- Email: [email protected]
Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy
Clinical Radiology xxx (2016) 1e7
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy K. Higashigaito a, A.S. Becker a, K. Sprengel b, H.-P. Simmen b, G. Wanner b, H. Alkadhi a, * a b
Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, University of Zurich, Switzerland Division of Trauma Surgery, University Hospital Zurich, University of Zurich, Switzerland
art icl e i nformat ion Article history: Received 11 September 2015 Received in revised form 15 April 2016 Accepted 29 April 2016
AIM: To demonstrate the feasibility and accuracy of automatic radiation dose monitoring software for computed tomography (CT) of trauma patients in a clinical setting over time, and to evaluate the potential of radiation dose reduction using iterative reconstruction (IR). MATERIALS AND METHODS: In a time period of 18 months, data from 378 consecutive thoraco-abdominal CT examinations of trauma patients were extracted using automatic radiation dose monitoring software, and patients were split into three cohorts: cohort 1, 64-section CT with filtered back projection, 200 mAs tube currentetime product; cohort 2, 128-section CT with IR and identical imaging protocol; cohort 3, 128-section CT with IR, 150 mAs tube current etime product. Radiation dose parameters from the software were compared with the individual patient protocols. Image noise was measured and image quality was semi-quantitatively determined. RESULTS: Automatic extraction of radiation dose metrics was feasible and accurate in all (100%) patients. All CT examinations were of diagnostic quality. There were no differences between cohorts 1 and 2 regarding volume CT dose index (CTDIvol; p¼0.62), doseelength product (DLP), and effective dose (ED, both p¼0.95), while noise was significantly lower (chest and abdomen, both 38%, p<0.017). Compared to cohort 1, CTDIvol, DLP, and ED in cohort 3 were significantly lower (all 25%, p<0.017), similar to the noise in the chest (e32%) and abdomen (e27%, both p<0.017). Compared to cohort 2, CTDIvol (e28%), DLP, and ED (both e26%) in cohort 3 was significantly lower (all, p<0.017), while noise in the chest (þ9%) and abdomen (þ18%) was significantly higher (all, p<0.017). CONCLUSION: Automatic radiation dose monitoring software is feasible and accurate, and can be implemented in a clinical setting for evaluating the effects of lowering radiation doses of CT protocols over time. Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction * Guarantor and correspondent: H. Alkadhi, Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, University of Zurich, Raemistrasse 100, CH-8091 Zurich, Switzerland. Tel.: þ41 44 255 3662; fax: þ41 44 255 4443. E-mail address: [email protected] (H. Alkadhi).
Computed tomography (CT) today is characterized by its robustness, fast data-acquisition speed, high spatial and temporal resolution, high diagnostic accuracy, and wide
http://dx.doi.org/10.1016/j.crad.2016.04.023 0009-9260/Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
2
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
availability, and is, therefore, the technique of choice for trauma patients. The drawback of CT, however, is the associated radiation dose.1e3 Radiation dose delivered that does not benefit the patient should be avoided, with the aim of lowering the risk of stochastic effects that could increase the individual’s lifetime baseline risk of cancer.4 Thus, techniques have been developed that enable the radiation dose to be reduced to a level that is “as low as reasonably achievable” (ALARA),5 including reduction of the number of scan phases, limiting the scan range in the z-axis, low kilovoltage CT, and application of iterative reconstruction (IR) algorithms. IR primarily improves image quality through a reduction of image noise.6 This, in turn, can be translated into a reduction of radiation dose while maintaining image quality of conventional filtered back projection (FBP) reconstructions.7 Despite the many options for optimizing the radiation dose of CT, it is also important to continuously monitor the correct usage of these techniques to ensure optimal radiation dose for each individual CT examination. To demonstrate the ALARA principle in patients, there needs to be repeated aggregation and analysis of dose metric data according to the protocol,4 which represents the cornerstone of quality assurance in CT.8 Recently, software tools became available that automatically collect and monitor radiation dose metrics from imaging methods employing ionizing radiation. Such software tools have the advantage of eradicating the need for potentially erroneous and time-consuming manual entry9 by automatically drawing exposure information directly from the picture archiving and communication system (PACS) or directly from the imaging system. Moreover, such automatic software allows for collection of a larger number of patients than by manual typing, resulting in more generalizable results regarding radiation doses. The purpose of the present study was to demonstrate the feasibility and accuracy of automatic radiation dose monitoring software in trauma patients undergoing CT with different protocols over time, and for evaluating the potential of radiation dose reduction using IR.
monitoring program, which automatically collects and analyses dosimetric data of all CT examinations performed. DoseWatch includes various embedded analysis features, such as dose per protocol displays or dose comparison between different study protocols (Fig 1). Protocols were filtered by the software based on the protocol names, which remained the same over time despite replacement of the CT system as described below. The program’s export feature was used to export the data collected into an Excel spreadsheet. This Excel file contained all important patient-related information and dosimetric data, but did not contain any information about the clinical indication, image quality, and imaging findings of an examination. As a first step, the criterion “study’s protocol name” was used to identify all CT studies in trauma patients (n¼752). The next step was to identify all CT studies performed in trauma patients, defined as two or more severe injuries, with at least one injury or the sum of all injuries being life threatening.10 As a standard at University Hospital Zurich, all patients with polytrauma are scanned with arms positioned on top of a pillow placed ventrally on the body with both arms flexed at the elbow, as shown previously.11 In the second step, CT images of all exported examinations were reviewed and those patients who did not suffer from polytrauma (but rather a minor trauma) and/or who were scanned with a different arm position were excluded (n¼373; Fig 2). Another reason for exclusion was if noise measurement (see below) was not feasible due to the lack of fat tissue in a thin patient (n¼1). Thus, 378 CT examinations of the chest and abdomen of trauma patients were included (68% male, mean age 5021 years; age range 16e99 years) in the present study. To determine the accuracy of the automatic software, all radiation dose parameters in the same 378 patients were also manually extracted and noted from the electronically logged patient protocols of each examination by another reader.
Material and methods
The 378 CT examinations were split into three different cohorts (Table 1) depending on the CT system and protocol used (Fig 3). CT examinations between March 2013 and February 2014 (n¼188; 66.5% male, mean age 48.720.7 years; age range 16e92 years) were assigned to cohort 1. All examinations in this cohort were performed with a 64section dual-source CT system (Somatom Definition; Siemens Healthcare, Forchheim, Germany) in the singlesource mode, using protocol A: 120 kVp tube voltage, 200 mAs reference tube currentetime product per rotation using automatic attenuation-based tube current modulation (CareDose 4D; Siemens). Images were reconstructed with a section thickness of 2 mm, an increment of 1.6 mm, and using FBP. CT examinations between March 2014 and May 2014 (n¼92; 70.7% male, mean age 50.620.8 years; age range 16e99 years) were assigned to cohort 2. All examinations of cohort 2 were performed using a 128-section dual-source
The study was conducted at a Level 1 trauma centre after institutional review board was obtained. A Level 1 trauma centre requires the availability of 24/7 trauma patient care including radiology services, as well as standardized and structured data acquisition to a registry, which records the quality of processes and results. Written informed consent was waived by the local ethics committee because of the retrospective nature of the study.
