A novel protocol for abdominal low-dose CT scans adapted with a model-based iterative reconstruction method

Abstract

PURPOSE:

This study aims to introduce a novel low-dose abdominal computed tomography (CT) protocol adapted with model-based iterative reconstruction (MBIR), To validate the adaptability of this protocol, objective image quality and subjective clinical scores of low-dose MBIR images are compared with the normal-dose images.

METHODS:

Normal-dose abdominal CT images of 58 patients and low-dose abdominal CT images of 52 patients are reconstructed using both conventional filtered back projection (FBP) and MBIR methods with and without smooth applying. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) are used to compare image quality between the normal-dose and low-dose CT scans. CT dose indices (CTDI) of normal-dose and low-dose abdominal CT images on post-contrast venous phase are also compared.

RESULTS:

The SNR, CNR and clinical score of low-dose MBIR images all show significant higher values (Bonferroni p < 0.05) than those of normal-dose images with conventional FBP method. A total of around 40% radiation dose reduction (CTDI: 5.3 vs 8.7 mGy) could be achieved via our novel abdominal CT protocol.

CONCLUSIONS:

With the higher SNR/CNR and clinical scores, the low-dose CT abdominal imaging protocol with MBIR could effectively reduce the radiation for patients and provide equal or even higher image quality and also its adaptability in clinical abdominal CT image diagnosis.

Keywords

Low-dose computed tomography (low-dose CT)model-based iterative reconstruction (MBIR)filtered back projection (FBP)Signal-to-noise ratio (SNR)contrast-to-noise ratio (CNR)

List of abbreviations

Computed tomography

Iterative reconstruction

MBIR

Model-based iterative reconstruction

FBR

Filtered back projection

SNR

Signal-to-noise ratio

CNR

Contrast-to-noise ratio

1 Introduction

Since its development in the 1970 s, computed tomography (CT) has proven to be a versatile imaging technique in diagnostic imaging, and the use of CT in the assessment of abdominal diseases rises exponentially in the decades [1]. The ionizing radiation is considered potentially harmful, and its risk from stochastic effects is thought to increase with radiation dose. Therefore, radiologists/radiographers face the challenge of balancing image quality and radiation dose.

The growing use of CT, with its higher radiation exposure relative to other imaging modalities, has driven interest in CT protocol optimization, especially the low-dose optimizing. Common methods of reducing dose include lowering X-ray tube amperage and X-ray tube voltage. Given continued advances in CT technology, radiation exposure to patients must be continuously reassessed and be kept as low as reasonably possible while maintaining diagnostic yield.

Dose reduction techniques, such as tube voltage reduction [2], noise reduction filters [3], and automated tube current modulation [4], lead to enhancement of the image contrast, and those methods are often limited due to serious image noise encountered when low-dose images are reconstructed with conventional filtered back projection (FBP) [5]. A complete abdomen imaging protocol includes several scans: prior to contrast agent injection (pre-contrast scan), one scan immediately after contrast agent injection (post-contrast arterial-phase scan), and several scans post to contrast agent injection (post-contrast venous-phase/delayed phase scan). The purpose of the pre-contrast scans is chiefly used to explore the abnormal tissues with high density, e.g., the calcium or fresh hematoma. In the post-contrast arterial phase scan, it is used to acquire the arterial structures or contrast extravasation (bleeding). Therefore, the pre-contrast and post-contrast arterial-phase scans can be applied with low-Kvp method, which would not only enhance the image contrast but also reduce the overall radiology doses in square fashion. On the contrary, the purpose of the post-contrast venous-phase scan is to explore the abnormal anatomy and lesions with small various intensities difference to the normal tissues. It would be difficult for low-dose method implemented into the post-contrast venous phase scans. An effective way to reduce noise within CT images is to use iterative reconstruction (IR) algorithms instead of conventional FBP reconstruction, which suffers from noise-related artifacts and limited low contrast detectability in low-dose settings [6, 7].

The model-based iterative reconstruction (MBIR) (Veo; GE Healthcare, GE Medical Systems, Waukesha, WI, USA), using a more complex system of prediction models, is a type of iterative reconstruction and can reduce image noise and improve image quality compared with conventional reconstruction [8 –10]. Previous literatures show that it can reduce up to 80% of dose in phantom scans [11], in vivo adult [12], and in vivo pediatric studies [13]. About the application of MBIR in low-dose CT, previous literature of liver imaging demonstrated about dose reductions in the order of 59% with even better imaging quality while compared MBIR with conventional hybrid iterative reconstruction [14]. In clinical diagnosis, low-dose abdominal-pelvic CT performed with MBIR is a potential radiation dose reduction strategy for imaging patients presenting with acute abdominal pain [15].

