International Congress Series 1281 (2005) 1177 – 1181
www.ics-elsevier.com
Image quality analysis and low dose dental CT S.D. Bianchia,T, O. Rampadob, L. Lubertoa, A.F. Genovesioa, C.C. Bianchia, R. Ropolob a
Radiodiagnostic Section, Medical–Surgical Sciences Department, University of Turin, Italy b Medical Physics Department, bSan Giovanni BattistaQ Hospital of Turin, Italy
Abstract. Due to the rapid increase of the number of computed tomography examinations for dental purposes we are facing the problem of dose reduction without loss of image quality. The purpose of this work is to investigate the image quality variation in relation to scan parameters that affect dose. Subjective evaluations and quantitative measurements of image quality indices were performed on images obtained from geometrical and anatomical phantoms. Obtained results agree on the possibility of a significant dose reduction maintaining a diagnostic image quality. D 2005 CARS & Elsevier B.V. All rights reserved. Keywords: Low dose dental CT; Dental CT; Dose reduction; Image quality
1. Introduction The use of computed tomography (CT) for dental purposes is rapidly increasing. CT equipments specifically developed for dentomaxillofacial imaging [1] (low dose cone beam CT) as well as total body CT equipments with dedicated software for dental imaging [2] are commercially available. The image acquisition with these last equipments is normally performed using a standard protocol. The choice of the parameter values that define the acquisition protocols should be done considering the balance between the image quality required and the radiation dose to patient. Radiologists and clinicians may tend to consider the former point as predominant, but the higher radiation dose to patient correlated with a better image quality is only justified if there is an increase in diagnostic information [3].
T Corresponding author. E-mail address:
[email protected] (S.D. Bianchi). 0531-5131/ D 2005 CARS & Elsevier B.V. All rights reserved. doi:10.1016/j.ics.2005.03.133
1178
S.D. Bianchi et al. / International Congress Series 1281 (2005) 1177–1181
In dental CT for preoperative planning the structures of interest are high contrast objects. Some recent studies [4–8] indicate that a consistent dose reduction without significant loss of image quality is possible. The image quality is often assessed by a subjective comparative analysis performed by several radiologists, using a rank scale to judge differentiation of picture quality. In the present work we use both subjective evaluation and quantitative measurements of image quality indices in order to investigate the variation of the image quality in relation to scan parameters that affect dose. 2. Material and methods CT scans were performed with a multislice CT (Lightspeed 16Pro, GE, Milwakee, WI, USA). Test images were obtained on two different phantoms: a geometric phantom (Catphan 500 phantom, The Phantom Laboratory, Salem, NY, USA) and an anatomical phantom (mandibular tract inserted in a cylindrical water phantom). The Catphan 500 is a cylindrical plastic phantom composed of four modules with several object details. One module contains high contrast details (a bead and a bar pattern) for power resolution evaluations. Many acquisition protocols were used with different parameters (tube currents ranged from 10 to 160 mA, rotation time 0.7 s, pitch factors ranged from 0.562 to 1.75, 120 kV, slice thickness 0.625 mm, FOV 128 mm). Cross section images of the anatomical phantom were reformatted by the software DentaScan (GE, Milwakee, WI, USA). Quantitative evaluation of image quality concerned spatial resolution, noise and contrast noise ratio. The spatial resolution was evaluated on the geometrical phantom images by means of automated quality assurance software provided by IRIS-Inc. (Frederick, MD, USA). Noise and contrast noise ratio of high contrast details were evaluated on both phantoms by means of self developed java plugins of the software ImageJ provided by NIH. An automatic region of interest (ROI) selection procedure was implemented to perform these quantitative measurements on the anatomical phantom. An example of ROI definition on a cross section is shown in Fig. 1(h), with a ROI 1 inside the cortical bone tract and a ROI 2 in the adjacent soft tissue. The mean and standard deviation values of these ROIs were evaluated for many cortical tracts and for every acquisition modality. We considered the standard deviation in the ROI 2 as a measure of noise and the ratio between the difference of the mean values in the two ROI and the noise as contrast noise ratio. Comparative analysis of the pictures obtained with 28 different modalities (7 mA values and 4 pitch values) was performed independently by six observers (1 radiologist 8 years of experience in dental CT, 5 residents in radiology and in maxillofacial surgery) blinded to each other. The rank scale used was the following: 1 as insufficient, 2 as poor, 3 as sufficient, 4 as discrete, and 5 as good. Three evaluation criteria were adopted: the representation quality of the spongy bone, the mandibular canal, and alveolar crest edge. The results of the picture analysis were established with the Friedman’s non parametric test ( P b 0.05) and multiple rank pairs comparisons. Prior to the comparative study interobserver reproducibility among the six observers was evaluated on 20 dental CT studies randomly selected. Kappa statistics showed a substantial agreement (kappa = 0.82).
