Comparison of different iterative reconstruction algorithms with contrast-enhancement boost technique on the image quality of CT pulmonary angiography for obese patients

Abstract

Objective

To evaluate the effect of the contrast-enhancement-boost (CE-boost) postprocessing technique on improving the image quality of obese patients in computed tomography pulmonary angiography (CTPA) compared to hybrid iterative reconstruction (HIR) and model-based iterative reconstruction (MBIR) algorithms.

Methods

This prospective study was conducted on 100 patients who underwent CTPA for suspected pulmonary embolism. Non-obese patients with a body mass index (BMI) under 25 were designated as group 1, while obese patients (group 2) had a BMI exceeding 25. The CE-boost images were generated by subtracting non-contrast HIR images from contrast-enhanced HIR images to improve the visibility of pulmonary arteries further. The CT value, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were quantitatively assessed. Two chest radiologists independently reviewed the CT images (5, best; 1, worst) across three subjective characteristics including diagnostic confidence, subjective image noise, and vascular contrast. The Friedman test and Dunn-Bonferroni correction were used for statistical analysis.

Results

The CE-boost had significantly higher CT values than HIR and MBIR in both groups (all p < 0.001). The MBIR yielded the lowest image noise compared with HIR and CE-boost (all p < 0.001). The SNR and CNR of main pulmonary artery (MPA) were significantly higher in CE-boost than in MBIR (all p < 0.05), with HIR showing the lowest values (all p < 0.001). Group 2 MBIR received significantly better subjective image noise scores, while the diagnostic confidence and vascular contrast scored highest with the group 2 CE-boost (all p < 0.05).

Conclusion

Compared to the HIR algorithm, both the CE-boost technique and the MBIR algorithm can improve the image quality of CTPA in obese patients. CE-boost had the greatest potential in increasing the visualization of pulmonary artery and its branches.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12880-024-01447-6.

Introduction

Pulmonary embolism (PE) is highly prevalent, with a complex etiology and nonspecific symptoms, standing as the third leading cause of acute cardiovascular death worldwide [1]. Computed tomography pulmonary angiography (CTPA) has emerged as the first-choice modality for confirming PE due to its minimal invasiveness and excellent spatial resolution [2, 3]. However, in obese patients, the larger body mass significantly contributes to increased image noise and a reduced concentration of contrast medium per unit volume, potentially impacting the diagnostic accuracy of PE. Striving to maintain optimal image quality, medical professionals often resort to increasing the tube current, which exposes obese patients to higher radiation doses. Previous studies have indicated that innovative techniques and strategies can mitigate the challenges associated with image noise in obese patients [4–6].

Several reconstruction algorithms have been employed to reduce image noise without compromising spatial resolution and diagnostic accuracy. One such reconstruction algorithm is the hybrid iterative reconstruction (HIR), which uses only statistical system modeling and forward projection steps, and it has been widely applied in clinical practice [7]. However, owing to its limited performance on noise reduction, patients with a higher BMI may still not benefit from HIR [8–11]. Previous studies have demonstrated that MBIR is more effective at visualizing small arteries and subtle structures [8].

Besides reconstruction methods, a purely postprocessing contrast enhancement boost (CE-boost) technique was developed to increase the visualization of blood vessels, aiming to enhance the image quality further [12]. With an accurate deformable registration algorithm, the iodine image is generated by subtracting a non-contrast image from an enhanced image. Then, the CE-boost image with increased degree of contrast effect, is obtained by adding the iodine image to the original enhanced image with an automatic denoising procedure. This technique has been applied in chest and abdominal CT angiography to improve the image quality of peripheral vasculature [13, 14].

We hypothesized that the CE-boost technique could improve the image quality of CTPA in obese patients, resulting in a clearer visualization of peripheral pulmonary arteries compared to the conventional HIR and MBIR algorithms. Therefore, the purpose of this study was to assess the effect of CE-boost on quantitative and qualitative image quality of CTPA in obese patients.

