Dose reduction in paediatric cranial CT via iterative reconstruction: a clinical study in 78 patients

Clinical Radiology 71 (2016) 1168e1177

Contents lists available at ScienceDirect

Clinical Radiology journal homepage: www.clinicalradiologyonline.net

Dose reduction in paediatric cranial CT via iterative reconstruction: a clinical study in 78 patients € ning a, D.M. Renz c, D. Kaul a, b, *, y, J. Kahn a, y, L. Huizing a, E. Wiener a, G. Bo F. Streitparth a Department of Radiology, Charit e School of Medicine and University Hospital, Charit eplatz 1, 10117 Berlin, Germany Department of Radiation Oncology, Charit e School of Medicine and University Hospital, Augustenburger Platz 1, 13353 Berlin, Germany c Department of Radiology, Jena University Hospital, Erlanger Allee 101, 07747 Jena, Germany a

b

article in formation Article history: Received 8 March 2016 Received in revised form 24 May 2016 Accepted 24 June 2016

AIM: To assess how adaptive statistical iterative reconstruction (ASIR) contributes to dose reduction and affects image quality of non-contrast cranial computed tomography (cCT) in children. MATERIALS AND METHODS: Non-contrast cranial CT acquired in 78 paediatric patients (age 0e12 years) were evaluated. The images were acquired and processed using four different protocols: Group A (control): 120 kV, filtered back projection (FBP), n¼18; Group B: 100 kV, FBP, n¼22; Group C: 100 kV, scan and reconstruction performed with 20% ASIR, n¼20; Group D1: 100 kV, scan and reconstruction performed with 30% ASIR, n¼18; Group D2: raw data from Group D1 reconstructed using a blending of 40% ASIR and 60% FBP, n¼18. The effective dose was calculated and the image quality was assessed quantitatively and qualitatively. RESULTS: Compared to Group A, Groups C and D1/D2 showed a significant reduction of the doseelength product (DLP) by 34.4% and 64.4%, respectively. All experimental groups also showed significantly reduced qualitative levels of noise, contrast, and overall diagnosability. Diagnosis-related confidence grading showed Group C to be adequate for everyday clinical practice. Quantitative measures of Groups B and C were comparable to Group A with only few parameters compromised. Quantitative scores in Groups D1 and D2 were mainly lower compared to Group A, with Group D2 performing better than Group D1. Group D2 was considered adequate for follow-up imaging of severe acute events such as bleeding or hydrocephalus. DISCUSSION: The use of ASIR combined with low tube voltage may reduce radiation significantly while maintaining adequate image quality in non-contrast paediatric cCT. Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.

Introduction * Guarantor and correspondent: D. Kaul, Department of Radiation  School of Medicine and University Hospital, Campus Oncology, Charite Virchow-Klinikum, Augustenburger Platz 1, 13353 Berlin, Germany. Tel.: þ49 30 450 527152. E-mail address: [email protected] (D. Kaul). y These authors contributed equally.

During the last decades, computed tomography (CT) scanners have become widely available in the first world, leading to an overall higher total dose of ionising radiation

http://dx.doi.org/10.1016/j.crad.2016.06.115 0009-9260/Ó 2016 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

in the population.1 Due to the carcinogenic potential of ionising radiation, the increased use of CT is particularly concerning when it comes to paediatric patients because of higher radiosensitivity and because these patients have a longer life expectancy, leading to a higher probability of radiation-induced carcinogenesis.2,3 Strategies to reduce radiation dose include using ultrasound (US) or magnetic resonance imaging (MRI) instead of CT when possible, as well as dose-saving CT technologies, such as automated tube current modulation and noise reduction filters4,5; however, the potential of these dose reduction technologies is limited when examining through dense skull bone. In addition, lowering the tube potential in the acquisition of cranial CT reduces radiation effectively, but comes at the cost of increased image noise.6 With advances in computer technology, iterative reconstruction (IR) algorithms, which were first introduced for single-photon-emission CT (SPECT) and positron-emission tomography (PET), have been identified as a possible tool for dose reduction in CT. Subsequently, all major CT manufactures have introduced IR algorithms for their CT units. IR algorithms have the potential to eliminate the increased image noise resulting from the use of lower tube currents. Pilot studies in adults have shown that IR algorithms may reduce the radiation dose of cranial CT by 20e45%.7e10 This clinical study analysed the effect of IR on radiation dose, image quality, and interpretability in comparison with routine filtered back projection (FBP) CT of the head in a large paediatric patient population.