To implement an effective low-dose abdomen imaging protocol, this study aims to introduce one novel protocol adapted with MBIR in venous phase imaging. To validate the adaptability of MBIR in low-dose CT imaging, we compared the low-dose CT image with MBIR reconstruction to the normal-dose CT abdominal images with FBP reconstruction in this study. These comparisons were based on the objective signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and subjective clinical scores of the abdominal images.

2 Materials and methods

2.1 Participants

Raw data of one hundred and ten patients were collected retrospectively in this study between October 2020 and June 2021. Fifty-eight patients (17 females; mean age, 65±11 years-old) were scanned using the normal dose (noise-index-15, NI-15) protocol, and another fifty-two (18 females; mean age, 65±13 years-old) were scanned using low dose (NI-22) protocol. The indications for abdominal CT exams were for follow-up of intra-abdominal malignancy (43 patients in NI15, and 35 patients in NI22), unknown nature of intra-abdominal mass lesions (7 in NI15, and 12 in NI22), liver cirrhosis for regular follow up (2 in NI15, and 3 in NI22), and others including fever of unknown origin, acute abdominal pain or ascites (6 in NI15, and 2 in NI 22). This study was approved by the Research Ethics Committee of Kaohsiung Veterans General Hospital - Joint Institutional Review Board (IRB21-CT12-16).

2.2 CT scanning Technique

All patients (both the normal dose and low-dose groups) underwent CT of the abdomen using the same imaging protocol performed on a GE Revolution HD CT scanner (GE Healthcare, Waukesha, WI) with the following parameters in pre-contrast and post-contrast arterial phase scanning: gantry speed of 0.5 seconds, pitch of 0.984:1, table speed of 39.375 mm/rotation, beam collimation of 40 mm with detector configuration of 64×0.625 mm, 100 kVp, and tube current modulation (Smart mA, GE Healthcare) was set by NI values of 22 according to the characteristics of the scanners. The imaging parameters in post-contrast venous phase scanning are listed as follows: gantry speed of 0.5 seconds, pitch of 0.984:1, table speed of 39.375 mm/rotation, beam collimation of 40 mm with detector conuration of 64×0.625 mm, 120 kVp, and tube current modulation was set by NI values (NI = 15 for normal-dose scans and NI = 22 for low-dose scans), according to the characteristics of the scanners. To validate the adaptability of low-dose venous-phase imaging, the axial images at 0.625 mm acquired and reconstructed with 2.5 mm of spatial resolution were performed with MBIR by Veo^TM software, (Veo 2.0, GE Healthcare) and conventional FBP methods. Additionally, a smoothing process is also applied with those two reconstructions for comparison. The details of the novel low-dose CT abdominal protocol are listed in the Table 1.

Table 1
The protocol for the low-dose CT scan

Pre-contrast Post-contrast Post-contrast

phase A-phase V-phase

kVp 100 100 120

MAS mode Smart MA mode Smart MA mode Smart MA mode

Noise Index 15 15 22

Scan Mode Helical mode Helical mode Helical mode

Spatial coverage 230 mm 230 mm 230 mm

Reconstruction FBP FBP MBIR (VEO)

Scan time 3-4 seconds 3-4 seconds 3-4 seconds

Reconstruction time 7seconds 7seconds 30 minutes

	Pre-contrast	Post-contrast	Post-contrast
kVp	100	100	120
MAS mode	Smart MA mode	Smart MA mode	Smart MA mode
Noise Index	15	15	22
Scan Mode	Helical mode	Helical mode	Helical mode
Spatial coverage	230 mm	230 mm	230 mm
Reconstruction	FBP	FBP	MBIR (VEO)
Scan time	3-4 seconds	3-4 seconds	3-4 seconds
Reconstruction time	7seconds	7seconds	30 minutes

2.3 The noise-index used in the low-dose venous-phase imaging

Prior to this study, low-dose imaging scans implemented with NI of 18, 20, and 22 were collected to investigate the adaptability in clinical diagnosis. With the subjective comparing, two abdominal radiologists of 15 and 30 years’ experiences made consensus to adopt NI-22 as our novel low-dose venous-phase abdominal imaging.