Materials and methods

Study population

A single-centered prospective study of 122 patients who underwent CTPA examination from February to March 2023 were enrolled. Exclusion criteria were pregnant (n = 4), untreated hyperthyroidism (n = 6), iodine hypersensitivity (n = 4), and renal insufficiency (n = 8). A total of 100 cases were finally collected. Patients were divided into two groups according to the body mass index (BMI): group 1 (n = 50, BMI < 25) and group 2 (n = 50, BMI ≥ 25), which allowed us to compare the differences of vascular attenuation between non-obese and obese patients. A BMI exceeding 25 was considered indicative of obesity according to the guidelines for the management of obesity disease [5]. This study was approved by the Institutional Review Board of our institution, and written informed consent was obtained.

CT scanning protocol

All CT examinations were performed with a 320-row detector CT scanner (Aquilion ONE Genesis, Canon Medical Systems Corporation, Japan) from the head to foot direction. The scanning parameters were as follows: tube voltage of 120 kV, automatic tube current adjustment (^SUREExposure 3D, Canon Medical Systems), noise index set at 12.5, collimator width of 80 × 0.5 mm, pitch factor of 0.813, rotation time of 0.35s. The patients were placed in the supine position with their arms raised above the head. The 18 G syringe was embedded in the right median cubital vein, connected to a double cartridge hyperbaric syringe. A fixed volume of 40 mL iodine contrast agent (350 mg I/ml, Jiangsu Hengrui Medicine Co., Ltd., Jiangsu, China) was injected at a flow rate of 4.5 ml/s, followed by 20 mL of saline at the same injection rate. The non-contrast scan was performed first, covering the area from the subclavian region to the costophrenic angle. An automatic scanning system was used with a trigger threshold of 120 HU in the pulmonary artery trunk. Two seconds after the trigger, CTA scans were automatically performed.

Image reconstruction and processing

Both non-contrast and contrast-enhanced images were reconstructed using HIR (Adaptive Iterative Dose Reduction 3-Dimensinal, AIDR 3D; FC18) and MBIR (forward-projected model-based iterative reconstruction solution, FIRST, Body standard), respectively. Moreover, the non-contrast and contrast-enhanced HIR images of all patients were sent to a dedicated postprocess software (^SURESubtraction Iodine mapping, Canon Medical Systems) to generate CE-boost datasets. Therefore, there were three datasets for comparison, including HIR, MBIR, and CE-boost.

Objective image quality assessment

The objective image quality evaluation was performed with a circular region of interest (ROI) drawn in the main pulmonary artery, right and left segmental pulmonary arteries, and paravertebral muscle, respectively. The ROIs covered at least 2/3 of the lumen section while avoiding the vessel edges. Two measurements of CT values in the segmental level was averaged. The background noise was defined as the standard deviation (SD) of attenuation measured in the paravertebral muscle. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated according to the following formula:

$$SNR=C{T}_{pulmonaryartery}/S{D}_{pulmonaryartery}$$

$$CNR=(C{T}_{pulmonaryartery}-C{T}_{paravertebralmuscle})/S{D}_{paravertebralmuscle}$$

Subjective image quality assessment

The subjective assessment included diagnostic confidence, subjective image noise, and vascular contrast. Two radiologists with 3 and 5 years of experience in chest CT independently graded the CT images, blinded to the image acquisition approaches and patient information. Diagnostic confidence was evaluated as follows: 5 = full confidence in diagnosis; 4 = predominately confident; 3 = confident; 2 = generally confident; 1 = no confidence to diagnose. Subjective image noise was rated as: 5 = minimal; 4 = mild; 3 = average; 2 = moderate; 1 = severe. The vascular contrast consisted of reviewing pulmonary artery trunks and their branches, which was scored as follows: 5 = excellent contrast, clearly identified to distal subsegmental branches; 4 = good contrast, clearly identified to subsegmental branches; 3 = acceptable contrast and somewhat vague, identified to segmental branches; 2 = blurry and not clear, identified to lobar branches; 1 = poor contrast, only trunks, left and right arteries identified [13, 15].

Radiation dose

The volume CT dose index (CTDIvol) and dose length product (DLP) were obtained from the dose reports. The effective dose (ED) was calculated as the product of DLP and a chest-specific conversion factor of 0.014 mSv/mGy*cm [16].