Materials and methods Study design The institutional ethics board approved this study. As patients were not exposed to additional radiation and their data were stored anonymously, the informed consent requirement was waived. Five protocols d A, B, C, D1 and D2 d with increasing dose reduction potential were used. Paediatric patients 0e12 years of age were included. The selected patients underwent cranial CT for one of the following acute events: trauma, loss of consciousness, seizure, or focal neurological deficit.

CT protocol A summary of the protocols can be found in Table 1. All examinations were performed on a 64-slice multidetector Table 1 CT protocol characteristics.

Tube potential Noise Index ASIR Blending ratio

Group A

Group B

Group C

Group D1 Group D2

120 kV 2.8 0% 100% FBP 0% ASIR

100 kV 2.8 0% 100% FBP 0% ASIR

100 kV 4 20% 80% FBP 20% ASIR

100 kV 6 30% 70% FBP 30% ASIR

FBP, filtered back reconstruction.

projection;

ASIR,

adaptive

100 kV 6 30% 60% FBP 40% ASIR

statistical

iterative

1169

CT system (Lightspeed VCT, GE Healthcare, Milwaukee, WI, USA). Using tube current modulation, patients were scanned at 120 kV (Group A) or 100 kV (Groups B, C, D1, and D2) with a tube current range of 100e300 mA. Each scan was carried out in a craniocaudal direction. In Groups A and B, FBP was used for image reconstruction, whereas in Groups C and D1/2 adaptive statistical IR (ASIR) was applied and blended with FBP. A control group of 18 patients was scanned using CT Protocol A (120 kV, FBP, NI: 2.8¼reference NI). Twentytwo patients were scanned using CT Protocol B (100 kV, FBP, NI: 2.8). Twenty patients were scanned using CT Protocol C (100 kV, 20% ASIR, NI: 4). Eighteen patients were scanned using CT Protocol D1 (100 kV, 30% ASIR, NI: 6). By default, the use of 30% ASIR results in a tube current reduction of approximately 30%; the raw data were analysed using the FBP and the ASIR algorithms, resulting in blended images of 30% ASIR and 70% FBP. In Group D2, raw data from Group D1 were blended using 40% ASIR and 60% FBP.

Data reconstruction ASIR is an algorithm with an emphasis on noise reduction for the reconstruction of CT images using information obtained from the FBP algorithm as its basis for further image modelling. In ASIR, the values of each pixel (y) are transformed using matrix algebra to obtain a new estimate of the pixel (y’). This value is then compared with the ideal value predicted by the noise model and iterative steps are carried out until the final and ideal pixel values eventually converge.11 In GE CT systems, the tool routinely utilised in the user interface to set desired image quality is called the noise index (NI). Referenced to the standard deviation in radiodensity (HU) in a specific-sized water phantom, the NI is compared to the attenuation measured in the preliminary CT scout. A reduction in the NI results in lower noise, but requires a higher tube current. The application of ASIR introduces an alternative option for adjusting tube current settings. The operator can choose the ASIR level in 10% steps from 0% to 50%. The percentage of ASIR by default correlates directly to the percentage reduction in tube current during the scan. It is, however, also possible to select values for NI and ASIR, which mutually exclude each other, e.g., very low NI and high ASIR levels. In such a case, NI values are prioritised over ASIR. As a result, when a low NI is set the reduction in tube current via ASIR is restricted. Likewise, when the NI is high and the level of ASIR is low, tube current reduction may exceed expectation. Following the scan sequence, the raw data are reconstructed using ASIR and FBP alternatively and the images are then combined in a ratio of x% ASIR and 100ex% FBP. For example, when using 20% ASIR, tube current is reduced by 20%, and the final images are reconstructed with a 20% ASIR and 80% FBP blending. Varying blending ratios can, however, be applied after image acquisition, as performed in Group D2.

1170

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

Image quality Image quality was evaluated using both qualitative and quantitative measurements. In measuring image quality quantitatively, signal attenuation (SI), measured in Hounsfield units (HU), and noise were assessed by selecting the following regions of interest (ROIs): in the lentiform nucleus (ROI1), frontal white matter (WM; ROI2), frontal cortical layer (ROI3), ventricle (ROI4), internal capsule (ROI5), cortical layer of cerebellum (ROI6), WM of the middle cerebellar peduncle (ROI7) and vermis (ROI8; Fig 1). The following equation was used to calculate the signalto-noise ratio (SNR): SNR ¼

SIROIa SDROIa

(1)