Study population and radiation dose information Within a time period of 18 months (between March 2013 and August 2014) a total of 752 CT examinations were identified with automatic radiation dose monitoring software (DoseWatch, GE Healthcare, Waukesha, WI, USA). This software is a commercially available web-based dose-
CT protocols
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
3
Figure 1 (a) User interface of the automatic dose monitoring software showing the DLP of the two different study protocols (study protocol A versus study protocol B). Note the gap in dose monitoring between February and March 2014, representing the time of CT scanner change. (b) Example of the software interface illustrating the CTDIvol and DLP analysis of each acquisition with protocol B.
CT system (Somatom Definition Flash; Siemens Healthcare) operated in the single-source mode. All examinations were performed with identical parameters to cohort 1 (protocol
A). Images were also reconstructed with a section thickness of 2 mm (increment 1.6 mm), but using sinogram-affirmed IR (SAFIRE, Siemens) at a strength level of 3. CT examinations between June 2014 and August 2014 (n¼98; 68.4% male, mean age 51.820.3 years; age range 17e94 years) were assigned to cohort 3. All examinations of cohort 3 were performed on the same 128-section dualsource CT system as in cohort 2, but with a reduced tube
Table 1 Patient demographics.
Age (meanSD) in years Sex (% male) ISS (medianIQR) NISS (medianIQR) Blunt trauma Penetrating trauma
Figure 2 Flowchart of the study.
Cohort 1 (n¼188)
Cohort 2 (n¼92)
Cohort 3 (n¼98)
p-Values
48.720.7
50.620.8
51.820.3
0.53
66.5 21.117.9 2826.5 168/188 (89%) 20/188 (11%)
70.7 22.419 28.124.2 70/92 (76%) 12/92 (13%)
68.4 22.318.4 29.226.1 89/98 (91%) 9/98 (9%)
0.25 0.32 0.37 0.11 0.14
ISS, injury severity score; NISS, new injury severity score; SD, standard deviation; IQR, interquartile range.
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
4
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
Objective image quality
Figure 3 CT systems and protocols of each cohort.
currentetime product of 150 mAs per rotation, also using automatic attenuation-based tube current modulation (CareDose 4D; Siemens). Images were similarly reconstructed with a section thickness of 2 mm, a reconstruction increment of 1.6 mm; and using SAFIRE at a strength level of 3. The reduction in tube current was based on the known amount of image noise reduction with SAFIRE.12 In all three cohorts an identical bolus of 100 ml isoosmolar, non-ionic iodinated contrast material (300 mg iodine/ml, iopromide; Ultravist, Bayer Healthcare, Leverkusen, Germany) followed by a saline flush of 30 ml was injected into an antecubital vein at a flow rate of 3 ml/s. All reconstructed images were archived in the hospital’s PACS (Impax 6.0, Agfa HealthCare, Mortsel, Belgium) for documentary purposes and further image analysis.
Radiation dose Dose metrics, such as volume CT dose index (CTDIvol) and doseelength product (DLP) were automatically collected by the software and analysed in the exported Excel spreadsheet. The same radiation dose metrics, CTDIvol and DLP, were taken from the electronically logged scan protocols from each patient. Effective dose (ED) was calculated by multiplying the DLP with an organ-specific factor.13 As only thoracic (kthorax¼0.014 mSv/mGy$cm) and abdominal (kabdomen¼0.015 mSv/mGy$cm) conversion coefficients exist (but not thoraco-abdominal coefficients), the mean of both region-specific conversion coefficients was used: kmean¼0.0145 mSv/mGy$cm to calculate the ED, as previously shown.13
Subjective image quality First, all final reports issued by the resident and the attending radiologist (having more than 5 years of experience in reading CT images) on call were reviewed regarding the diagnostic quality of the CT examination: specifically, the presence of a reduced diagnostic confidence due to deterioration of image quality was investigated. Second, all CT images were reviewed independently by another two radiologists (2 and 3 years of experience in radiology) regarding the quality of the examination in a dichotomic way: diagnostic or non-diagnostic. Reasons for nondiagnostic image quality were noted.
For each examination, noise measurements were performed in the chest and abdomen by one of the readers who performed the subjective image-quality analysis. For this, attenuation and the standard deviation (SD) of attenuation were measured using a circular region of interest (ROI, average diameter 10 mm). This ROI was set in the subcutaneous fat of the anterior right chest wall at the level of the mammilla and in the perirenal fat of the right kidney in each CT examination. The SD of attenuation was taken as a measure of image noise. If noise measurement in the right thoracic wall was not possible, measurements were conducted in the fat of the right axilla.
Statistical analysis Descriptive data are given as mean values and standard deviations. Normality was tested using the ShapiroeWilk test. For multiple comparisons, KruskaleWallis one-way analysis of variance was used to test for differences regarding CTDIvol, DLP, ED, and image noise among all three cohorts, and the ManneWhitney U-test was used to test for differences regarding CTDIvol, DLP, ED, and image noise between each two cohorts. The Bonferroni correction was used to control the family-wise error rate for testing three hypotheses on the same data set. Therefore, a two-sided pvalue of <0.017 (0.05/3) was considered statistically significant. Cohen’s kappa coefficient was used to measure the interobserver agreement between the two readers who performed the subjective image-quality analysis. Statistical analyses were performed using SPSS Software (SPSS, version 22, Chicago, IL, USA).
Results Patient demographics There were no significant differences in age and gender among the three patient cohorts (all p>0.05). Similarly, there were no significant differences in injury severity score (ISS) score, new injury severity score (NISS), and trauma mechanism between cohorts (p>0.05, Table 1).
