Comparison of the image quality and radiation dose of different scanning modes in head-neck CT angiography

Abstract

Objectives:

To analyze and compare the radiation dose and image quality of different CT scanning modes on head-neck CT angiography.

Methods:

A total of 180 patients were divided into Group A and Group B. The groups were further subdivided according to different scanning modes: subgroups A1, A2, A3, B1, B2, and B3. Subgroups A1 and B1 used conventional CT protocol, subgroups A2 and B2 used the kV-Assist scan mode, and subgroups A3 and B3 used the dual-energy gemstone spectral imaging protocol. The CT dose index and dose–length product were recorded. The objective image quality and subjective image evaluation was conducted by two independent radiologists.

Results:

The signal-to-noise ratios, contrast-to-noise ratios, and subjective scores of subgroups A3 and B3 were higher than the other subgroups. In subgroups B1 and B2, the subjective scores of 9 patients and 12 patients were lower than 3, respectively. The subjective scores of subgroups B1 and B2 were lower than the other subgroups. There was no statistically significant difference in signal-to-noise ratios, contrast-to-noise ratios, and subjective scores between subgroups A1 and A2. The effective dose of subgroup A2 was 41.7 and 36.4% lower than that in subgroups A1 and A3, respectively (p < 0.05). In Group B, there were no statistically significant differences in CT dose indexvol, dose–length product, and ED among the subgroups (p > 0.05).

Conclusion:

In the head-neck CT angiography, the kV-Assist scan mode is recommended for patients with body mass index between 18.5 and 34.9 kg m⁻²; gemstone spectral imaging scanning mode is recommended for patients with body mass index ≥34.9 kg m⁻².

Introduction

With the rapid development of CT, head-neck CT angiography (CTA) examination is becoming increasingly important in the evaluation of head and neck angiopathy and intracranial aneurysm.¹ However, the high radiation dose from the head-neck CTA examination is a concern for radiologists.^2,3 Since the radiation exposure in CT examination is associated with the risk of cancer,⁴ it is important to reduce the radiation dose while maintaining the image quality and diagnostic accuracy.^5–8 Therefore, the risk of adverse effects from radiation should be carefully considered when selecting CT scan modes, scan protocols, and clinical applications.

In recent years, many scan modes are developed and used to reduce the radiation dose of CT in clinical practice, such as kV-Assist scanning mode, and dual-energy gemstone spectral imaging (GSI) scanning mode. The kV-Assist scanning mode is a tube voltage selection feature that can suggest the optimal kV based on the attenuation from the patient topogram along the Z-axis. There is a consistent desired image quality despite the difference in patients. The different tube voltages required in the current tube benchmark and curves with the volumetric CT dose index (CTDI_vol) are automatically calculated according to the image quality requirement. In the case of X-ray tube system hardware, the minimum tube voltage scanning is selected. This feature works in conjunction with both Smart mA and Manual mA to adjust the noise index/mA as needed once the kV is selected. The kV-Assist then optimizes the scan parameters based on the selected clinical task.

GSI is an innovative spectral CT technique that uses fast kV switching dual energy acquisition technology, fast sampling, low afterglow, and high light output scintillator detector. The two different energies (80 and 140 kVp) switching in 0.25 ms can acquire paired images simultaneously at the same projection angle, which can feasibly display the target vessel system from the best contrast-to-noise ratio (CNR) curve. GSI improves image quality over the conventional imaging techniques because of its higher CNR, reduced beam hardening artifacts, enhanced material separation, and quantitative material information.

In clinical use, the reduction of tube voltage and tube current reduces the radiation dose, however, it also increases the image noise and has a negative impact on image quality.^9,10 Therefore, reducing the radiation dose without affecting the diagnosis has been intensely investigated.^11,12 Adaptive Statistical iterative Reconstruction (ASiR)-V is a reconstruction technique that enables reduction in image noise (standard deviation) and streak artifacts at low signal conditions while preserving the structure details of the image. Further, it reduces the dose required for routine imaging.