The contrast-to-noise ratio (CNR) was measured in the supratentorial (ST) region between ROI3/ROI2 (ST d CNR C/WM) and between ROI1/ROI2 (ST d CNR LN/WM). In the infratentorial (IT) region, CNR was measured using ROI6/ ROI7 (IT d CNR C/WM) and ROI8/ROI7 (IT d CNR V/WM). CNR was calculated using the following equation:

DðSIROIa ; SIROIb Þ CNR ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðSDROIa Þ2 þ ðSDROIb Þ2

(2)

In order to evaluate the images qualitatively, two qualified radiologists with 5 and 11 years of clinical experience analysed the images following a joint training session. Using a blind assessment, all technical information was removed from the images to reduce expectation bias. Image quality was then evaluated in seven categories: noise, ST contrast between the cortex and WM, ST contrast between lentiform nucleus and internal capsule, IT contrast between cortex and WM, overall diagnosability and confidence (in patients with diagnosed acute pathology). A five-point Likert scale

was used to evaluate each of the seven categories, in which the reference was an “ideal examination”. The following Likert scale format was implemented: 1: non-diagnostic image quality, 2: uncertainty concerning the evaluation, 3: restricted assessment, 4: unrestricted diagnostic image evaluation possible, 5: excellent image quality.

Radiation dose The doseelength product (DLPs) and CT dose index (CTDIvol) were measured.

Statistical analysis The data were analysed using GraphPad Prism version 5.0f for Mac (GraphPad Software, San Diego, CA, USA) and IBM SPSS Statistics 19 (New York, NY, USA). Continuous data were analysed using Student’s t-test, while the chi-square test was used for categorical variables. p0.05 was considered statistically significant and p0.1 was considered a trend. Interobserver agreement between the two readers was assessed using Cohen’s kappa statistic.

Results Patient characteristics Patient characteristics are summarised in Table 2. Patients in Groups B and C were significantly younger than in the control group; however, there was no significant difference between the four groups regarding cranial diameter or scan range. The groups were also well balanced in terms of male-to-female ratio. Thirty-one patients (39.7%) were referred after experiencing a head injury, 13 patients (16.7%) were referred due to seizures, 12 patients (15.4%) were referred

Figure 1 Sites of ROIs for quantitative image analysis. Supratentorial ROIs included the lentiform nucleus (ROI1), frontal white matter (ROI2), frontal cortical layer (ROI3), ventricle (ROI4) and internal capsule (ROI5). Infratentorial ROIs included the cortical layer of cerebellum (ROI6), WM of middle cerebellar peduncle (ROI7) and vermis (ROI8).

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

1171

Table 2 Patient characteristics. Overall

N Age (years) Male:female ratio AP diameter (cm) Transverse diameter (cm) No pathology Epi-/subdural haemorrhage Fracture Hydrocephalus Cerebral haemorrhage Subacute stroke Tumour SAH Other

78 4.83.6 48:31 16.61.2 13.60.95 29 (37.2%) 17 (21.8%) 13 (16.7%) 9 (11.5%) 5 (6.4%) 4 (5.1%) 2 (2.6%) 1 (1.3%) 4 (5.1)

Group A

Group B

A vs. B

Group C

A vs. C

Group D1/D2

A vs. D1/D2

120 kV/FBP

100 kV/ASIR20

p-Value

100 kV/ASIR20

p-Value

100 kV/ASIR30

p-Value

18 7.73.4 11:7 16.71.0 13.70.86 9 (50%) 1 (5.6%) 2 (11.1%) 1 (5.6%) 0 (0%) 1 (5.6%) 0 (0%) 1 (5.6%) 3 (16.7%)

22 3.23 16:6 16.21.58 13.41.1 8 (36.4%) 7 (31.8%) 2 (9.1%) 1 (4.5%) 1 (4.5%) 3 (13.6%) 1 (4.5%) 0 (0%) 1 (4.5%)

<0.0001 0.44 0.25 0.2

20 2.82 11:10 16.81.1 13.70.89 8 (40%) 4 (20%) 5 (25%) 4 (20%) 2 (10%) 0 (0%) 1 (5%) 0 (0%) 0 (0%)

<0.0001 0.7 0.81 0.98

18 6.33.4 10:8 16.61.1 13.7 0.86 4 (22.2%) 5 (27.8%) 4 (22.2%) 3 (16.7%) 2 (11.1%) 0 (0%) 0 (0%) 0 (0%) 0 (0%)