Radiation dose metrics Comparison between the radiation dose metrics (i.e., CTDIvol and DLP) that were retrieved manually from the electronically logged patient protocols in each patient and those that were automatically extracted by the software were identical, indicating that the software tool was 100% accurate. The mean CTDIvol was 19.75.8 mGy (range 8.2e38.1 mGy) for cohort 1, 20.57 mGy (range 4.4e61.5 mGy) for cohort 2, and 14.83 mGy (range 9.1e14.8 mGy) for cohort 3, being significantly different among groups (p<0.017). There was no significant difference between cohort 1 and 2 in CTDIvol (p¼0.62), whereas there was a significant decrease from cohort 1 to 3 (mean CTDIvol difference 24.9 %,
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
p<0.017) and from cohort 2 to 3 (mean CTDIvol difference 27.8%, p<0.017; Table 2). Similar to the CTDIvol, there were significant differences in DLP and ED between cohorts (both p<0.017), with significantly lower DLP and ED in cohort 3 compared to cohort 1 (mean DLP and ED difference 25.2 %, p<0.017) and between cohort 2 and cohort 3 (mean DLP and ED difference 26.4 %, p<0.017). As with CTDIvol, there were no significant differences in DLP and ED between cohort 1 and cohort 2 (p¼0.95; Table 2).
Subjective image quality All examinations were regarded as being fully diagnostic by the two radiologists. Cohen’s Kappa showed excellent agreement (k¼1) between the additional two radiologists, who also read all CT images and considered all examinations to be of diagnostic image quality.
Objective image quality In cohort 1, mean image noise in the chest was 112 HU (range 5e18 HU) and the mean noise in the abdomen was 122 HU (9e25 HU). In cohort 2, the mean noise in the chest was 71 HU (4e12 HU) and the mean noise in the abdomen was 71 HU (5e11 HU). In cohort 3, the mean noise in the chest was 71 HU (4e12 HU) and the mean noise in the abdomen was 91 HU (6e13 HU). There was a significant difference among cohorts (p<0.017). Noise was significantly lower in cohort 2 compared to cohort 1 (chest: 37.6%; abdomen: 38.3%, both p<0.017). Noise was also significantly lower in cohort 3 compared to cohort 1 (chest: 31.8%; abdomen: 27.4%, both p<0.017), and significantly lower in cohort 2 compared to cohort 3 (chest: 9.2; abdomen: 17.5 %, both p<0.017, Fig 4).
Discussion The present study illustrates how automatic radiation dose monitoring software can be used in a clinical setting for collecting and analysing radiation dose metrics of a large cohort of trauma patients over time, eventually allowing for comparisons between various CT systems and protocols. Comparison of the metrics obtained by the software with those retrieved manually from the individual patient Table 2 Mean values and standard deviations of CTDIvol, DLP, and ED of each cohort. Cohort 1 CTDIvol (mGy) DLP (mGy$cm) ED (mSv)
19.75.8 1456412 21.16
Cohort 2
Cohort 3
20.57
14.83
1481522
1090246
21.57.6
8.83.6
p-Value 0.62a;< 0.017b; <0.017c 0.95a;< 0.017b; <0.017c 0.95a;< 0.017b; <0.017c
CTDIvol, volume CT dose index, DLP, doseelength product, ED, effective dose. a Cohort 1 versus cohort 2. b cohort 1 versus cohort 3. c cohort 2 versus cohort 3.
5
protocols showed that the automatic approach was accurate in all patients included in the study. Studies reporting the radiation dose of CT in trauma patients are rare, with only a few manually retrieving radiation dose metrics and determining radiation doses.2,14,15 In most studies on trauma patients so far, radiation doses were only roughly estimated.16,17 This further highlights the need for tools such as automatic radiation dose monitoring software, which enable efficient and accurate collection of radiation dose parameters in a clinical setting. Such radiation dose information could also be transferred to a central database of the department or hospital, which could be used for quality assurance. Compared to manual dose tracking, which is time consuming and may be prone to typographical errors,9 automatic radiation dose monitoring tools overcome these limitations, allowing analysis of larger numbers of examinations relatively quickly. It is important to note that radiation dose information gathered by the software was fully accurate, which is mandatory before such tools are used in a clinical setting. Notwithstanding the accuracy of the software, however, there may be situations in which discrepancies between the actual doses related to a certain protocol and the automatically extracted metrics occur. This typically happens when additional series are added to a CT examination while the study was being undertaken and read,4 where the original protocol name remains in the header of the examination although changes to the protocol were made. Another issue that was encountered in the present study was the fact that different patient populations were scanned with the same CT protocol, i.e., patients with severe trauma and those with minor trauma only. Differences between those patients consist of different arm positions, according to the hospital’s trauma work-up algorithm, which leads to differences in radiation dose. When the protocol name is identical, such information can be only obtained when visually going through the entire image data and extracting those that are not related to the actual analysis (a total of 373 patients in the present study). Based on this experience, different CT protocol names were generated for patients with trauma and those with minor trauma, respectively, to enable differentiation of the radiation dose metrics in different populations. Translation of such automatic software tools into clinical routine can be done through small dedicated teams, consisting of technologists and radiologists, who regularly meet and analyse the data provided by the software and decide when actions need be taken. After adding subjective and objective image-quality assessments to the analysis, which could not be performed automatically by the software release used in the present study, the authors were able to demonstrate how the use of IR (comparison of cohort 1 versus 2) resulted in a significant improvement in image quality through a reduction of noise. Moreover, applying IR and reducing the tube current (cohort 1 versus 3) resulted in a significantly lower radiation dose, while image quality remained diagnostic. IR is a reconstruction algorithm that improves image quality by reducing noise without a corresponding
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
6
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
Figure 4 Boxplots showing the differences in noise and CTDIvol between the three different cohorts. The horizontal lines in the boxes correspond to the mean. The top and bottom lines of the boxes correspond to the first and third quartiles. The whiskers represent 1.5 the interquartile range (IQR). Circles represent outliers (1.5 IQR e3 IQR from the near edge of the box) and stars represent extreme outliers (>3 IQR from the near edge of the box).