The purpose of this study was to comprehensively analyze the image quality and radiation dose of head-neck CTA in patients with different body mass index (BMI) by using three scanning modes (conventional scanning mode, kV-Assist scanning mode, GSI scanning mode) with ASiR-V, in order to select the optimal scanning mode conferring the lowest radiation dose. To the best of our knowledge, no previous studies have made this comparison.

Methods and materials

Patient population

This study was approved by our institutional review board, and written informed consent was obtained from all patients.

We prospectively enrolled patients who were recommended by clinicians to undergo head-neck CTA examination from August 2019 to January 2020. Clinical exclusion criteria included allergy to iodine contrast, pregnancy, heart failure, kidney failure, and metal implants in the scanning region. Finally, 4 patients were excluded due to prior severe allergic reaction to iodine contrast agent, 2 patients were excluded due to metal implants in their necks and 180 patients (92 males and 88 females; mean age 61 years; range, 22–85 years) were included in the study. These patients were divided according to their BMI into Group A (BMI between 18.5 and 34.9 kg m⁻) and Group B (BMI ≥34.9 kg m⁻²). Groups A and B were further divided into A1, A2, A3, B1, B2, and B3 subgroups according to the different scanning modes, with 30 subjects in each subgroup (Table 1).

Table 1.

Patient characteristics

Group	No. cases	Female/male	Age (years)	Scan length (mm)
A1	30	16/14	60 ± 12	364 ± 21
A2	30	13/17	63 ± 11	365 ± 24
A3	30	17/13	65 ± 14	362 ± 22
B1	30	16/14	67 ± 12	364 ± 23
B2	30	17/13	64 ± 13	363 ± 26
B3	30	14/16	62 ± 14	365 ± 25
F value		1.69	0.48	1.60
p value		0.430	0.738	0.213

CT protocol

All 180 patients were scanned on a Revolution CT scanner (GE Healthcare, Waukesha, WI). The conventional scanning mode was used in subgroups A1 and B1, the kV-Assist scanning mode in subgroups A2 and B2, and the GSI scanning mode in subgroups A3 and B3. The scan parameters included: (1) tube voltage: subgroups A1, B1 − 120 kVp; subgroups A2, B2 − 70–140 kVp; subgroups A3, B3 − rapid dual kVp (80 kVp and 140 kVp) switching in 0.25 ms; (2) tube current: subgroup A1, A2, B1 and, B2 − 3D Smart mA, ranging from 100 to 600 mA; (3) noise index: 6 at 5 mm thickness for subgroups A3, B3: GSI Assist, 280 mA, noise index: 6 at 5 mm thickness. The A1, B1, A2 and B2 subgroups applied 40% pre-ASIR for reducing radiation dose requirement. The other parameters that were the same for each group were as follows: 0.5 s tube rotation time, pitch factor of 0.992:1, 32 cm DFOV, matrix of 512 × 512. All images were reconstructed with 70% post-ASIR. The scan area extended from 1 cm below the aortic arch to the calvaria. The axial length ranged from 325 to 400 mm.

The contrast medium (Iopamidol, 370 mg ml⁻¹, Bracco, Munroe Township, NJ) was injected at 4.0 ml s⁻¹ into the right antecubital vein (median volume: 0.4 ml kg⁻¹), followed by a saline flush (40 ml) at the same rate using a high pressure injector (Missouri XD2001; Ulrich, Ulm, Germany). Bolus-tracking technology was used, and the region of interest (ROI: 10 mm²) was placed in the center of the aortic arch with a threshold value of 145 Hounsfield units to trigger the scan with a 2.3 s scan delay.

Objective image quality evaluation

All images with 0.625 mm thickness were transferred to the GE AW4.7 workstation for analyses and measurements. The ROI (2–5 mm²) was placed at the aortic arch, bilateral common carotid artery, middle cerebral artery, and same level of muscle tissue/brain parenchyma area of these vessels (Figure 1). The clone function of comparison was adapted to ensure the consistency of the size and shape of the ROI. The mean CT attenuation values and mean standard deviation (SD) values from three blood vessels were calculated as CT_artery and SD_artery, respectively. We used the values from the same level of muscle tissue/brain parenchyma area as the CT_background and SD_background to calculate signal-to-noise ratio (SNR) and CNR for vessels with the following calculations: SNR = CT_artery/SD_artery and CNR = (CT_artery − CT_background)/SD_background.