0.2 0.74 0.69 0.9

There was no significant difference between the four groups regarding gender or cranial diameter. Age was lower in Groups B and C compared to the control group. The high number of patients with no intracranial pathology underlines the necessity to keep the level of ionising radiation as low as reasonably possible. FBP, filtered back projection; ASIR, adaptive statistical iterative reconstruction; AP, anteroposterior; SAH, subarachnoid haemorrhage.

following a cranial operational procedure, six patients (7.7%) were referred due to a reduced level of consciousness, four patients (5.1%) were referred as a result of an extracranial malignoma, three patients (3.8%) were referred due to a cranial malformation, and nine patients (11.5%) were referred for other reasons. Of the 78 patients referred for cranial CT, 62.8% showed cranial pathologies. The most common diagnoses were epi- and subdural bleeding (21.8%), fracture (16.7%), or hydrocephalus (11.5%; Table 2).

Quantitative analysis of image quality Table 3 summarises the results of quantitative analysis of image quality. Compared to Group A (control), Group B showed similar ST SNRs; IT SNRs showed mixed results with a higher value for the SNR of the cerebellar cortex and lower values for peduncle and vermis. ST CNRs in Group B were similar to the control group; regarding the IT CNRs vermis/WM values were increased in Group B, while cortex/WM contrast corresponded to the control group. In Group C, 20% ASIR was combined with a reduced tube voltage of 100 kV. Here, the ST SNRs in the frontal WM and frontal cortical layer decreased significantly, values of IT SNRs also decreased significantly when looking at peduncle and vermis, and the other ST and IT SNRs remained similar to the control group. CNRs in Group C were comparable to those of the control group with the exception of the vermis/ WM CNR, which was significantly increased in comparison to the control group. In Group D1, the level of ASIR was further increased to 30%. As a result, all ST and IT SNRs decreased significantly. With regard to the CNRs, the supraventricular lentiform nucleus/WM CNR as well as the IT cortex/WM CNR decreased significantly. When ASIR blending was increased to 40% (Group D2), SNRs showed higher levels than in Group D1 (30% ASIR

blending), but remained significantly lower compared to the control group. Most CNRs showed higher levels than in Group D1, except for the lentiform nucleus/WM CNR.

Qualitative analysis of image quality Table 4 and Fig 2 present the results of qualitative analysis of image quality and interobserver agreement. Compared to control Group A, image quality in terms of overall diagnosability, noise, and ST cortical-, basal ganglia and IT contrast were significantly reduced in all experimental groups. Mean values in Group B were found to be slightly higher than 3.5, and mean values in Group C were mainly slightly lower than 3.5. In Group D1, quality levels of overall diagnosability, noise, ST cortical contrast, basal ganglia contrast and IT contrast fell below values of 3. The increased blending ratio (40% ASIR, 60% FBP) in Group D2 led to a slight improvement of these subjective quality values in comparison to Group D1. Diagnosis-related confidence was not compromised in Groups B and C compared to the control group; however, there was a trend (p¼0.1) towards decreased diagnosisrelated confidence in Groups D1 and D2.

Radiation dose Data on radiation doses are summarised in Table 5 and Fig 5. Using a tube voltage of 100 kV and 0% ASIR for CT in Group B led to a non-significant reduction of the DLP by 23.2% compared to Group A. Combining a tube voltage of 100 kV and 20% ASIR (Group C) reduced the DLP significantly by 34.4%. The combination of a tube voltage of 100 kV and 30% ASIR led to a significant DLP reduction of 64.4%.

Discussion The number of CT examinations performed worldwide has been increasing constantly and there is growing

1172

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

Table 3 Quantitative analysis of image quality.

SNR ROI1 SNR ROI2 SNR ROI3 SNR ROI4 SNR ROI5 SNR ROI6 SNR ROI7 SNR ROI8 ST-CNR C/WM ST-CNR LN/WM IT-CNR C/WM IT-CNR V/WM

Group A

Group B

A vs. B

Group C

A vs. C

Group D1

A vs. D1

Group D2

A vs. D2

120 kV/FBP

100 kV/FBP

p-Value

100 kV/ASIR20

p-Value

100 kV/ASIR30

p-Value

100 kV/ASIR30 (40%/60%)

p-Value

71.4 6.71.6 8.92.2 1.10.33 5.81.1 7.51.2 6.51.6 9.11.9 1.70.69 0.990.34 1.60.47 0.980.29

7.22.2 5.91.4 8.33.3 1.350.54 5.91.1 8.71.6 5.51.1 7.91.7 1.691.35 0.90.34 1.40.47 1.640.79

