Available at: http://posterng.netkey.at/esr/viewing/index.php? module¼viewing_poster&task¼&pi¼118196. Accessed December 23, 2014.
Reprint requests: Sam-Sun Lee Department of Oral and Maxillofacial Radiology School of Dentistry, Seoul National University Chang-kyeong-gung-ro 62-1 Jongno-gu Seoul 110-768 Korea [email protected]