Figure 1.

Example of identical ROI measurements simultaneously placed in axial CT images. CT value and standard deviation were measured by placing the ROI with an area of 2–5 mm² at the aortic arch, bilateral common carotid artery and middle cerebral artery respectively. ROI, Region of interest.

Evaluation of subjective image quality

All head-neck CTA images were evaluated independently with a 5-grade scale by two radiologists with at least 15 years of experience in CTA imaging.¹³ All images were randomly assigned for evaluation, and the readers were blinded to the CT scan protocols. Image scores of ≥3 were considered to be in accordance with the diagnostic requirements. The scoring criteria are shown in Table 2.

Table 2.

Grading score for qualitative image analysis

Grading score	Noise	Vessel contrast	sharpness	Diagnostic confidence
5	Very little	Very clear	Sharpest	Fully diagnostic
4	Mild	Clear	Better than three, poorer than five	Good diagnostic
3	Moderate	Less clear	Minor burring in an acceptable image	Diagnostic
2	Obvious	Unclear	Poorer than three	Affecting diagnosis
1	Severe	Cannot be displayed	Blurry	Non-diagnostic

Radiation dose

The CTDI_vol and dose–length-product (DLP) were recorded from the CT scans, and the effective dose (ED) was calculated using the formula: ED = DLP×K, where K represents the radiological protection conversion factor for the adult head-neck: K = 0.0023 mSv·mGy⁻¹cm⁻¹.¹⁴

Statistical analyses

All statistical analyses were performed with the software package SPSS 21.0 (IBM, Chicago, IL). p value of < 0.05 was considered statistically significant. The variables were expressed as mean ± SD. One-way analysis of variance was used to compare differences among subgroups. The variables included patient age, BMI, image noise, CT attention, SNR, CNR, and radiation dose. The κ test was used to analyze the consistency of subjective scores for image qualities between the two radiologists (the range of k values: poor: 0–0.2, average: 0.21–0.40, medium: 0.41–0.60, good: 0.61–0.80, and excellent: 0.8–1.0).

Results

Overall characteristics

There was no significant difference in gender, age, or scanning range among subgroups (p > 0.05) (Table 1). In subgroup A2, 16 patients used 80 kVp tube voltage, 12 patients used 100 kVp tube voltage, and two patients used 120 kVp tube voltage. All 30 patients in the B2 subgroup used 120 kVp tube voltage.

Quantitative image analysis

The mean SNRs of patients in subgroups A1–B3 were 35.8 ± 7.4, 37.1 ± 9.7, 44.3 ± 13.7, 28.7 ± 7.1, 28.4 ± 7.3 and 45.6 ± 12.7, respectively. The mean CNRs of patients in subgroups A1–B3 were 31.2 ± 6.8, 33.1 ± 9.6, 43.1 ± 11.2, 28.5 ± 6.6, 27.5 ± 9.3 and 43.6 ± 11.6, respectively. The mean values of SNRs and CNRs in the A3 and B3 subgroups were higher than in other subgroups (all p < 0.05). The mean value of SNRs and CNRs in the B1 and B2 subgroups were lower than other subgroups (all p < 0.05). (Table 3, Figures 2 and 3)

Table 3.

