Optimal CT Contrast Enhancement Protocol for the Fontan Pathway: An Analysis Based on a 15-year Single-center Experience

This retrospective study was performed to systemically compare several candidates for optimal contrast enhancement protocols of cardiothoracic CT dedicated for evaluating the Fontan pathway. Of 115 CT examinations from 89 patients, simultaneous injection of contrast agent via the arm and leg veins with 50% diluted contrast agent (group 1, n = 38), 60-second scan delay after leg vein injection (group 2, n = 41) or 3-minute scan delay (group 3, n = 36) was used to obtain optimal contrast enhancement. The degree and heterogeneity of cardiovascular enhancement, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were quantitatively evaluated. Histogram-assisted semi-quantitative evaluation was performed for heterogeneous enhancement, and a cut-off value indicating heterogeneous enhancement was determined by comparing the standard deviations between the cases showing homogeneous and heterogeneous enhancement. Contrast enhancement of the Fontan pathway, the standard deviation measured in the Fontan pathway, SNR, and CNR were more frequently lower in group 3 compared to groups 1 and 2 (p < 0.001). Homogeneous enhancement of the Fontan pathway based on the histogram-assisted semi-quantitative evaluation was more frequently seen in group 3 compared to groups 1 and 2 (p < 0.043 ‒ 0.001). Receiver operating characteristic curve analysis demonstrated that the standard deviation was an excellent diagnostic test in determining the homogeneity of contrast enhancement (area under the curve = 0.991; 95% condence interval, 0.988 ‒ 0.994; p < 0.001). Standard deviations > 105.1 HU could be considered to indicate heterogeneous enhancement with 73.2% sensitivity and 100.0% specicity. The protocol using a 3-minute scan delay best achieved homogeneous contrast enhancement in the Fontan pathway on cardiothoracic CT among the tested protocols. However, low contrast enhancement, SNR, and, CNR need to be improved.


Introduction
Cardiothoracic CT is frequently requested for identifying stenosis and thromboembolism in the Fontan pathway [1]. It is also often used as an alternative imaging modality in cases with contraindications to cardiac MRI, including MR-incompatible cardiac pacemakers and de brillators and claustrophobia, and severe artifacts from metallic implants [2,3]. However, it is quite challenging to achieve homogeneous contrast enhancement in the Fontan pathway on cardiothoracic CT due to turbulent and preferential nonpulsatile ow as clearly demonstrated in time-resolved MRA [4] and four-dimensional ow MRI [5]. The typical contrast enhancement protocols used to evaluate congenital heart disease almost always result in non-uniform, non-diagnostic contrast enhancement on cardiothoracic CT; therefore, additional delayed scanning is often required in patients with the Fontan pathway. Consequently, several contrast enhancement protocols dedicated to evaluate the Fontan pathway on cardiothoracic CT have been suggested [6][7][8][9][10][11][12]. An expert consensus document recommended a venous two-phase or split injection method, in which 30-50% of the contrast agent is given and followed by a 30-60 seconds pause and then by the second injection of contrast agent [13]. Another expert consensus guideline recommended a 60-70-second delayed scanning because the method appears to offer balanced opaci cation of all thoracic cardiovascular structures and requires a simple injection technique [14]. However, systematic comparisons between the dedicated contrast enhancement protocols for cardiothoracic CT after the Fontan operation have been seldom conducted to determine the best contrast enhancement protocol.
Contrast enhancement protocols dedicated to evaluate the Fontan pathway have also been developed and used in our institution since 2006. Therefore, this retrospective study was performed to systematically compare several candidates for optimal contrast enhancement protocols of cardiothoracic CT dedicated to evaluate the Fontan pathway performed for 15 years in our institution and suggest the best alternative.

Materials And Methods
This retrospective study was approved and the requirement for informed consent was waived by the local Institutional Review Board.