0.75 0.12 0.49 0.073 0.83 0.013 0.015 0.039 0.98 0.38 0.2 0.002

6.51.3 4.60.97 6.81.7 1.20.36 5.11.3 7.81.8 4.81.2 71.1 1.650.59 1.130.56 1.480.36 1.670.47

0.26 <0.0001 0.017 0.26 0.11 0.65 0.0005 0.0002 0.8 0.39 0.36 <0.0001

4.20.77 41.1 5.61.3 0.670.36 3.40.62 5.40.93 3.70.56 5.91.4 1.410.5 0.720.28 1.240.49 0.920.39

<0.0001 <0.0001 <0.0001 0.0011 <0.0001 <0.0001 <0.0001 <0.0001 0.15 0.013 0.032 0.6

4.40.96 4.11.1 6.62.2 0.790.36 3.70.82 5.91.3 4.71.4 7.11.6 1.570.67 0.670.29 1.550.53 1.020.39

<0.0001 <0.0001 0.0032 0.015 <0.0001 0.0004 0.0008 0.0015 0.56 0.0044 0.77 0.76

Compared to Group A Group B showed similar supratentorial SNRs; infratentorial SNRs showed a higher value for the SNR of the cerebellar cortex and lower values for peduncle and vermis regions. Infratentorial CNRs vermis/white matter values were increased in Group B. In Group C, supratentorial SNRs in the frontal white matter and frontal cortical layer decreased significantly, while values of infratentorial SNRs also decreased significantly. CNRs in Group C were comparable to the control group with the exception of the vermis/white matter CNR, which was significantly increased. In Group D1, SNRs decreased significantly. Looking at the CNRs, the supraventricular lentiform nucleus/white matter CNR as well as the infratentorial cortex/ white matter CNR decreased significantly. Group D2 SNRs showed higher levels than Group D1 but remained significantly decreased compared to the control group. Most CNRs also showed higher levels than in Group D1, except for the lentiform nucleus/white matter CNR. FBP, filtered back projection; ASIR, adaptive statistical iterative reconstruction; SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio; ST-CNR C/WM, supratentorial CNR (cortex/white matter); ST-CNR NL/WM, supratentorial CNR (lentiform nucleus/white matter); IT-CNR C/WM, infratentorial CNR (cortex/ white matter); IT-CNR V/WM, infratentorial CNR (vermis/white matter). Lentiform nucleus (ROI1), frontal white matter (ROI2), frontal cortical layer (ROI3), ventricle (ROI4), internal capsule (ROI5), infratentorial ROIs included the cortical layer of cerebellum (ROI6), WM of middle cerebellar peduncle (ROI7), and vermis (ROI8).

discussion on its potential risks.12 In the present study, approximately one third of the paediatric patients referred for cranial CT did not show any radiological pathology. Considering the harmful potential of ionising radiation, cranial CT examinations should consequently be performed with the lowest possible radiation dose that still allows for adequate diagnosability, especially when it comes to paediatric patients; however, it should be kept in mind that even with the most sophisticated dose-reduction algorithms the most important method to reduce dose in any patient is the physician who chooses the correct diagnostic method (e.g., MRI for the examination of hydrocephalus). The wide availability of IR algorithms, in the context of CT particularly, may play a major role in reaching these aims. It was shown that IR algorithms have the potential to

significantly reduce dose while maintaining, or in some cases even improving, image quality.2,11,13e17 The results of the present study show that the use of a CT protocol with a tube voltage of 100 kV and 20% ASIR (Group C) reduced the DLP by 34.4% compared to the 120 kV/FBP control Group A. The combination of a tube voltage of 100 kV and 30% ASIR further led to a DLP reduction of 64.4%. All experimental groups also showed significantly reduced qualitative levels of noise, contrast, and overall diagnosability; however, diagnosis-related confidence grading was not significantly compromised in Groups BeD. There was a trend towards reduced diagnosis-related confidence in Groups D1 and D2. Quantitative measures of Groups B and C were, for the most part, comparable to Group A with only a few parameters compromised; quantitative scores in

Table 4 Qualitative analysis of image quality and interobserver agreement k.