Comparison of image quality and radiation dose of head-neck CTA patients in each subgroup

Group	Objective score				Subjective score	CTDIvol（mGy）	DLP（mGy）	ED（mSv）
	CT value（HU）	SD	SNR	CNR
A1	361.5 ± 44.3	10.1 ± 1.1	35.8 ± 7.4	31.2 ± 6.8	3.8 ± 0.2	11.3 ± 0.6	473.0 ± 38.3	1.2 ± 0.2
A2	416.6 ± 88.5	11.2 ± 1.4	37.1 ± 9.7	33.1 ± 9.6	4.1 ± 0.3	6.4 ± 1.6	275.4 ± 86.6	0.7 ± 0.1
A3	513.3 ± 92.7	11.6 ± 5.2	44.3 ± 13.7	43.1 ± 11.2	4.4 ± 0.3	11.2 ± 0.8	465.1 ± 58.2	1.1 ± 0.2
B1	352.7 ± 43.4	12.3 ± 1.3	28.7 ± 7.1	28.5 ± 6.6	3.2 ± 0.2	11.8 ± 1.2	493.9 ± 44.1	1.3 ± 0.4
B2	365.8 ± 84.2	12.9 ± 1.2	28.4 ± 7.3	27.5 ± 9.3	3.1 ± 0.2	11.5 ± 1.3	484.5 ± 62.5	1.1 ± 0.3
B3	505.7 ± 90.6	11.1 ± 5.1	45.6 ± 12.7	43.6 ± 11.6	4.4 ± 0.4	11.6 ± 1.1	477.8 ± 41.2	1.2 ± 0.4
F value	18.89	4.14	8.08	10.81	6.19	116.40	92.91	92.91
p value	All: <0.05	All: <0.05	A3 vs B3: 0.661	A3 vs B3: 0.649	A1 vs A2: 0.929	A2 vs A1: 0.000	A2 vs A1: 0.000	A2 vs A1: 0.000
			B1 vs B2: 0.806	B1 vs sB2: 0.784	B1 vs B2: 0.466	A2 vs A3: 0.000	A2 vs A3: 0.000	A2 vs A3: 0.000
			Rest: <0.001	Rest: <0.001	Rest: <0.05	A2 vs B1: 0.000	A2 vs B1: 0.000	A2 vs B1: 0.000
						A2 vs B2: 0.000	A2 vs B2: 0.000	A2 vs B2: 0.000
						A2 vs B3: 0.000	A2 vs B3: 0.000	A2 vs B3: 0.000
						Rest: >0.05	Rest: >0.05	Rest: >0.05

CNR, contrast-to-noise ratio; CTDI, CT dose index; DLP, dose–length product; HU, Hounsfield unit; SD, standard deviation; SNR, signal-to-noise ratio.

Figure 2.

Comparison of image quality and radiation dose of head and neck CTA obtained by different scanning modes. Aa, Ab, Ac, Ad, Ba, Bb, Bc, Bd, Ca, Cb, Cc, and Cd are the transverse axial images of 0.625 mm head and neck vessels, and MPR images of neck vessels obtained by three patients using the conventional scanning mode, kV assist scanning mode, and GSI scanning mode. The BMI of the three patients was 26.41, 22.92, and 29.41 kg m⁻², respectively. The mean CT value of the bilateral common carotid artery in Figure a and figure b is 367.0 HU, that in Figure e and figure f is 480.1 HU, that in Figure i and j is 593.1 HU, and that in three patients is higher than 300 HU. The subjective scores of CTA images of the three patients were 4.1, 4.3, and 4.6, respectively, which met the needs of clinical diagnosis. The ED values of three patients were 1.13 mSv, 0.73 mSv and 1.23 mSv, respectively, but the ED values of patients B (using kV assist) were 35.40 and 40.65% lower than those of patients a and c. BMI, Body mass index; CTA, CT angiography; HU, Hounsfield unit; MPR, Multiplanar reconstruction.

Figure 3.

Image quality of neck and shoulder from three scanning modes. Aa, Ab, Ac, Ba, Bb, Bc, Ca, Cb, and Cc are the MPR images of right and left carotid arteries and VR images of carotid arteries obtained by conventional scanning mode, kV commit scanning mode, and GSI scanning mode, respectively. Their BMI are 35.15, 35.45 and 35.78 kg m⁻² respectively. The subjective scores of three patients were 3.9, 4.1, and 4.5, respectively. The image noise of neck shoulder junction in patients C (GSI scanning mode) was significantly lower in groups a and b. BMI, Body mass index; GSI, Gemstone spectral imaging; MPR, Multiplanar reconstruction; VR, Volume rendering.