increasing in radiation dose. The IR process is characterized by a loop where the forward projection of an image estimate is compared with the original data, which are then corrected. This step is repeated several times, and every step renders images less noisy, ultimately leading to an optimized final image.18 So far, only a few studies have shown the potential of IR in trauma patients for improving image quality of the cervical spine19 and abdomen.20 Similar to these studies, the present study showed (at a constant radiation dose because of identical CT parameters) an improvement in image quality via an approximately 38% reduction in noise in the chest and abdomen of trauma patients when using SAFIRE (at a strength level 3) as opposed to FBP. Only a few studies in trauma patients have shown how the improvement in image quality can be used for reducing radiation dose. Grupp et al.21 showed that the use of adaptive statistical IR resulted in a radiation dose reduction of 31%, and Maxfield et al.22 showed that the implementation of adaptive statistical IR led to a nearly 20% reduction in radiation dose, while image quality was not compromised in both studies. Comparable to these results, the present study showed how the improvement of image quality with IR can be translated into a reduction in tube current and hence, in a significantly decreased radiation dose by 26%. As expected, image noise increased as a consequence of tube current reduction, however, without rendering the image quality non-diagnostic. Most importantly, compared to the initial situation without the possibility of IR (cohort 1), the combination of reduced tube current and IR resulted in a
significant reduction of radiation dose by 25% at a diagnostic image quality. There are some limitations of the present study. First, the retrospective nature of this study has inherent shortcomings. Second, the full potential of dose reduction by applying a tailored tube current reduction was not assessed. Instead, the reduction of tube current was based on the known amount of noise reduction with SAFIRE from the literature.12 Third, it is known that the determination of image noise based on the SD of attenuation in a ROI may fall short in images reconstructed with IR techniques. Nevertheless, this approach was used by many previous studies6,7,12,19,20,22 and provides at least a rough estimate of the noise level of the respective examination. Fourth, there exist alternative organ-specific conversion factors than those used in this study, leading to different EDs.23 Finally, the lowest possible radiation dose for dedicated trauma findings, such as for example traumatic liver or spleen lacerations, was not determined. This is also true for the present criteria defining diagnostic image quality. Although image quality based on objective measures, such as noise, differed between cohorts, indicating that they were not equivalent, they were considered subjectively as being diagnostic for all readers. In conclusion, the present study illustrates that automatic radiation dose monitoring software has potential value in a clinical setting and is an accurate tool for determining the radiation dose received by patients in a large cohort over time. Protocol names must be standardized, so that the software automatically selects the correct data for
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
review. Given the sparse literature on radiation doses in trauma patients, automatic radiation dose monitoring software helps not only for the awareness of true radiation doses in these patients, but can be also used for tracking changes after implementing dose-lowering techniques. In the present study, the combination of IR and reduced tube current resulted in significantly lower radiation doses without deteriorating the diagnostic yield of the CT studies, both representing techniques that should be applied widely in modern trauma centres.
References 1. Salibi PN, Agarwal V, Panczykowski DM, et al. Lifetime attributable risk of cancer from CT among patients surviving severe traumatic brain injury. AJR Am J Roentgenol 2014;202(2):397e400. 2. Gordic S, Alkadhi H, Hodel S, et al. Whole-body CT-based imaging algorithm for multiple trauma patients: radiation dose and time to diagnosis. Br J Radiol 2015;88(1047):20140616. 3. Leung V, Sastry A, Woo TD, et al. Implementation of a split-bolus singlepass CT protocol at a UK major trauma centre to reduce excess radiation dose in trauma pan-CT. Clin Radiol 2015;70(10):1110e5. 4. MacGregor K, Li I, Dowdell T, et al. Identifying institutional diagnostic reference levels for CT with radiation dose index monitoring software. Radiology 2015:141520. 5. McCollough CH, Primak AN, Braun N, et al. Strategies for reducing radiation dose in CT. Radiol Clin N Am 2009;47(1):27e40. 6. Gordic S, Desbiolles L, Stolzmann P, et al. Advanced modelled iterative reconstruction for abdominal CT: qualitative and quantitative evaluation. Clin Radiol 2014;69(12):e497e504. 7. Hara AK, Paden RG, Silva AC, et al. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. AJR Am J Roentgenol 2009;193(3):764e71. 8. Talati RK, Dunkin J, Parikh S, et al. Current methods of monitoring radiation exposure from CT. J Am Coll Radiol 2013;10(9):702e7. 9. AlSuwaidi JS, Bayoumi M, Al Shibli N, et al. Utilisation of PACS to monitor patient CT doses. Radiat Prot Dosimetry 2011;147(1e2):321e4. 10. Tscherne H, Oestern HJ, Sturm JA. [Stress tolerance of patients with multiple injuries and its significance for operative care]. Langenbecks Arch Chir 1984;364:71e7.
7
11. Karlo C, Gnannt R, Frauenfelder T, et al. Whole-body CT in polytrauma patients: effect of arm positioning on thoracic and abdominal image quality. Emerg Radiol 2011;18(4):285e93. 12. Winklehner A, Karlo C, Puippe G, et al. Raw data-based iterative reconstruction in body CTA: evaluation of radiation dose saving potential. Eur Radiol 2011;21(12):2521e6. 13. Shrimpton PC, Hillier MC, Lewis MA, et al. National survey of doses from CT in the UK: 2003. Br J Radiol 2006;79(948):968e80. 14. Ptak T, Rhea JT, Novelline RA. Radiation dose is reduced with a singlepass whole-body multi-detector row CT trauma protocol compared with a conventional segmented method: initial experience. Radiology 2003;229(3):902e5. 15. Winslow JE, Hinshaw JW, Hughes MJ, et al. Quantitative assessment of diagnostic radiation doses in adult blunt trauma patients. Ann Emerg Med 2008;52(2):93e7. 16. Surendran A, Mori A, Varma DK, et al. Systematic review of the benefits and harms of whole-body computed tomography in the early management of multitrauma patients: are we getting the whole picture? J Trauma Acute Care Surg 2014;76(4):1122e30. 17. Asha S, Curtis KA, Grant N, et al. Comparison of radiation exposure of trauma patients from diagnostic radiology procedures before and after the introduction of a panscan protocol. Emerg Med Aus 2012;24(1):43e51. 18. Nuyts J, De Man B, Fessler JA, et al. Modelling the physics in the iterative reconstruction for transmission computed tomography. Phys Med Biol 2013;58(12):R63e96. € rner M, Hempel R, et al. Evaluation of a dedicated MDCT 19. Geyer LL, Ko protocol using iterative image reconstruction after cervical spine trauma. Clin Radiol 2013;68(7):e391e6. 20. Steenburg SD, Persohn S, Shen C, et al. Iterative reconstruction improves image quality and preserves diagnostic accuracy in the setting of blunt solid organ injuries. Emerg Radiol 2015;22(1):43e51. 21. Grupp U, Schafer ML, Meyer H, et al. Reducing radiation dose in emergency CT scans while maintaining equal image quality: just a promise or reality for severely injured patients? Emerg Med Int 2013;2013:984645. 22. Maxfield MW, Schuster KM, McGillicuddy EA, et al. Impact of adaptive statistical iterative reconstruction on radiation dose in evaluation of trauma patients. J Trauma Acute Care Surg 2012;73(6):1406e11. 23. Gosling O, Loader R, Venables P, et al. Are we underestimating the dose? A radiation dose study utilizing the 2007 ICRP tissue weighting factors and a cardiac specific scan volume. Clin Radiol 2010;65(12):1013e7.