Statistical analysis

Statistical analysis was performed using the SPSS software (Version 22.0; IBM, New York, USA). The Kolmogorov-Smirnov test was used to test whether the data satisfied the normal distribution. The Friedman test and Bonferroni correction were used to compare the objective and subjective image quality between different acquisition techniques. In addition, the Mann-Whitney test was used to compare the results between different BMI populations. P-value < 0.05 was considered as a statistical difference for all comparisons. Cohen’s Kappa was used to test the consistency of the subjective analysis results; the obtained kappa values were categorized as follows: 0.81-1.00, excellent; 0.61–0.80, good; 0.41–0.60, moderate; 0.21–0.40, fair; and 0.00–0.20, poor.

Results

Study population

A total of one hundred patients with suspected PE were finally included in this study (51 males and 49 females). The mean BMI of group 1 and 2 were 21.46 ± 1.79 kg/m² and 28.58 ± 3.75 kg/m², respectively. Detailed demographic characteristics of patients can be found in Table 1.

Table 1.

Demographic characteristics of patients

Parameter	Low BMI (group 1)	High BMI (group 2)	P value
Number of participants	50	50
Age (years)	65.38 ± 13.72	64.86 ± 14.79	0.87a
Male	23 (46.0)	28 (56.0)	0.42b
Body mass index (kg/m2)	21.46 ± 1.79	28.58 ± 3.75	< 0.001a
Radiological findings			0.76b
Pulmonary embolism	13 (26.0)	10 (20.0)
Pneumonia	10 (20.0)	14 (28.0)
Carcinoma	9 (18.0)	8 (16.0)
Pulmonary tuberculosis	9 (18.0)	6 (12.0)
Cardiovascular disease	8 (16.0)	9 (18.0)
Metastases	1 (2.0)	3 (6.0)

Data expressed as mean ± standard deviation or counts (percentage)

BMI body mass index

^aIndependent samples t-test

^bFisher’s exact test

Objective image quality analysis

Compared to HIR and MBIR images in group 1, the CT value was decreased and the image noise was increased in group 2 (all p < 0.001). For both MPA and SPA in each group, CE-boost images revealed the highest CT values among the three datasets, while the image noise was significantly lower in MBIR images than in HIR and CE-boost images (all p < 0.001). A slightly higher increase in image noise was observed in group 2 CE-boost images compared to group 2 HIR images (p < 0.05). In group 1, the SNRs and CNRs of the MPA and SPA were superior in CE-boost images compared to HIR and MBIR images (all p < 0.05). Group 2 CE-boost images showed the highest SNR and CNR for the MPA compared with group 2 HIR and MBIR images (all p < 0.05), with SNR and CNR for the SPA comparable to MBIR images (all p > 0.30). The detailed results of objective image quality are exhibited in Table 2; Fig. 1. P-values for pairwise and non-pairwise comparisons of objective image quality are shown in Supplementary Tables S1-S2. One representative case is presented in Fig. 2.

Table 2.

The objective image quality assessment

	HIR	MBIR	CE-boost		HIR	MBIR	CE-boost
Parameters	Low BMI (group 1)			P-value overall	High BMI (group 2)			P-value overall
Image noise (HU)
MPA	21.41 ± 3.13	15.92 ± 3.12	22.30 ± 4.97	< 0.001	22.76 ± 2.74	16.73 ± 2.89	23.62 ± 3.58	< 0.001
SPA	25.47 ± 1.73	18.38 ± 1.93	26.42 ± 2.84	< 0.001	28.12 ± 1.76	21.27 ± 1.91	29.30 ± 2.71	< 0.001
CT value (HU)
MPA	407.75 ± 92.57	407.79 ± 97.55	589.30 ± 139.77	< 0.001	351.68 ± 88.44	349.01 ± 83.53	509.31 ± 132.75	< 0.001
SPA	383.36 ± 57.25	384.31 ± 60.11	562.76 ± 80.79	< 0.001	334.51 ± 37.00	335.29 ± 37.82	480.78 ± 56.82	< 0.001
SNR
MPA	18.27 ± 5.33	25.21 ± 6.81	26.84 ± 6.75	< 0.001	14.61 ± 4.48	20.13 ± 5.50	21.22 ± 6.94	< 0.001
SPA	15.08 ± 2.35	20.98 ± 3.57	22.34 ± 3.28	< 0.001	11.91 ± 1.38	15.78 ± 1.81	16.42 ± 1.97	< 0.001
CNR
MPA	15.94 ± 5.10	21.07 ± 9.54	22.95 ± 8.02	< 0.001	11.68 ± 4.21	15.75 ± 5.99	17.15 ± 6.59	< 0.001
SPA	13.84 ± 2.38	19.83 ± 3.88	21.20 ± 3.89	< 0.001	10.50 ± 1.46	14.53 ± 1.99	15.18 ± 2.36	< 0.001