Subjective image analysis

All images were assessed by two radiologists with good subjective consistency (k = 0.85). The scores of subjective image qualities in the A1–B3 subgroups were 3.8 ± 0.2, 4.1 ± 0.3, 4.4 ± 0.3, 3.2 ± 0.2, 3.1 ± 0.2, and 4.4 ± 0.4, respectively. In particular, even though the mean subjective image quality scores of B1 and B2 subgroups were 3.2 and 3.1, meeting the diagnostic requirements, the two imaging experts believed that the noise of images at the junction of the neck and shoulder in B1 and B2 subgroups affected the analysis of plaque at the arterial wall. Hence, the subjective image quality scores of 9 patients in B1 subgroup and 12 patients in B2 subgroups were below 3 (Table 3), while the subjective scores of all patient images in the other 4 subgroups were above 3. Statistical analysis showed that the mean value of the subjective score of the B1 and B2 subgroups was lower than other subgroups (all p < 0.05). The mean values of the subjective score of A3 and B3 subgroups were higher than other subgroups (all p < 0.05). There was no significant difference found between subgroups A1 and A2 (p > 0.05) (Table 3, Figure 2).

Radiation dose estimations

The ED values of the A1–B3 subgroups were 1.2 ± 0.2 mSv, 0.7 ± 0.1 mSv, 1.1 ± 0.2 mSv, 1.3 ± 0.4 mSv, 1.1 ± 0.3 mSv, and 1.2 ± 0.4 mSv, respectively. The mean ED (0.7 ± 0.1 mSv) of patients in subgroup A2 was 41.7 and 36.4% lower than in subgroups A1 (1.2 ± 0.2 mSv) and A3 (1.1 ± 0.2 mSv), respectively (all p < 0.05). There was no significant difference in the radiation dose of each subgroup in Group B (p > 0.05). (Table 3)

Discussion

The tube voltage determines the X-ray penetrating power, the image quality of CTA, and the radiation dose for the examinees. The selection of appropriate tube voltage according to the size of the subject to reduce the radiation dose and ensure the quality of images during CT examinations has become a research hotspot.^15–18 The kV-Assist technology automatically selects the optimal kV value according to patient position, patient size, relative inspection density, initial kV value, and initial noise index/mA. It automatically selects the lowest CTDI_vol that can be generated to generate the desired image quality level. The combination of tube voltage and tube current/noise index, accompanied by automatic adjustment of window width and window level, can make the image quality meet the clinical requirements.^19–23 This avoids the deviation caused by selecting the tube voltage through personal experience. Thus, the combination of tube voltage and tube current is more personalized, and the optimal tube voltage and tube current can be automatically selected according to the body shape of each examinee, the size of the targeted part, and the tissue composition.

Our study showed that the selection of optimal scan protocol was dependent on patient BMI values. For patients in group A with BMI between 18.5 and 34.9 kg m⁻², although the GSI scanning mode provided the highest image quality, the image quality of all three scanning modes met the clinical diagnosis standard (subjective score >3 points). On the other hand, the mean ED of patients with the kV-Assist scan mode (0.7 ± 0.1 mSv) was 41.7 and 36.4% lower than the conventional scanning mode (1.2 ± 0.2 mSv) and GSI scan mode (1.1 ± 0.2 mSv), respectively (p < 0.05). This was because more low tube voltages were used in the kV-Assist scan mode. The decrease in the tube voltage causes the increase in the proportion of photoelectric effect between the X-ray and the iodine contrast agent in the head and neck vessels, and the decrease in the proportion of Compton scattering effect. At the same time, the energy of the X-ray photons is closer to the K boundary absorption limit of the iodine atoms (33.2 keV), such that the iodine of the contrast agent absorbs more X-ray energy and thereby increases the contrast between the blood vessels and the perivascular tissues during head-neck CTA examination. With the increase in contrast, the edge of the vessels in the head and neck becomes sharper and clearer. Further, the image contrast becomes more conducive to clinical diagnosis. With the decrease in kV, the noise of the CTA image of the head and neck will increase, especially at the junction of the head and shoulder. The ASiR technique is a reconstruction approach that improves the image quality and can also reduce the X-ray quantum noise induced by the lower kV. In our study, we found that the ASiR-V technique could reduce the X-ray quantum noise and help to improve the image quality in head-neck CTA using the kV-Assist. This is because ASiR-V includes a system noise model, object model, and physical model, which can reduce image noise and improve image density resolution through multiple iterations.^24–29