Overall diagnosability Noise Supratentorial cortical contrast Basal ganglia contrast Infratentorial contrast Diagnosis-related Confidence

Group A

Group B

A vs. B

Group C

A vs. C

Group D1

A vs. D1

Group D2

120 kV/ FBP

100 kV/ FBP

p-Value

100 kV/ ASIR20

p-Value

100 kV/ ASIR30

p-Value

100 kV/ASIR30 p-Value (40%/60%)

A vs. D2

4.440.5 4.330.48 4.330.48 4.390.49 4.390.49 50

3.570.55 3.60.59 3.640.53 3.60.5 3.520.5 50

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 >0.99

3.50.55 3.450.5 3.530.51 3.550.5 3.430.5 4.970.17

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.48

2.970.55 2.680.48 2.970.17 2.820.39 2.910.29 4.860.36

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.1

30 30.25 3.090.29 30.25 2.970.17 4.860.36

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.1

Interobserver agreement k

0.67 0.51 0.62 0.65 0.71 0.88

Overall diagnosability, noise, supratentorial-, basal ganglia- and infratentorial contrast were significantly reduced in Group B, C and D1/2. Group D2 showed a slight improvement of these subjective quality values compared to Group D1. The diagnosis related confidence was not compromised in Group B and C compared to the control group. There was a trend (p¼0.1) towards decreased diagnosis related confidence in Group D1 and Group D2. Interobserver agreement was excellent (>0.75) for diagnosis-related confidence and good (>0.4) for other subjective values. FBP, filtered back projection; ASIR, adaptive statistical iterative reconstruction.

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

1173

Figure 2 Image quality of cCTs obtained in patients with no pathology. (a) Patients without acute or subacute pathology, supratentorial image quality. Scan performed using 120 kV and FBP (Group A), 120 kV and FBP (Group B), 100 kV and 20% ASIR (Group C), 100 kV and 30% ASIR for dose reduction and 30% ASIR/70% FBP blending (Group D1), and 30% ASIR for dose reduction and 40% ASIR/60% FBP blending (Group D2). (b) Patients without acute or subacute pathology, infratentorial image quality, Groups A, B, C, D1 and D2.

Table 5 Total doseelength product (DLP), computed tomography dose index (CTDIvol), and effective dose.

CTDIvol Total DLP (mGy$cm)

Group A

Group B

A vs. B

Group C

A vs. C

Group D1/2

A vs. D1/2

120 kV/FBP

100 kV/FBP

p-Value

100 kV/ASIR20

p-Value

100 kV/ASIR30

p-Value

31.912.2 396185

25.410 304124

0.072 0.07

18.32.45 26041

<0.0001 0.0028

10.83.55 14168

<0.0001 <0.0001

Compared to Group A, Groups B, C, D1/D2 showed a significant reduction of the doseelength product (DLP) by 23.2%, 34.4%, and 64.4%, respectively. FBP, filtered back projection; ASIR, adaptive statistical iterative reconstruction.

1174

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

Figure 3 Protocol C is adequate for everyday clinical imaging. (a) Occipital fracture in a 2-year-old boy after trauma. (b) Massive intracranial intraventricular hemorrhage in a 2-year-old girl under anticoagulation due to mitral valve replacement. (c) Chronic subdural hemorrhage, hygroma and ventricular drainage in a 4-year-old boy after tumor resection. (d) Massive edema in the thalamic regions in a 3-year-old boy with hemolytic-uremic syndrome.

Groups D1 and D2 were mainly lower compared to Group A with Group D2 performing better than Group D1 (Fig 4). The protocol used in Group C is considered adequate for everyday clinical imaging, and this CT protocol is now routinely used in paediatric patients in the authors’ clinics (Fig 3). The CT protocol with 30% ASIR and an increased noise index (Group D1/D2) further reduced both quantitative and qualitative image quality to such an extent that it was not considered adequate for everyday clinical use; however, the quality remains high enough for the diagnosis of life-threatening conditions, such as acute bleeding, fractures, or for the assessment of hydrocephalus, especially when blending is increased to 40% ASIR/60% FBP. In these cases, this protocol achieved a mean DLP of only 260 mGy$cm, which is particularly useful for repeated follow-up examinations of paediatric neurosurgical patients.