S.D. Bianchi et al. / International Congress Series 1281 (2005) 1177–1181
1179
Fig. 1. Cross sections obtained with 10 mA (a), 20 mA (b), 40 mA (c), 60 mA (d), 80 mA (e), 120 mA (f), and 160 mA (g). In (h) an example of automatic ROI definition is shown.
The computed tomography dose index (CTDI) for every exposure condition was provided by the equipment user interface. The accuracy of these provided values was verified by comparison with measured values, finding a maximum difference of 5% during the routinely quality controls. 3. Results Measurements performed on the geometrical phantom on axial sections resulted in modulation transfer function (MTF) values at 50%, 10%, and 2% of 4.56, 7.15, and 8.44 lp/cm, with 120 mA. Similar MTF values were obtained with lower mA. Noise values ranged from 93.0 HU at 10 mA to 18.8 HU at 120 mA. The contrast noise ratio ranged from 10.2 to 40.6. Fig. 2 shows contrast noise ratio values and noise values obtained from the automatic ROI analysis performed on the cross sections of the anatomical phantom. Noise values
Fig. 2. Noise and contrast noise ratio values vs mA and for different pitches, obtained from the automatic ROI analysis performed on the anatomical phantom.
1180
S.D. Bianchi et al. / International Congress Series 1281 (2005) 1177–1181
Table 1 Average of independent image quality evaluation of six radiologists for images acquired with different mA, with a pitch of 0.562 Diagnostic criteria and CTDI
10 mA
20 mA
40 mA
60 mA
80 mA
120 mA
160 mA
Spongy bone Mandibular canal Alveolar ridge CTDI (mGy)
1.7 2.7 2.8 2.6
2.5 3.2 3.3 5.2
3.8 4.3 4.5 10.5
3.5 4.0 4.3 15.7
4.0 4.8 4.8 20.9
3.8 4.2 4.7 31.3
3.2 4.2 4.5 41.8
ranged from 238.3 HU (10 mA, pitch 1.75) to 41.8 HU (160 mA, pitch 0.562). The contrast noise ratio ranged from 7.7 to 43.2. The application of Friedman’s test to the subjective image quality evaluation data set highlighted that there was no significant difference in image quality between images acquired at 40 mA and images acquired at higher mA, for all the three diagnostic criteria chosen and for all the pitch values. The multiple comparison for different pitches at 40 mA evidenced a difference only between pitch 0.562 and pitches higher than 0.938 for the visibility of the spongy bone. There was no significant differences between images acquired with different pitches for the other two criteria. Examples of averages of independent subjective image quality evaluation are shown in Table 1 (different mA and pitch 0.562) and in Table 2 (different pitches and 40 mA). 4. Discussion The evaluation of image quality in dental CT especially performed for implant purposes has been done up until now mainly by means of qualitative methods [5–8]. For these reasons the study started with an attempt to approach a quantitative method, with automated software analysis of the axial images of the Catphan geometrical phantom. This phantom has high contrast targets for the resolution power measure, that should be performed in low noise conditions. As a consequence, this phantom lacks of elements and analysis procedure suitable to assess the best compromise between image quality and dose. Indeed MTF values at 50%, 10%, and 2% obtained with different exposure conditions did not show a clear dependence on mA, as differences were not due to actual correlations between these quantities but to errors in the measurement method related to high noise and low contrast noise ratio also measured. Even if these last indices show a clear dependence on mA and pitches, it is difficult to define a proper threshold of their values related to the necessary image quality. Table 2 Average of independent image quality evaluation of six radiologists for images acquired with different pitches, with a mA value of 40 Diagnostic criteria and CTDI