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy K. Higashigaito a, A.S. Becker a, K. Sprengel b, H.-P. Simmen b, G. Wanner b, H. Alkadhi a, * a b
Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, University of Zurich, Switzerland Division of Trauma Surgery, University Hospital Zurich, University of Zurich, Switzerland
art icl e i nformat ion Article history: Received 11 September 2015 Received in revised form 15 April 2016 Accepted 29 April 2016
AIM: To demonstrate the feasibility and accuracy of automatic radiation dose monitoring software for computed tomography (CT) of trauma patients in a clinical setting over time, and to evaluate the potential of radiation dose reduction using iterative reconstruction (IR). MATERIALS AND METHODS: In a time period of 18 months, data from 378 consecutive thoraco-abdominal CT examinations of trauma patients were extracted using automatic radiation dose monitoring software, and patients were split into three cohorts: cohort 1, 64-section CT with filtered back projection, 200 mAs tube currentetime product; cohort 2, 128-section CT with IR and identical imaging protocol; cohort 3, 128-section CT with IR, 150 mAs tube current etime product. Radiation dose parameters from the software were compared with the individual patient protocols. Image noise was measured and image quality was semi-quantitatively determined. RESULTS: Automatic extraction of radiation dose metrics was feasible and accurate in all (100%) patients. All CT examinations were of diagnostic quality. There were no differences between cohorts 1 and 2 regarding volume CT dose index (CTDIvol; p¼0.62), doseelength product (DLP), and effective dose (ED, both p¼0.95), while noise was significantly lower (chest and abdomen, both 38%, p<0.017). Compared to cohort 1, CTDIvol, DLP, and ED in cohort 3 were significantly lower (all 25%, p<0.017), similar to the noise in the chest (e32%) and abdomen (e27%, both p<0.017). Compared to cohort 2, CTDIvol (e28%), DLP, and ED (both e26%) in cohort 3 was significantly lower (all, p<0.017), while noise in the chest (þ9%) and abdomen (þ18%) was significantly higher (all, p<0.017). CONCLUSION: Automatic radiation dose monitoring software is feasible and accurate, and can be implemented in a clinical setting for evaluating the effects of lowering radiation doses of CT protocols over time. Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction * Guarantor and correspondent: H. Alkadhi, Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, University of Zurich, Raemistrasse 100, CH-8091 Zurich, Switzerland. Tel.: þ41 44 255 3662; fax: þ41 44 255 4443. E-mail address: [email protected] (H. Alkadhi).
Computed tomography (CT) today is characterized by its robustness, fast data-acquisition speed, high spatial and temporal resolution, high diagnostic accuracy, and wide
http://dx.doi.org/10.1016/j.crad.2016.04.023 0009-9260/Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
2
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
availability, and is, therefore, the technique of choice for trauma patients. The drawback of CT, however, is the associated radiation dose.1e3 Radiation dose delivered that does not benefit the patient should be avoided, with the aim of lowering the risk of stochastic effects that could increase the individual’s lifetime baseline risk of cancer.4 Thus, techniques have been developed that enable the radiation dose to be reduced to a level that is “as low as reasonably achievable” (ALARA),5 including reduction of the number of scan phases, limiting the scan range in the z-axis, low kilovoltage CT, and application of iterative reconstruction (IR) algorithms. IR primarily improves image quality through a reduction of image noise.6 This, in turn, can be translated into a reduction of radiation dose while maintaining image quality of conventional filtered back projection (FBP) reconstructions.7 Despite the many options for optimizing the radiation dose of CT, it is also important to continuously monitor the correct usage of these techniques to ensure optimal radiation dose for each individual CT examination. To demonstrate the ALARA principle in patients, there needs to be repeated aggregation and analysis of dose metric data according to the protocol,4 which represents the cornerstone of quality assurance in CT.8 Recently, software tools became available that automatically collect and monitor radiation dose metrics from imaging methods employing ionizing radiation. Such software tools have the advantage of eradicating the need for potentially erroneous and time-consuming manual entry9 by automatically drawing exposure information directly from the picture archiving and communication system (PACS) or directly from the imaging system. Moreover, such automatic software allows for collection of a larger number of patients than by manual typing, resulting in more generalizable results regarding radiation doses. The purpose of the present study was to demonstrate the feasibility and accuracy of automatic radiation dose monitoring software in trauma patients undergoing CT with different protocols over time, and for evaluating the potential of radiation dose reduction using IR.
monitoring program, which automatically collects and analyses dosimetric data of all CT examinations performed. DoseWatch includes various embedded analysis features, such as dose per protocol displays or dose comparison between different study protocols (Fig 1). Protocols were filtered by the software based on the protocol names, which remained the same over time despite replacement of the CT system as described below. The program’s export feature was used to export the data collected into an Excel spreadsheet. This Excel file contained all important patient-related information and dosimetric data, but did not contain any information about the clinical indication, image quality, and imaging findings of an examination. As a first step, the criterion “study’s protocol name” was used to identify all CT studies in trauma patients (n¼752). The next step was to identify all CT studies performed in trauma patients, defined as two or more severe injuries, with at least one injury or the sum of all injuries being life threatening.10 As a standard at University Hospital Zurich, all patients with polytrauma are scanned with arms positioned on top of a pillow placed ventrally on the body with both arms flexed at the elbow, as shown previously.11 In the second step, CT images of all exported examinations were reviewed and those patients who did not suffer from polytrauma (but rather a minor trauma) and/or who were scanned with a different arm position were excluded (n¼373; Fig 2). Another reason for exclusion was if noise measurement (see below) was not feasible due to the lack of fat tissue in a thin patient (n¼1). Thus, 378 CT examinations of the chest and abdomen of trauma patients were included (68% male, mean age 5021 years; age range 16e99 years) in the present study. To determine the accuracy of the automatic software, all radiation dose parameters in the same 378 patients were also manually extracted and noted from the electronically logged patient protocols of each examination by another reader.
Material and methods
The 378 CT examinations were split into three different cohorts (Table 1) depending on the CT system and protocol used (Fig 3). CT examinations between March 2013 and February 2014 (n¼188; 66.5% male, mean age 48.720.7 years; age range 16e92 years) were assigned to cohort 1. All examinations in this cohort were performed with a 64section dual-source CT system (Somatom Definition; Siemens Healthcare, Forchheim, Germany) in the singlesource mode, using protocol A: 120 kVp tube voltage, 200 mAs reference tube currentetime product per rotation using automatic attenuation-based tube current modulation (CareDose 4D; Siemens). Images were reconstructed with a section thickness of 2 mm, an increment of 1.6 mm, and using FBP. CT examinations between March 2014 and May 2014 (n¼92; 70.7% male, mean age 50.620.8 years; age range 16e99 years) were assigned to cohort 2. All examinations of cohort 2 were performed using a 128-section dual-source
The study was conducted at a Level 1 trauma centre after institutional review board was obtained. A Level 1 trauma centre requires the availability of 24/7 trauma patient care including radiology services, as well as standardized and structured data acquisition to a registry, which records the quality of processes and results. Written informed consent was waived by the local ethics committee because of the retrospective nature of the study.