2.4 Data analysis

To acquire objective image quality (Fig. 1) of normal dose and low dose CT images, ROIs of 4-slice images including liver, spinal muscle, and fat were selected manually from each patient by two experienced radiology technologists to calculate the SNR and CNR by the following equations,

Fig. 1

The ROI selection for SNR/CNR calculation. Two 12-cm²-ROIs of liver (ROI1) and fat (ROI2) of each participant were selected by two experienced radiographers.

$SNR = \frac{average (S_{liver})}{standard deviation (S_{fat})}$ (1)

$CNR = \frac{average (S_{liver}) - average (S_{spinalmuscle})}{standard deviation (S_{fat})}$ (2) where S means signal intensity (CT numbers). Wilcoxon Ranksum test was used to compare the image quality between normal dose and low dose CT images.

The subjective imaging quality in clinical diagnosis was scored by two abdominal radiologists by using clinical rating of 1–10 scale. Of which, 10 represents the most excellent imaging quality, and 1 illustrates the poorest imaging quality. Clinical rating between radiologists were compared by Pearson Correlation. The results were compared with Wilcoxon Ranksum test between normal dose and low dose CT images.

3 Results

The details of the included patients are listed in Table 2. With the low-100kVp method, we can reduce the CTDI dose to about 3.83±1.22 mGy in pre-contrast and 3.87±1.22mGy in post-contrast arterial-phase scans, respectively (listed in the Supplementary Table 1). In the post-contrast venous-phase scans, the CT dose index (CTDI) of images with NI-22 protocol (5.3±1.6 mGy) is significantly lower (z-value = –7.6365; p < 10^-13) than those of NI-15 protocol (8.7±2.8 mGy) (Fig. 2 and Table 2). SNR and CNR of all the datasets were listed in the Table 3, and the details of intensity and statistic value are listed in the Supplementary Table 2 and 3. Compared with the FBP with normal-dose (NI-15), SNR of low dose (NI-22) images reconstructed by FBP with smooth and MBIR with/without smooth showed higher values, and so did the CNR (the details are listed in the Table 3 and Fig. 3a-b, and the linearity of each value are listed in the Supplementary Table 4).

Table 2
The demographic data

Scanned with Normal-dose Scanned with Low-dose

Number 58 (17 females) 52 (18 females)

Age 64.8 (11.2) 64.9 (12.9)

BMI 25.1 (5.5) 25.1 (3.2)

CTDI (mGy) 8.7 (2.8) 5.3 (1.6)

DLP (mGy-cm) 474.4 (190.5) 230.5 (100.4)

	Scanned with Normal-dose	Scanned with Low-dose
Number	58 (17 females)	52 (18 females)
Age	64.8 (11.2)	64.9 (12.9)
BMI	25.1 (5.5)	25.1 (3.2)
CTDI (mGy)	8.7 (2.8)	5.3 (1.6)
DLP (mGy-cm)	474.4 (190.5)	230.5 (100.4)

Fig. 2

The CT dose values of NI-15 and NI-22 V-phase scans. The Dose reduction rate is about 40% (*: multiple comparison corrected p < 0.05).

Table 3

The mean and standard deviation of SNR and CNR of different reconstruction methods among 2 NI protocols

Normal dose (NI-15)		Low dose (NI-22)
		Mean	Standard	Mean	Standard		P-value
			deviation		deviation
FBP	SNR	30.83	3.88	23.29	4.06	<0.001
	CNR	0.90	0.40	0.75	0.32	0.030
FBP (smooth)	SNR	52.69	8.22	39.27	6.87	<0.001
	CNR	1.56	0.68	1.28	0.54	0.018
MBIR	SNR	51.59	6.51	46.94	4.89	<0.001
	CNR	1.49	0.61	1.48	0.51	0.974
MBIR (smooth)	SNR	75.93	10.84	67.07	7.86	<0.001
	CNR	2.17	0.87	2.12	0.74	0.732

Fig. 3

The boxplot of SNR, CNR, and clinical scores of all datasets. (*: multiple comparison corrected p < 0.05).