Pitch 0.562
Pitch 0.938
Pitch 1.375
Pitch 1.75
Spongy bone Mandibular canal Alveolar ridge CTDI (mGy)
3.8 4.3 4.5 10.5
3.2 3.7 4.2 6.3
2.7 3.7 3.7 4.3
2.7 3.7 4.0 3.4
S.D. Bianchi et al. / International Congress Series 1281 (2005) 1177–1181
1181
For this reason, a simple anatomical phantom was made in order to better investigate the role of this quality indices, in conditions more similar to the actual clinical use. Results of both subjective evaluation and quantitative measurements of image quality indices were consistent with each other. The main increment of noise was observed in the images obtained with less than 40 mA, which also received a significantly lower score in the subjective evaluation. In terms of contrast noise ratio (CNR) between cortical bone and soft tissue, a threshold of about 20 could be considered a reasonable compromise between dose reduction and good representation of anatomical details for diagnostic use, as shown by the approximately uniform scores assigned by all the observers to images with CNR values above this threshold. Different pitch values only affected the score of the spongy bone visibility. CTDI differences between a standard protocol (120 mA, pitch 0.5) and a low dose protocol (40 mA, pitch 1.75) indicate the possibility of about a 90% dose reduction. Higher dose reduction might be obtained by an increase of pitch value and or a much lower mA, when the sheer visibility of the mandibular canal and the identification of the upper level of the alveolar crest edge are the primary aim of the study. Further investigations could be done in order to compare results obtained on this equipment with low dose cone beam CT, also using other anatomical phantoms with different structural characteristics regarding spongy bone and mandibular canal. Acknowledgements This study was partially supported by grants from bRicerca Sanitaria Finalizzata 2004Regione PiemonteQ and 60% research funds 2004, University of Turin. References [1] P. Mozzo, et al., A new volumetric CT machine for dental imaging based on the cone-beam technique: preliminary results, Eur. Radiol. 8 (1998) 1558 – 1564. [2] J.J. Abrahams, Anatomy of the jaw revisited with a dental CT software program, Am. J. Neuroradiol. 14 (1993) 979 – 990. [3] Annals of the ICRP, 1990 Recommendations of the International Commission on Radiological Protection, 60, publication 60, vol. 21 (1990). [4] M. Cohnen, et al., CT of the head by use of reduced current and kilovoltage: relationship between image quality and dose reduction, Am. J. Neuroradiol. 21 (2000) 1654 – 1660. [5] P. Rustemeyer, U. Streubuhr, J. Suttmoeller, Low-dose dental computed tomography: significant dose reduction without loss of image quality, Acta Radiol. 45 (2004) 847 – 853. [6] P. Rustemeyer, et al., Low-dosage dental CT, RoFo 171 (1999) 130 – 135. [7] A. Ekestubbe, K. Grondahl, H.G. Grondahl, Quality of preimplant low-dose tomography, Oral Surg., Oral Med., Oral Pathol., Oral Radiol. Endod. 88 (1999) 738 – 744. [8] C. Schorn, et al., Dental CT: image quality and radiation exposure in relation to scan parameters, RoFo 170 (1999) 137 – 144.