Study population and radiation dose information Within a time period of 18 months (between March 2013 and August 2014) a total of 752 CT examinations were identified with automatic radiation dose monitoring software (DoseWatch, GE Healthcare, Waukesha, WI, USA). This software is a commercially available web-based dose-
CT protocols
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
3
Figure 1 (a) User interface of the automatic dose monitoring software showing the DLP of the two different study protocols (study protocol A versus study protocol B). Note the gap in dose monitoring between February and March 2014, representing the time of CT scanner change. (b) Example of the software interface illustrating the CTDIvol and DLP analysis of each acquisition with protocol B.
CT system (Somatom Definition Flash; Siemens Healthcare) operated in the single-source mode. All examinations were performed with identical parameters to cohort 1 (protocol
A). Images were also reconstructed with a section thickness of 2 mm (increment 1.6 mm), but using sinogram-affirmed IR (SAFIRE, Siemens) at a strength level of 3. CT examinations between June 2014 and August 2014 (n¼98; 68.4% male, mean age 51.820.3 years; age range 17e94 years) were assigned to cohort 3. All examinations of cohort 3 were performed on the same 128-section dualsource CT system as in cohort 2, but with a reduced tube
Table 1 Patient demographics.
Age (meanSD) in years Sex (% male) ISS (medianIQR) NISS (medianIQR) Blunt trauma Penetrating trauma
Figure 2 Flowchart of the study.
Cohort 1 (n¼188)
Cohort 2 (n¼92)
Cohort 3 (n¼98)
p-Values
48.720.7
50.620.8
51.820.3
0.53
66.5 21.117.9 2826.5 168/188 (89%) 20/188 (11%)
70.7 22.419 28.124.2 70/92 (76%) 12/92 (13%)
68.4 22.318.4 29.226.1 89/98 (91%) 9/98 (9%)
0.25 0.32 0.37 0.11 0.14
ISS, injury severity score; NISS, new injury severity score; SD, standard deviation; IQR, interquartile range.
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
4
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
Objective image quality
Figure 3 CT systems and protocols of each cohort.
currentetime product of 150 mAs per rotation, also using automatic attenuation-based tube current modulation (CareDose 4D; Siemens). Images were similarly reconstructed with a section thickness of 2 mm, a reconstruction increment of 1.6 mm; and using SAFIRE at a strength level of 3. The reduction in tube current was based on the known amount of image noise reduction with SAFIRE.12 In all three cohorts an identical bolus of 100 ml isoosmolar, non-ionic iodinated contrast material (300 mg iodine/ml, iopromide; Ultravist, Bayer Healthcare, Leverkusen, Germany) followed by a saline flush of 30 ml was injected into an antecubital vein at a flow rate of 3 ml/s. All reconstructed images were archived in the hospital’s PACS (Impax 6.0, Agfa HealthCare, Mortsel, Belgium) for documentary purposes and further image analysis.
Radiation dose Dose metrics, such as volume CT dose index (CTDIvol) and doseelength product (DLP) were automatically collected by the software and analysed in the exported Excel spreadsheet. The same radiation dose metrics, CTDIvol and DLP, were taken from the electronically logged scan protocols from each patient. Effective dose (ED) was calculated by multiplying the DLP with an organ-specific factor.13 As only thoracic (kthorax¼0.014 mSv/mGy$cm) and abdominal (kabdomen¼0.015 mSv/mGy$cm) conversion coefficients exist (but not thoraco-abdominal coefficients), the mean of both region-specific conversion coefficients was used: kmean¼0.0145 mSv/mGy$cm to calculate the ED, as previously shown.13
Subjective image quality First, all final reports issued by the resident and the attending radiologist (having more than 5 years of experience in reading CT images) on call were reviewed regarding the diagnostic quality of the CT examination: specifically, the presence of a reduced diagnostic confidence due to deterioration of image quality was investigated. Second, all CT images were reviewed independently by another two radiologists (2 and 3 years of experience in radiology) regarding the quality of the examination in a dichotomic way: diagnostic or non-diagnostic. Reasons for nondiagnostic image quality were noted.
For each examination, noise measurements were performed in the chest and abdomen by one of the readers who performed the subjective image-quality analysis. For this, attenuation and the standard deviation (SD) of attenuation were measured using a circular region of interest (ROI, average diameter 10 mm). This ROI was set in the subcutaneous fat of the anterior right chest wall at the level of the mammilla and in the perirenal fat of the right kidney in each CT examination. The SD of attenuation was taken as a measure of image noise. If noise measurement in the right thoracic wall was not possible, measurements were conducted in the fat of the right axilla.
Statistical analysis Descriptive data are given as mean values and standard deviations. Normality was tested using the ShapiroeWilk test. For multiple comparisons, KruskaleWallis one-way analysis of variance was used to test for differences regarding CTDIvol, DLP, ED, and image noise among all three cohorts, and the ManneWhitney U-test was used to test for differences regarding CTDIvol, DLP, ED, and image noise between each two cohorts. The Bonferroni correction was used to control the family-wise error rate for testing three hypotheses on the same data set. Therefore, a two-sided pvalue of <0.017 (0.05/3) was considered statistically significant. Cohen’s kappa coefficient was used to measure the interobserver agreement between the two readers who performed the subjective image-quality analysis. Statistical analyses were performed using SPSS Software (SPSS, version 22, Chicago, IL, USA).
Results Patient demographics There were no significant differences in age and gender among the three patient cohorts (all p>0.05). Similarly, there were no significant differences in injury severity score (ISS) score, new injury severity score (NISS), and trauma mechanism between cohorts (p>0.05, Table 1).
Radiation dose metrics Comparison between the radiation dose metrics (i.e., CTDIvol and DLP) that were retrieved manually from the electronically logged patient protocols in each patient and those that were automatically extracted by the software were identical, indicating that the software tool was 100% accurate. The mean CTDIvol was 19.75.8 mGy (range 8.2e38.1 mGy) for cohort 1, 20.57 mGy (range 4.4e61.5 mGy) for cohort 2, and 14.83 mGy (range 9.1e14.8 mGy) for cohort 3, being significantly different among groups (p<0.017). There was no significant difference between cohort 1 and 2 in CTDIvol (p¼0.62), whereas there was a significant decrease from cohort 1 to 3 (mean CTDIvol difference 24.9 %,
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
p<0.017) and from cohort 2 to 3 (mean CTDIvol difference 27.8%, p<0.017; Table 2). Similar to the CTDIvol, there were significant differences in DLP and ED between cohorts (both p<0.017), with significantly lower DLP and ED in cohort 3 compared to cohort 1 (mean DLP and ED difference 25.2 %, p<0.017) and between cohort 2 and cohort 3 (mean DLP and ED difference 26.4 %, p<0.017). As with CTDIvol, there were no significant differences in DLP and ED between cohort 1 and cohort 2 (p¼0.95; Table 2).