Patient Population
For 15 years since 2006, 97 consecutive patients underwent cardiothoracic CT after the Fontan operation using one of three dedicated contrast enhancement protocols in our institution. Eight patients were excluded from this study because they had the following reasons substantially affecting cardiovascular enhancement of the Fontan pathway: inferior vena cava (IVC) interruption with azygos continuation (n = 5); failed intravenous injection of contrast agent (n = 2); and the use of a central venous catheter placed in the superior vena cava (SVC) for injection (n = 1). Therefore, 89 patients were included in this study; a single CT scan was performed in 63 patients; early-(60-second delay) and late-(3-minute delay) phase scans were performed in 26 patients. According to the contrast enhancement protocols, 115 CT examinations from 89 patients were divided into three groups as follows: group 1, simultaneous injection of the contrast agent via the arm and leg veins with 50% diluted contrast agent in 38 examinations; group 2, a 60-second scan delay after leg vein injection at slow injection rates for 55 seconds and stopping the injection 5 seconds before starting the scan in 41 examinations; and group 3, a 3-minute scan delay irrespective of the intravenous injection sites at slow injection rates for 55 seconds in 36 examinations.
The demographics and clinical characteristics of the study population are described in Table 1.

Cardiothoracic CT
For the 15 years, cardiothoracic CT examinations were performed using one of three different CT scanners as follows: a 16-slice CT scanner (SOMATOM Sensation 16; Siemens Healthineers), a rstgeneration dual-source CT scanner (SOMATOM De nition; Siemens Healthineers), and a secondgeneration dual-source CT scanner (SOMATOM De nition Flash; Siemens Healthineers); and one of which was assigned as a cardiovascular CT scanner in our institution depending on the time period. The scanning technique was determined based on the clinical requests and patients' conditions [15]. Body size-adapted CT scan protocols [15][16][17] and an attenuation-based tube current modulation (CARE Dose 4D; Siemens Healthineers) [18,19] were used to optimize the radiation dose. Electrocardiography (ECG)controlled tube current modulation (MinDose; Siemens Healthineers) and the biphasic chest pain protocol were used to further reduce the radiation dose of the retrospectively ECG-gated spiral scan [20]. When available, an iterative reconstruction algorithm was used to reduce the image noise while maintaining the image details [21][22][23][24]. The scan range included the whole thorax in all patients. Patients younger than 6 years of age were initially sedated with oral chloral hydrate (50 mg/kg) and subsequently with intravenous midazolam (0.1 mg/kg) or ketamine (1 mg/kg) as required. The CT scan was performed during free breathing in these sedated patients or while the conscious and cooperative patients were holding their breath. The presence of metallic implants, such as cardiac pacemakers and vascular coils, on the CT scout images was recorded.
To maximize the radiation dose e ciency and the iodine contrast-to-noise ratio (CNR), the lowest possible tube voltage was selected ( Table 2). The volume CT dose index and dose-length product values based on a 32-cm phantom were recorded. The effective dose estimate was calculated by multiplying the dose-length product and the age, sex, and tube voltage-speci c conversion factors for the chest CT [16].
The detailed CT parameters are described in Table 2.
Iodinated contrast agent (Iopamidol; Pamiray-370, Dongkook Pharmaceutical; 2.0 mL/kg) was administered intravenously with a dual-head power injector at an injection rate of 0.5 -2.5 mL/s. The scan delay time was determined by a bolus tracking technique in 17 of 38 CT examinations of group 1, while a xed delay of either 60 seconds or 3 minutes was used for the remaining CT examinations.

Quantitative Evaluation of the Image Quality
To quantitatively evaluate the degree and heterogeneity of cardiovascular enhancement, image noise, signal-to-noise ratio (SNR), and CNR, a biggest possible region of interest was placed in the right and left internal jugular veins, right and left SVC, right and left pulmonary arteries, inferior cavopulmonary connection, IVC, ascending aorta, ventricle, atrium, ventricular myocardium, and air. When a vessel was not visualized or contained a vascular catheter, CT densitometry could not be performed and these vessels were treated as having missing values. Mean CT density and standard deviation were measured for each location.
Contrast enhancement < 200 Houns eld units (HU) was regarded to be suboptimal. The standard deviation of the measured air density was regarded as the image noise because it is less affected by different tube voltages. Image noise ≥ 10 HU was regarded to be suboptimal. SNR was calculated using the following formula: SNR = mean density of the ventricle/image noise. CNR was calculated using the following formula: CNR = (mean density of the ventricle -mean density of the ventricular myocardium)/image noise. SNR < 35 and CNR < 20 were regarded to be suboptimal.

Semi-quantitative Evaluation of Heterogeneous Enhancement
There is no established objective criterion indicating heterogeneous enhancement, and the subjective evaluation of heterogeneous enhancement is considerably in uenced by the window settings of the CT images ( Fig. 1). Therefore, the evaluation in this study was assisted by the shape of the histogram of the obtained CT densitometry. When a histogram showed a single peak with a narrow base, the contrast enhancement was regarded to be homogeneous (Fig. 2). In contrast, contrast enhancement was regarded to be heterogeneous when a histogram demonstrated a broad-based peak (Figs. 3, 4). The standard deviations were compared between the cases showing homogeneous and heterogeneous contrast enhancement to determine a cut-off value indicating heterogeneous contrast enhancement.