Table 4

The z-values of comparison of SNR and CNR between 2 NI protocols. Positive value means that images with normal dose protocol is better than images with low dose protocol, and vice versa

SNR		Normal dose (NI-15)
		FBP	FBP	MBIR	MBIR
			(smooth)		(smooth)
Low dose (NI-22)	FBP	7.3192*	8.9656*	9.0255*	9.0255*
	FBP (smooth)	–6.7444*	7.0977*	7.0857*	9.0255*
	MBIR	–8.8638*	4.3137*	4.0083*	9.0015*
	MBIR (smooth)	–9.0255*	–7.2533*	–8.1933*	4.3436*
CNR		Normal dose (NI-15)
		FBP	FBP	MBIR	MBIR
			(smooth)		(smooth)
Low dose (NI-22)	FBP	2.2362	6.6127*	6.6067*	7.9778*
	FBP (smooth)	–3.9664*	2.3619	1.9667	5.7925*
	MBIR	–5.9422*	0.4221	–0.0210	4.6070*
	MBIR (smooth)	–7.9778*	–4.0083*	–4.4214*	0.3083

*: multiple comparison corrected p < 0.05.

In the result of clinical rating, the clinical scores of normal-dose FBP CT images showed significantly higher (Bonferroni p < 0.05) than those of low-dose FBP CT images while no significance was observed with that of low-dose CT images by FBP with smooth method. On the contrary, the clinical scores of low-dose CT images by MBIR with/without smooth methods were significantly higher (Bonferroni p < 0.05) than those of normal-dose conventional FBP CT images (shown on Fig. 3 c, Table 5 and Fig. 4).

Table 5

The comparisons of clinical scores of different reconstruction methods among 2 NI protocols

	Reviewer #1	Reviewer #2	Pearson
			correlation
FBP (Normal-dose)	5.4 (0.6)	5.5 (0.5)	0.878
FBP (Low-dose)	4.5 (0.7)	4.7 (0.7)	0.752
FBP with smoothing (Low-dose)	5.6 (0.7)	5.7 (0.7)	0.888
MBIR (Low-dose)	8.5 (0.6)	8.4 (0.6)	0.891
MBIR with smoothing (Low-dose)	8.0 (0.7)	8.0 (0.8)	0.929

Fig. 4

CT images of different NI values and reconstruction techniques. NI15 reconstructed by MBIR (a) and FBP (b) in one patient. NI 22 reconstructed by MBIR (a) and FBP (b) in another patient.

3 Discussions

This study aims to introduce the low-dose abdominal protocol and to validate the adaptability of MBIR in low-dose abdominal CT image based on the objective image quality and the subjective clinical scores. MBIR method with/without smooth kernels shows significant higher SNR, CNR, and clinical scores as compared with the normal-dose FBP CT images. The image quality of normal-dose FBP CT images are the minimal requirement in imaging diagnosis, and our result demonstrates the practicality and feasibility of low-dose V-phase images with MBIR in the diagnosis.

In our result, the MBIR images with smoothing shows better SNR and CNR than that without smoothing while the subjective scores of MBIR without smoothing are better than those without smoothing. In smoothing, the data points of images are modified, and individual points would contain value from the adjacent points. So, points with higher intensity than the adjacent points (presumably because of noise) are reduced, and points that are lower than the adjacent points are increased leading to a smoother signal. However, the adjacent information would also be introduced in the individual points, and the smoothing would lead to a blurred image and to reduced clinical qualities for diagnosis. Therefore, we adopt the application of MBIR without smoothing as our novel low-dose CT protocol. This protocol can provide acceptable image quality and good diagnostic power while reduces 40% radiation dose for patients’ safety. Shuman et al. [14] comparing the imaging quality and radiation dose of abdominal CT images with MBIR and hybrid iterative reconstruction, reported a dose reduction of 59% (6.2 vs 15.2 mGy). In the present study, the CTDI is even lower (5.3 mGy) than that of Shuman’s series.

One main disadvantage of applying the MBIR method is processing time. In the MBIR method, it costs about 30 minutes of processing time which is much longer than that of the regular FBP method. Considering with the convenience and dose-reducing purpose of the novel protocol, we apply MBIR with the setting of 120 kVp and low-mAs only to the post-contrast V-phase scans, while adopting the setting of low-kVp with FBP reconstruction to the pre-contrast and post-contrast A-phase scans. With the novel setting, the abdominal protocol can not only reach the balance of dose reducing for better patient safety and image quality for equal diagnosis as the images with normal dose setting within an acceptable procedure time. With the novel protocol, a total of around 40% dose reduction could be achieved.