MPA main pulmonary artery, SPA segmental pulmonary artery, SNR signal-to-noise ratio, CNR contrast-to-noise ratio, BMI body mass index, HIR hybrid iterative reconstruction, MBIR model-based iterative reconstruction, CE-boost contrast enhancement boost

Fig. 1.

Bar plots show the CT value (a), image noise (b), signal-to-noise (SNR; c), and contrast-to-noise (CNR; d) in MPA and SPA of all patients. The image acquisition techniques include HIR, MBIR, and CE-boost. MPA, main pulmonary artery; SPA, segmental pulmonary artery. **: P < 0.001; *: P < 0.05; ns: P ≥ 0.05

Fig. 2.

Axial CTPA images (a-c) and corresponded volume-rendered (VR) images (d-f) of a 39-year-old women with a BMI of 30. Reconstruction was with HIR (a, c) and MBIR (b). Compared with the reference HIR image, both MBIR and CE-boost images demonstrate equal or even better image quality of pulmonary trunks for obese patients. In VR images, MBIR and CE-boost show more visualization of distal blood vessels (white arrows)

Subjective image quality analysis

Figures 3 and 4 show a stacked bar graph and image examples of observer ratings for each subjective criterion for both groups, respectively. Both MBIR and CE-boost images obtained higher scores than HIR images in the two groups (all p < 0.001). In terms of diagnostic confidence, group 2 CE-boost images were superior to group 2 HIR and MBIR images (all p < 0.05). MBIR images revealed the highest subjective image noise scores among the three datasets (all p < 0.001), while there was no statistical difference between HIR and CE-boost (p = 0.648). Compared to HIR and MBIR images in both groups, CE-boost images showed a significantly greater improvement in vascular contrast scores (all p < 0.05). The inter-observer agreement of subjective evaluation was good, with kappa values of 0.770 for diagnostic confidence, 0.734 for subjective image noise, and 0.790 for vascular contrast (Table 3). P-values for non-pairwise comparisons of subjective image quality were shown in Supplementary Table S3.

Fig. 3.

Stacked bar graph exhibiting observer ratings for each subjective criterion. The image acquisition techniques include HIR, MBIR, and CE-boost. In terms of diagnostic confidence and vascular contrast, CE-boost images had significantly better visual scores than HIR and MBIR images

Fig. 4.

Axial CTPA images (A-E) and corresponded VR images (F-J) illustrate examples of the scoring criteria for the vascular contrast. From left to right: Score 5 (Excellent), clear visualization extending to the distal subsegmental branches; Score 4 (Good), clearly visible up to the subsegmental branches; Score 3 (Acceptable),  indistinct visualization, identifiable up to the segmental branches; Score 2 (Blurry), limited visualization to the lobar branches; Score 1 (Poor), only the main trunk and left and right arteries are identifiable

Table 3.