Korn et al.9 examined quantitative and qualitative image quality at reduced tube current rates in CT using sinogramaffirmed IR (SAFIRE, Siemens, Erlangen, Germany) cranial CT examinations compared to standard dose FBP (320 versus 255 mAs) in adult patients. At 20% dose reduction, reconstruction of cranial CT examinations using SAFIRE provided better objective and subjective image quality than FBP reconstruction (adult patients, 30 FBP and 30 SAFIRE cranial CT examinations).9 Contrary to the present study, the purpose of the study of Korn et al. was not to reduce dose while maintaining image quality, but to improve image quality while maintaining dose. At the time of manuscript editing, the largest study on IR in adult cranial CT was performed by Komlosi et al.18 who investigated 200 patients and showed that the use of an NI of 5 (compared to FBP and an NI of 4) and 40% ASIR blending

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

1175

Figure 4 Protocol D2 is adequate for follow-up imaging of acute severe events. (a) Fracture of the petrous bone in a 1-year-old boy after trauma. (b) Massive intracranial hemorrhage and hydrocephalus in a 10-year-old boy with Moyamoya disease. (c) Epidural hemorrhage in a 7-year-old girl after trauma. (d) Iarge chronic subdural hemorrhage and smaller acute hemorrhage in a 3-year-old girl under anticoagulation.

led to a 10.5% reduction in DLPs in adult cranial CT examinations, while image quality and noise remained comparable (100 FBP and 100 ASIR cranial CT examinations).18 In this study, the NI was gradually increased and varying levels of ASIR/FBP blending were used to compensate for the higher NI. Unfortunately, the authors did not analyse SNRs or CNRs in the brain, which makes it difficult to judge overall image quality and noise objectively. This may be problematic as image quality in IT regions is often more sensitive to dose variations during the scan due to the higher bone thickness. One of the first studies investigating the use of ASIR in paediatric cranial CT was conducted by Vorona et al.14 In 12 patients, the authors showed a 23.9% DLP reduction of cranial CT scans when 20% ASIR was applied during acquisition at 120 kV tube voltage. There was no significant reduction in image quality and interpretability (age 3e18 years, 12 FBP and 12 ASIR cranial CT examinations). The

authors did not find significant differences between the ASIR studies and non-ASIR studies with respect to diagnostic acceptability, sharpness, noise, or artefacts. Contrary to the work by Vorona et al.,14 the present study found reduced quantitative and qualitative levels to some extent in all groups. The fact that the tube voltage was modified in combination with ASIR in the present experimental set-up may explain these findings. McKnight et al.15 showed that the application of 30% ASIR in paediatric head CT examinations allowed for a 28% CTDIvol reduction in the subgroup of 3- to 12-year-old patients without substantially compromising sharpness, greyeWM differentiation or overall diagnostic quality (age 0e18 years, 49 FBP and 33 ASIR cranial CT examinations).15 Similar to Vorona et al., the authors used 120 kV in both the control and the ASIR collective, which may explain why the authors did not find significant differences in image quality.

1176

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177

determining whether the present results can be applied to other scanners and algorithms. In conclusion, IR algorithms have the potential to reduce radiation exposure in cranial CT examinations of children without compromising image quality. A CT protocol using 100KV, 20% ASIR and a 20% ASIR/80% FBP blending ratio decreases the DLP by 34.4% while producing scans with adequate image quality compared to a routine dose cranial CT. This CT protocol is recommended for everyday clinical practice in children 0e12 years of age. The use of a CT protocol with 100 kV, 30% ASIR, and a blending of 40% ASIR/60% FBP reduces the DLP by 64.4% and can be considered for follow-up imaging.

References Figure 5 Dose-length product (DLP) in cranial CTs using 120 kV and FBP (Group A), 100 kV and FBP (Group B), 100 kV and 20% ASIR (Group C) as well as 30% ASIR (Group D1 and Group D2).

Ono et al.19 reconstructed the same paediatric CT examinations using FBP and SAFIRE (age 3 months to 5 years, 78 cranial CT examinations).19 A similar study by Ho20 and colleagues compared image quality in paediatric head CT examinations reconstructed using different levels of iDOSE (Philips Healthcare, Andover, MA, USA) versus FBP.20 Both studies were able to show improved objective and subjective scores for IR reconstructed images. Like Korn et al. both groups did not try to reduce the dose while maintaining image quality, but to improve image quality while maintaining the dose. Although there are few studies examining IR in paediatric cranial CT, the present is one of the largest studies published thus far. It is the first study that combines ASIR and reduces tube voltage, resulting in significant dose reduction while maintaining adequate image quality for everyday clinical use. The present study did have some limitations. Firstly, no explicit patient group matching was performed and there were differences in patient ages between Groups B/ C and Group A; however, cranial diameters were well balanced in all groups. Secondly, image-quality assessment was based on the subjective grading of two radiologists, which may not have been completely blind, as it is possible to identify an ASIR image by its typical appearance; however, additional objective quantitative image analyses were employed to corroborate qualitative evaluation. Nevertheless, it has been questioned whether quantitative measures are an appropriate tool for evaluating the effectiveness of IR algorithms. Jensen et al.21 showed that lesion detection was not improved in ASIR reconstructed images compared to FBP reconstructed images of a liver phantom, even though noise decreased and CNR increased significantly.21 This may explain the differences in objective and subjective grading shown here. Considering the plethora of IR algorithms used by different companies further studies might be helpful in