Subjective image quality All examinations were regarded as being fully diagnostic by the two radiologists. Cohen’s Kappa showed excellent agreement (k¼1) between the additional two radiologists, who also read all CT images and considered all examinations to be of diagnostic image quality.
Objective image quality In cohort 1, mean image noise in the chest was 112 HU (range 5e18 HU) and the mean noise in the abdomen was 122 HU (9e25 HU). In cohort 2, the mean noise in the chest was 71 HU (4e12 HU) and the mean noise in the abdomen was 71 HU (5e11 HU). In cohort 3, the mean noise in the chest was 71 HU (4e12 HU) and the mean noise in the abdomen was 91 HU (6e13 HU). There was a significant difference among cohorts (p<0.017). Noise was significantly lower in cohort 2 compared to cohort 1 (chest: 37.6%; abdomen: 38.3%, both p<0.017). Noise was also significantly lower in cohort 3 compared to cohort 1 (chest: 31.8%; abdomen: 27.4%, both p<0.017), and significantly lower in cohort 2 compared to cohort 3 (chest: 9.2; abdomen: 17.5 %, both p<0.017, Fig 4).
Discussion The present study illustrates how automatic radiation dose monitoring software can be used in a clinical setting for collecting and analysing radiation dose metrics of a large cohort of trauma patients over time, eventually allowing for comparisons between various CT systems and protocols. Comparison of the metrics obtained by the software with those retrieved manually from the individual patient Table 2 Mean values and standard deviations of CTDIvol, DLP, and ED of each cohort. Cohort 1 CTDIvol (mGy) DLP (mGy$cm) ED (mSv)
19.75.8 1456412 21.16
Cohort 2
Cohort 3
20.57
14.83
1481522
1090246
21.57.6
8.83.6
p-Value 0.62a;< 0.017b; <0.017c 0.95a;< 0.017b; <0.017c 0.95a;< 0.017b; <0.017c
CTDIvol, volume CT dose index, DLP, doseelength product, ED, effective dose. a Cohort 1 versus cohort 2. b cohort 1 versus cohort 3. c cohort 2 versus cohort 3.
5
protocols showed that the automatic approach was accurate in all patients included in the study. Studies reporting the radiation dose of CT in trauma patients are rare, with only a few manually retrieving radiation dose metrics and determining radiation doses.2,14,15 In most studies on trauma patients so far, radiation doses were only roughly estimated.16,17 This further highlights the need for tools such as automatic radiation dose monitoring software, which enable efficient and accurate collection of radiation dose parameters in a clinical setting. Such radiation dose information could also be transferred to a central database of the department or hospital, which could be used for quality assurance. Compared to manual dose tracking, which is time consuming and may be prone to typographical errors,9 automatic radiation dose monitoring tools overcome these limitations, allowing analysis of larger numbers of examinations relatively quickly. It is important to note that radiation dose information gathered by the software was fully accurate, which is mandatory before such tools are used in a clinical setting. Notwithstanding the accuracy of the software, however, there may be situations in which discrepancies between the actual doses related to a certain protocol and the automatically extracted metrics occur. This typically happens when additional series are added to a CT examination while the study was being undertaken and read,4 where the original protocol name remains in the header of the examination although changes to the protocol were made. Another issue that was encountered in the present study was the fact that different patient populations were scanned with the same CT protocol, i.e., patients with severe trauma and those with minor trauma only. Differences between those patients consist of different arm positions, according to the hospital’s trauma work-up algorithm, which leads to differences in radiation dose. When the protocol name is identical, such information can be only obtained when visually going through the entire image data and extracting those that are not related to the actual analysis (a total of 373 patients in the present study). Based on this experience, different CT protocol names were generated for patients with trauma and those with minor trauma, respectively, to enable differentiation of the radiation dose metrics in different populations. Translation of such automatic software tools into clinical routine can be done through small dedicated teams, consisting of technologists and radiologists, who regularly meet and analyse the data provided by the software and decide when actions need be taken. After adding subjective and objective image-quality assessments to the analysis, which could not be performed automatically by the software release used in the present study, the authors were able to demonstrate how the use of IR (comparison of cohort 1 versus 2) resulted in a significant improvement in image quality through a reduction of noise. Moreover, applying IR and reducing the tube current (cohort 1 versus 3) resulted in a significantly lower radiation dose, while image quality remained diagnostic. IR is a reconstruction algorithm that improves image quality by reducing noise without a corresponding
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
6
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
Figure 4 Boxplots showing the differences in noise and CTDIvol between the three different cohorts. The horizontal lines in the boxes correspond to the mean. The top and bottom lines of the boxes correspond to the first and third quartiles. The whiskers represent 1.5 the interquartile range (IQR). Circles represent outliers (1.5 IQR e3 IQR from the near edge of the box) and stars represent extreme outliers (>3 IQR from the near edge of the box).