The subjective image quality assessment

Parameters	HIR	MBIR	CE-boost	P value				Kappa value
				Overall	HIR vs. MBIR	HIR vs. CE-boost	MBIR vs. CE-boost
Diagnostic confidence								0.770
All patients	3.47±0.72	4.07±0.78	4.53±0.52	<0.001	<0.001	<0.001	0.002
Low BMI (group 1)	3.68±0.65	4.22±0.89	4.64±0.49	<0.001	<0.001	<0.001	0.009
High BMI (group 2)	3.26±0.71	3.92±0.63	4.42±0.54	<0.001	0.005	<0.001	<0.001
Subjective image noise								0.734
All patients	3.63±0.80	4.49±0.54	3.68±0.76	<0.001	<0.001	0.648	<0.001
Low BMI (group 1)	3.88±0.72	4.66±0.48	3.76±0.73	<0.001	<0.001	0.660	<0.001
High BMI (group 2)	3.34±0.77	4.32±0.55	3.60±0.81	<0.001	0.002	0.372	<0.001
Vascular contrast								0.790
All patients	3.51±0.85	4.22±0.56	4.69±0.51	<0.001	<0.001	<0.001	<0.001
Low BMI (group 1)	3.74±0.75	4.30±0.58	4.78±0.42	<0.001	0.024	<0.001	0.003
High BMI (group 2)	3.28±0.88	4.14±0.54	4.60±0.57	<0.001	0.032	<0.001	0.008

BMI Body mass index, HIR Hybrid iterative reconstruction, MBIR Model-based iterative reconstruction, CE-boost Contrast enhancement boost

Radiation dose

The mean CTDI_vol, DLP, and ED for both non-contrast and contrast-enhanced phases of group 1 were 3.65 ± 1.15 mGy, 137.53 ± 39.80 mGy*cm, and 1.93 ± 0.56 mSv, respectively; they were 5.19 ± 1.78 mGy, 191.25 ± 69.83 mGy*cm, and 2.68 ± 0.97 mSv in group 2.

Discussion

In this study, we investigated the clinical application of CE-boost in pulmonary vascular abnormalities for both non-obese and obese patients, and explored how its effect differed from iterative reconstruction algorithms. Compared to conventional HIR, our results demonstrated that both MBIR and CE-boost resulted in an enhanced improvement in the image quality of CTPA for obese patients, although the image quality of obese patients was inferior to that of normal-weight patients. The CE-boost images demonstrated a higher degree of contrast enhancement in the pulmonary trunk and its branches, as well as a higher subjective score for diagnostic confidence and vascular contrast. On the other hand, our results also showed that MBIR images had the lowest image noise among the three datasets even in obese patients.

Previous studies have indicated that the CE-boost technique can improve peripheral vascular visualization in CT angiography without increasing the contrast dose [13]. However, to our knowledge, no studies have compared the influence of different reconstruction algorithms with CE-boost on the image quality of CTPA in obese patients. The prevalence of obesity among patients is rising steadily, making it imperative to conduct research on this population.Theoretically, a sufficient amount of contrast agent was used to evaluate the presence of abnormal vascular structures during chest CT angiography, while obese patients had a lower contrast enhancement due to reduced iodine concentration in the blood. In this study, the objective results showed that the CT values, SNRs, and CNRs of pulmonary arteries of HIR images in group 1 were significantly lower than those in group 2 HIR images. Besides, accurate diagnostic of subsegmental pulmonary embolism also represents essential value for patients with cardiopulmonary limitation and pulmonary hypertension [17]. Missed subsegmental emboli not only cause hemodynamic abnormalities leading to chronic pulmonary embolism but can even cause thrombus spread to form large areas of PE [18, 19]. Nevertheless, excessive image noise in large patients will affect the observation of micro-emboli in the distal pulmonary artery and small lesions in the peripheral lung field [19–21]. The image noise was reduced in MBIR images, and the CT value of CE-boost images was significantly improved, which can clearly show the vascular structure of MPA and SPA regardless of normal weight or obese patients. In the subsequent application of clinical research, patients with embolism will be collected to study the display of embolism in distal small pulmonary vessels, and the accuracy of pulmonary artery embolism will be improved, which will be helpful for the clinical diagnosis of pulmonary embolism.