1. Hall EJ, Brenner DJ. Cancer risks from diagnostic radiology. Br J Radiol 2008;81(965):362e78. 2. Strauss KJ, Goske MJ, Kaste SC, et al. Image gently: ten steps you can take to optimize image quality and lower CT dose for pediatric patients. AJR Am J Roentgenol 2010;194(4):868e73. 3. Pearce MS, Salotti JA, Little MP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. Lancet 2012;380(9840): 499e505. 4. Kalender WA, Wolf H, Suess C. Dose reduction in CT by anatomically adapted tube current modulation. II. Phantom measurements. Med Phys 1999;26(11):2248e53. 5. McCollough CH, Bruesewitz MR, Kofler Jr JM. CT dose reduction and dose management tools: overview of available options. RadioGraphics 2006;26(2):503e12. 6. Alkim E, Gurbuz E, Kilic E. A fast and adaptive automated disease diagnosis method with an innovative neural network model. Neural Networks 2012;33:88e96. 7. Wu TH, Hung SC, Sun JY, et al. How far can the radiation dose be lowered in head CT with iterative reconstruction? Analysis of imaging quality and diagnostic accuracy. Eur Radiol 2013;23(9):2612e21. 8. Kilic K, Erbas G, Guryildirim M, et al. Lowering the dose in head CT using adaptive statistical iterative reconstruction. AJNR Am J Neuroradiol 2011;32(9):1578e82. 9. Korn A, Bender B, Fenchel M, et al. Sinogram affirmed iterative reconstruction in head CT: improvement of objective and subjective image quality with concomitant radiation dose reduction. Eur J Radiol 2013;82(9):1431e5. 10. Kaul D, Kahn J, Huizing L, et al. Reducing radiation dose in adult head CT using iterative reconstruction d a clinical study in 177 patients. RoFo 2016 Feb;188(2):155e62. 11. Willemink MJ, de Jong PA, Leiner T, et al. Iterative reconstruction techniques for computed tomography. Part 1: technical principles. Eur Radiol 2013;23(6):1623e31. 12. Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med 2009;169(22):2078e86. 13. Willemink MJ, Leiner T, de Jong PA, et al. Iterative reconstruction techniques for computed tomography. Part 2: initial results in dose reduction and image quality. Eur Radiol 2013;23(6):1632e42. 14. Vorona GA, Zuccoli G, Sutcavage T, et al. The use of adaptive statistical iterative reconstruction in pediatric head CT: a feasibility study. AJNR Am J Neuroradiol 2013;34(1):205e11. 15. McKnight CD, Watcharotone K, Ibrahim M, et al. Adaptive statistical iterative reconstruction: reducing dose while preserving image quality in the pediatric head CT examination. Pediatr Radiol 2014;44(8):997e1003. 16. Kahn J, Grupp U, Kaul D, et al. Computed tomography in trauma patients using iterative reconstruction: reducing radiation

D. Kaul et al. / Clinical Radiology 71 (2016) 1168e1177 exposure without loss of image quality. Acta Radiol 2016 Mar;57(3):362e9. 17. Boning G, Schafer M, Grupp U, et al. Comparison of applied dose and image quality in staging CT of neuroendocrine tumor patients using standard filtered back projection and adaptive statistical iterative reconstruction. Eur J Radiol 2015;84(8):1601e7. 18. Komlosi P, Zhang Y, Leiva-Salinas C, et al. Adaptive statistical iterative reconstruction reduces patient radiation dose in neuroradiology CT studies. Neuroradiology 2014;56(3):187e93.

1177

19. Ono S, Niwa T, Yanagimachi N, et al. Improved image quality of helical computed tomography of the head in children by iterative reconstruction. J Neuroradiol 2015. 20. Ho C, Oberle R, Wu I, et al. Comparison of image quality in pediatric head computed tomography reconstructed using blended iterative reconstruction versus filtered back projection. Clin Imaging 2014;38(3):231e5. 21. Jensen K, Martinsen AC, Tingberg A, et al. Comparing five different iterative reconstruction algorithms for computed tomography in an ROC study. Eur Radiol 2014;24(12):2989e3002.