increasing in radiation dose. The IR process is characterized by a loop where the forward projection of an image estimate is compared with the original data, which are then corrected. This step is repeated several times, and every step renders images less noisy, ultimately leading to an optimized final image.18 So far, only a few studies have shown the potential of IR in trauma patients for improving image quality of the cervical spine19 and abdomen.20 Similar to these studies, the present study showed (at a constant radiation dose because of identical CT parameters) an improvement in image quality via an approximately 38% reduction in noise in the chest and abdomen of trauma patients when using SAFIRE (at a strength level 3) as opposed to FBP. Only a few studies in trauma patients have shown how the improvement in image quality can be used for reducing radiation dose. Grupp et al.21 showed that the use of adaptive statistical IR resulted in a radiation dose reduction of 31%, and Maxfield et al.22 showed that the implementation of adaptive statistical IR led to a nearly 20% reduction in radiation dose, while image quality was not compromised in both studies. Comparable to these results, the present study showed how the improvement of image quality with IR can be translated into a reduction in tube current and hence, in a significantly decreased radiation dose by 26%. As expected, image noise increased as a consequence of tube current reduction, however, without rendering the image quality non-diagnostic. Most importantly, compared to the initial situation without the possibility of IR (cohort 1), the combination of reduced tube current and IR resulted in a
significant reduction of radiation dose by 25% at a diagnostic image quality. There are some limitations of the present study. First, the retrospective nature of this study has inherent shortcomings. Second, the full potential of dose reduction by applying a tailored tube current reduction was not assessed. Instead, the reduction of tube current was based on the known amount of noise reduction with SAFIRE from the literature.12 Third, it is known that the determination of image noise based on the SD of attenuation in a ROI may fall short in images reconstructed with IR techniques. Nevertheless, this approach was used by many previous studies6,7,12,19,20,22 and provides at least a rough estimate of the noise level of the respective examination. Fourth, there exist alternative organ-specific conversion factors than those used in this study, leading to different EDs.23 Finally, the lowest possible radiation dose for dedicated trauma findings, such as for example traumatic liver or spleen lacerations, was not determined. This is also true for the present criteria defining diagnostic image quality. Although image quality based on objective measures, such as noise, differed between cohorts, indicating that they were not equivalent, they were considered subjectively as being diagnostic for all readers. In conclusion, the present study illustrates that automatic radiation dose monitoring software has potential value in a clinical setting and is an accurate tool for determining the radiation dose received by patients in a large cohort over time. Protocol names must be standardized, so that the software automatically selects the correct data for
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023
K. Higashigaito et al. / Clinical Radiology xxx (2016) 1e7
review. Given the sparse literature on radiation doses in trauma patients, automatic radiation dose monitoring software helps not only for the awareness of true radiation doses in these patients, but can be also used for tracking changes after implementing dose-lowering techniques. In the present study, the combination of IR and reduced tube current resulted in significantly lower radiation doses without deteriorating the diagnostic yield of the CT studies, both representing techniques that should be applied widely in modern trauma centres.
References 1. Salibi PN, Agarwal V, Panczykowski DM, et al. Lifetime attributable risk of cancer from CT among patients surviving severe traumatic brain injury. AJR Am J Roentgenol 2014;202(2):397e400. 2. Gordic S, Alkadhi H, Hodel S, et al. Whole-body CT-based imaging algorithm for multiple trauma patients: radiation dose and time to diagnosis. Br J Radiol 2015;88(1047):20140616. 3. Leung V, Sastry A, Woo TD, et al. Implementation of a split-bolus singlepass CT protocol at a UK major trauma centre to reduce excess radiation dose in trauma pan-CT. Clin Radiol 2015;70(10):1110e5. 4. MacGregor K, Li I, Dowdell T, et al. Identifying institutional diagnostic reference levels for CT with radiation dose index monitoring software. Radiology 2015:141520. 5. McCollough CH, Primak AN, Braun N, et al. Strategies for reducing radiation dose in CT. Radiol Clin N Am 2009;47(1):27e40. 6. Gordic S, Desbiolles L, Stolzmann P, et al. Advanced modelled iterative reconstruction for abdominal CT: qualitative and quantitative evaluation. Clin Radiol 2014;69(12):e497e504. 7. Hara AK, Paden RG, Silva AC, et al. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. AJR Am J Roentgenol 2009;193(3):764e71. 8. Talati RK, Dunkin J, Parikh S, et al. Current methods of monitoring radiation exposure from CT. J Am Coll Radiol 2013;10(9):702e7. 9. AlSuwaidi JS, Bayoumi M, Al Shibli N, et al. Utilisation of PACS to monitor patient CT doses. Radiat Prot Dosimetry 2011;147(1e2):321e4. 10. Tscherne H, Oestern HJ, Sturm JA. [Stress tolerance of patients with multiple injuries and its significance for operative care]. Langenbecks Arch Chir 1984;364:71e7.
7
11. Karlo C, Gnannt R, Frauenfelder T, et al. Whole-body CT in polytrauma patients: effect of arm positioning on thoracic and abdominal image quality. Emerg Radiol 2011;18(4):285e93. 12. Winklehner A, Karlo C, Puippe G, et al. Raw data-based iterative reconstruction in body CTA: evaluation of radiation dose saving potential. Eur Radiol 2011;21(12):2521e6. 13. Shrimpton PC, Hillier MC, Lewis MA, et al. National survey of doses from CT in the UK: 2003. Br J Radiol 2006;79(948):968e80. 14. Ptak T, Rhea JT, Novelline RA. Radiation dose is reduced with a singlepass whole-body multi-detector row CT trauma protocol compared with a conventional segmented method: initial experience. Radiology 2003;229(3):902e5. 15. Winslow JE, Hinshaw JW, Hughes MJ, et al. Quantitative assessment of diagnostic radiation doses in adult blunt trauma patients. Ann Emerg Med 2008;52(2):93e7. 16. Surendran A, Mori A, Varma DK, et al. Systematic review of the benefits and harms of whole-body computed tomography in the early management of multitrauma patients: are we getting the whole picture? J Trauma Acute Care Surg 2014;76(4):1122e30. 17. Asha S, Curtis KA, Grant N, et al. Comparison of radiation exposure of trauma patients from diagnostic radiology procedures before and after the introduction of a panscan protocol. Emerg Med Aus 2012;24(1):43e51. 18. Nuyts J, De Man B, Fessler JA, et al. Modelling the physics in the iterative reconstruction for transmission computed tomography. Phys Med Biol 2013;58(12):R63e96. € rner M, Hempel R, et al. Evaluation of a dedicated MDCT 19. Geyer LL, Ko protocol using iterative image reconstruction after cervical spine trauma. Clin Radiol 2013;68(7):e391e6. 20. Steenburg SD, Persohn S, Shen C, et al. Iterative reconstruction improves image quality and preserves diagnostic accuracy in the setting of blunt solid organ injuries. Emerg Radiol 2015;22(1):43e51. 21. Grupp U, Schafer ML, Meyer H, et al. Reducing radiation dose in emergency CT scans while maintaining equal image quality: just a promise or reality for severely injured patients? Emerg Med Int 2013;2013:984645. 22. Maxfield MW, Schuster KM, McGillicuddy EA, et al. Impact of adaptive statistical iterative reconstruction on radiation dose in evaluation of trauma patients. J Trauma Acute Care Surg 2012;73(6):1406e11. 23. Gosling O, Loader R, Venables P, et al. Are we underestimating the dose? A radiation dose study utilizing the 2007 ICRP tissue weighting factors and a cardiac specific scan volume. Clin Radiol 2010;65(12):1013e7.
Please cite this article in press as: Higashigaito K, et al., Automatic radiation dose monitoring for CT of trauma patients with different protocols: feasibility and accuracy, Clinical Radiology (2016), http://dx.doi.org/10.1016/j.crad.2016.04.023