CE-boost is a newly introduced post-processing technique designed to enhance the brightness of blood vessels. The CE-boost technique enhances the signal intensity of iodine by fusing the iodine distribution map with the enhanced image, thereby improving the overall image quality and making the poorly enhanced vascular branches, peripheral arteries, and veins appeared more clearly [22], which has been applied in portal vein, aortic, and pulmonary artery. Routine CTPA typically involves two scans: the first is a non-contrast scan, which clearly visualizes calcifications in the vessel walls and high-density lesions in the lungs. In some cases, pulmonary embolisms in patients who have undergone cement-filled surgery may also appear as high-density. The second scan is contrast-enhanced, allowing for a comparison of lesion density changes between the two scans, thus providing a more comprehensive diagnosis. By adding contrast-enhanced images to the original CTPA images, our study demonstrated that the vascular attenuation of CE-boost images in obese patients was even superior to that of HIR images in non-obese patients. Unexpectedly, both HIR and MBIR images revealed lower image noise than CE-boost images. Otgonbaatar et al. suggested that the slightly higher noise may be attributed to the denoising filter during image registration [13]. Nevertheless, we believe that CE-boost technique surpasses HIR and MBIR algorithms in terms of subjective and objective image quality without increasing the amount of contrast agent, which potentially offers greater benefits to obese patients. Since CE-boost technology requires less computing power, it can be theoretically adopted by many CT vendors. Therefore, the clinical potential of the CE-boost procedure for the evaluation of small vessel structures is promising.

According to Wu et al., MBIR have greater potential to reduce noise, especially for the display of small blood vessels and subtle structures [8]. In our study, MBIR images showed lower vascular attenuation and image noise in the pulmonary artery than HIR and CE-boost images. No significant difference was found in SNR between MBIR and CE-boost images of SPA. In terms of subsegmental arteries, radiologists believed that a lower noise of MBIR image was also observed in obese patients. However, their time-consuming hinders the clinical application of MBIR in daily practice. It requires more computational power than other techniques and specific hardware, which is only available in the most advanced CT providers. For these reasons, the use of MBIR algorithm is not possible in many institutions that are not equipped with specific scanners [23]. Since CE-boost technology requires less computing power and has a wide range of applications. Therefore, the clinical potential of CE-boost technology in the evaluation of small vessel structures is significant.

This study has some limitations. First, this is a prospective single-center study with a relatively small sample size. Second, all CT images in this study were reconstructed with HIR or MBIR, the latest deep learning-based CT image reconstruction algorithm (e.g., Advanced Intelligent Clear-IQ Engine [AiCE]) was not included in the analysis. Third, the performance of the CE-boost technique in low-dose scanning protocols also needs further evaluation. Forth, the embolic diagnoistic of small distal pulmonary vessels was not studied. We concerned about that the challenging nature of accurately measuring small-diameter tubes, which could lead to data distortion and potential inaccuracies in the results.

In summary, in obese patients, both CE-boost and MBIR images yielded an increased image quality than conventional HIR images. CE-boost can improve the contrast enhancement of pulmonary arteries without raising the amount of iodine delivery. Instead, MBIR shows better noise reduction in the distal branches of the pulmonary artery. These two approches enable clinicians to provide a guarantee for the diagnosis of obese patients with clinically suspected pulmonary embolism, while CE-boost was outperformed when considering the processing speed of CT images and the convenience of clinical application.

Abbreviations

MBIR

Model-based iterative reconstruction

CE-boost

Contrast-enhancement-boost

CTPA

Computed tomography pulmonary angiography

HIR

hybrid iterative reconstruction

BMI

Body mass index

SNR

Signal-to-noise ratio

CNR

Contrast-to-noise ratio

PE

Pulmonary embolism

AIDR 3D

Adaptive Iterative Dose Reduction 3-Dimensinal

MPA

Main Pulmonary Artery

SPA

Segmental Pulmonary Artery

ROI

Region of interest

CTDIvol

Volume CT dose index

DLP

Dose length product

ED

Effective dose

Authors' contributions

A.Conception and design: Mei Ye, Li Wang, Yan Xing, Yuxiang Li, Zicheng Zhao, Min XuB.Administrative support: Wenya Liu, Yan XingC. Provision of study materials or patients: Mei Ye, Li Wang, Min Xu, Yuxiang LiD.Collection and assembly of data: Mei Ye, Li Wang, Yan Xing, Zicheng ZhaoE.Data analysis and interpretation: Mei Ye, Zicheng ZhaoF.Manuscript writing: All authorsG.Final approval of manuscript: All authors.

Funding

Not Funding.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This prospective study was approved by our Research Ethics Committee (Medical Ethics Committee of the First Affiliated Hospital of Xinjiang Medical University) and written informed consent was obtained (No. K202310-11).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.