On the other hand, for large patients in group B with BMI ≥34.9 kg m⁻², the GSI scanning outperformed the other two scan protocols. The GSI scanning mode adopts instantaneous kVp switching technology to complete high- and low-energy scans in a very short time (0.25 ms). It can provide 101 sets of single energy images of 40–140 keV. The tissue structure of the targeted body parts has the best display under different single energies. Therefore, in the GSI scanning mode, its unique and best single energy imaging can increase the contrast of the image and ensure high quality.

The results of our study showed that for BMI ≥34.9 kg m⁻², the average CT value of arteries in GSI scanning mode, especially the CT value of the carotid shoulder, was higher than the other two scanning modes. Further, the higher CT value of arteries was more conducive to the multiple reconstruction display of head-neck CTA images, such as volume rendering and maximum density projection, which can be more intuitive and clearer through these reconstruction methods. It is helpful for the clinical diagnosis of head and neck artery disease to display the parent and branches of the head and neck arteries. Our results showed that for patients with BMI ≥34.9 kg m⁻², the image noise of head–neck shoulder junction CTA obtained using the conventional scanning mode and kV-Assist scanning mode was large, the inner wall of the blood vessel was unclear, and the subjective score was low. The subjective scores of 9 and 12 patients using the conventional and kV-Assist scanning mode were lower than 3, but the subjective scores of all patients using the GSI scanning mode were higher than 4. This is because GSI can optimize the best keV value of CTA blood vessels of the head and neck, especially at the junction of the neck and shoulder through the best contrast curve. The best keV values were closer to the boundary limit of absorption of iodine atoms to X-rays, hence provided more clearly display of the blood vessels of the head and neck, especially at the junction of the neck and shoulder.

This study also found that with the increase in BMI, the advantages of the kV-Assist scanning mode in reducing the radiation dose of the examinee gradually decreased, while the advantages of the GSI scanning mode gradually increased. This was due to the increasing value of tube voltage and tube current selected by the kV-Assist scanning mode with the increase in BMI. Among patients with BMI of 18.5−34.9 kg m⁻², 16 patients used 80 kVp tube voltage, 12 patients used 100 kVp tube voltage, and only 2 patients used 120 kVp tube voltage. On the other hand, for patients with BMI ≥ 34.9 kg m⁻², all 30 patients required 120 kVp tube voltage. However, GSI scanning mode could select the best keV value to display the contrast of the target blood vessel and surrounding tissue in the best state, according to the patient’s BMI and without the influence of tube voltage. In this study, the tube current of the GSI scanning mode was selected by the mA-Assist function of the scanner, and the mA was set as 280 mA. Therefore, with the increase in BMI, the radiation dose of the examinee did not rise, but the image quality remains ideal.

Our study had some limitations. First, only the image quality and radiation dose under the default parameters of the device were evaluated. Other parameters (such as noise index value, different weights of the iterative algorithm, etc.) that could also be applied to maintain the image quality and reduce radiation dose of the patient were not included. However, this is the direction of our future research. Second, this study had a small sample size and no other detailed grouping of BMI was done. We plan to increase sample size and make a more detailed grouping of BMI in our future research. Third, the three scanning modes were compared only in the head-neck CTA. In the future, the gradual expansion of the results to other body parts will be done.

In conclusion, it is recommended to use the kV-Assist scan mode for head-neck CTA in patients with BMI of 18.5 to 34.9 kg m⁻². GSI scan mode is recommended for head and neck CTA in patients with of BMI ≥ 34.9 kg m⁻².