This study was limited in several aspects. First, the study population was small. Second, inter-observer agreement for subjective image quality was not evaluated, and we preferred evaluating overall variation in quality of each radiologist. Third, the scanning tube current may differ between patients with low-dose scanning due to the usage of automatic current modulation; further studies may be needed.

In conclusion, with the better SNR/CNR and clinical scores, the low-dose V-phase abdominal CT images with MBIR reconstruction could effectively reduce the radiation dose for patients and provide equal or even greater quality in clinical abdominal CT diagnosis.

Competing financial and/or non-financial interests

Two co-authors, C.-W. Li, and A.-C. Chen, are employees of GE Healthcare, Taiwan, and they provided protocol information for the CT scans. C.-W. Li contributed to the statistical analysis and to the Figures illustration based on the result. Hwang. No funding was received from GE Healthcare for publication activities.

Data availability statements

The datasets generated and analyzed during the current study are not publicly available due to the policy of the Research Ethics Committee of Kaohsiung Veterans General Hospital - Joint Institutional Review Board but are available from the corresponding author on reasonable request.

Footnotes

The supplementary material is available in the electronic version of this article: .

References

Mettler

F.A.

Jr , Thomadsen

B.R

Bhargavan

et al. Medical radiation exposure in the U.S. in: preliminary results, Health Phys 95(5) (2008), 502–507.

Seyal

A.R.

, Arslanoglu

, Abboud

S.F.

et al. CT of the abdomen with reduced tube voltage in adults: A practical approach, Radiographics 35(7) (2015), 1922–1939.

Kalra

M.K.

, Maher

M.M.

, Sahani

D.V.

et al. Low-dose CT of the abdomen: evaluation of image improvement with use of noise reduction filters pilot study, Radiology 228(1) (2003), 251–256.

Kalra

M.K.

, Maher

M.M.

, Toth

T.L.

et al. Comparison of Z-axis automatic tube current modulation technique with fixed tube current CT scanning of abdomen and pelvis, Radiology 232(2) (2004), 347–353.

Schofield

, King

, Tayal

et al. Image reconstruction: Part 1 –understanding filtered back projection, noise and image acquisition, J Cardiovasc Comput Tomogr 14(3) (2020), 219–225.

Caywood

, Paxton

, Boll

et al. Effects of model-based iterative reconstruction on image quality for low-dose computed tomographic angiography of the thoracic aorta in a Swine model, J Comput Assist Tomogr 39(2) (2015), 196–201.

Huang

, Wan

, Chen

et al. An iterative reconstruction method for sparse-projection data for low-dose CT, J Xray Sci Technol 29(5) (2012), 797–812.

Deák

, Grimm

J.M.

, Treitl

et al. Filtered backprojection, adaptive statistical iterative reconstruction, and amodel-based iterative reconstruction in abdominal CT: anexperimental clinical study, Radiology 266(1) (2013), 197–206.

Notohamiprodjo

, Deak

, Meurer

et al. Image quality of iterative reconstruction in cranial CT imaging: comparison of model-based iterative reconstruction (MBIR) and adaptive statistical iterative reconstruction (ASiR), Eur Radiol 25(1) (2015), 140–146.

10.

Bie

, Yang

, Li

et al.Impact of deep learning-based image reconstruction on image quality compared with adaptive statistical iterative reconstruction-Veo in renal and adrenal computed tomography, J Xray Sci Technol 30(3) (2022), 409–418.

11.

Hérin

, Gardavaud

, Chiaradia

et al. Use of Model-basediterative reconstruction (MBIR) in reduced-dose CT for routinefollow-up of patients with malignant lymphoma: dose savings, imagequality and phantom study, Eur Radiol 25(8) (2015), 2362–2370.

12.

Neroladaki

, Botsikas

, Boudabbous

et al. Computed tomography of the chest with model-based iterative reconstruction using a radiation exposure similar to chest X-ray examination: preliminary observations, Eur Radiol 23(2) (2013), 360–366.

13.

Miéville

F.A.

, Berteloot

, Grandjean

et al. Model-basediterative reconstruction in pediatric chest CT: assessment of imagequality in a prospective study of children with cystic fibrosis, Pediatr Radiol 43(5) (2013), 558–567.

14.

Shuman

W.P.

, Chan

K.T.

, Busey