Statistical Analysis
Continuous variables are presented as mean ± standard deviation or median with range, and categorical variables are expressed as frequencies with percentages. For the comparisons of the continuous variables between the groups, an unpaired t-test was used. A chi-square test or Fisher's exact test was used for the comparisons of the categorical variables between the groups, depending on whether the expected cell counts were below ve or not. Receiver operating characteristic (ROC) curve analysis was used to assess the diagnostic ability of a binary classi er system (heterogeneous or homogeneous contrast enhancement). A p-value less than 0.05 was considered statistically signi cant. Statistical analyses were performed using statistical software (SPSS version 24.0; IBM Corp., Armonk, NY, USA).

Patient Population
No signi cant differences were found in the age and sex ratio between the groups (p > 0.365-0.728 and p > 0.081, respectively) ( Table 1). Among the three modi cations of the Fontan operation, the extracardiac conduit type was the most common in the range of 84.2-97.6% in all three groups (Table 1).
Metallic implants producing severe beam hardening artifacts were frequently present in the range of 16.7-22.2% in the three groups (Table 1).

Cardiothoracic CT
The non-ECG-synchronized spiral scan was most frequently performed in the initial period (2006-2014) of this study, while the ECG-synchronized scans prevailed in the later period (Table 2). Similarly, lower tube voltages including 70 kV and 80 kV were more frequently selected in a later period of this study ( Table 2). The dose-length product and effective dose in group 3 were signi cantly lower than those in group 1 (p < 0.046 and p < 0.012, respectively), whereas radiation dose metrics showed no signi cant differences in other comparisons (p > 0.060-0.522) ( Table 2).

Quantitative Evaluation of the Image Quality
The results of the quantitative evaluation of the image quality are described in Table 3. The mean densities in group 3 were signi cantly lower than those in groups 1 and 2 in all the cardiovascular structures (p < 0.001-0.005), except for the left SVC in which the mean density in group 1 was signi cantly higher than those in groups 2 and 3 (p < 0.005 and p < 0.043, respectively) ( Table 3). The mean densities of the IVC and inferior cavopulmonary connection in group 2 were signi cantly higher than those in groups 1 and 3 (p < 0.001) ( Table 3). In contrast, the mean density of the myocardium showed no signi cant differences among the three groups (p > 0.212-934) ( Table 3). Suboptimal contrast enhancement (< 200 HU) was more frequently seen in the IVC, pulmonary arteries, ascending aorta, ventricle, and atrium in group 3 than in groups 1 and 2 (p < 0.001-0.048) ( Table 3).
The standard deviations of the CT densitometries in group 3 were signi cantly lower than those in groups 1 and 2 in the internal jugular veins, vena cavae, inferior cavopulmonary connection, and pulmonary arteries (p < 0.001-0.016) ( Table 3). In the ventricle, the standard deviation of CT densitometry was signi cantly lower in group 1 than those in groups 2 and 3 (p < 0.001 and 0.017, respectively) ( Table 3). In the atrium, the standard deviation of CT densitometry was signi cantly higher in group 2 than those in groups 1 and 3 (p < 0.001) ( Table 3). In contrast, the standard deviations of CT densitometries measured in the ascending aorta and myocardium showed no signi cant differences among the three groups (p > 0.063-0.473 and p > 0.128-0.612) ( Table 3).
Image noise measured in the air in group 1 was signi cantly lower than those in groups 2 and 3 (p < 0.017 and 0.001, respectively), while image noises were comparable between groups 2 and 3 (p > 0.141) ( Table 3). In all three groups, suboptimal image noise (≥ 10 HU) was rarely observed in the range of 0.0-7.3% without signi cant differences in frequency (p > 0.233) ( Table 3). Notably, SNR and CNR in group 3 were signi cantly lower than those in groups 1 and 2 (p < 0.001), while they were comparable between groups 1 and 2 (p > 0.196 and p > 0.114, respectively) ( Table 3). The proportions of suboptimal SNR (< 35) and CNR (< 20) were also signi cantly higher in group 3 than in groups 1 and 2 (p < 0.001), while they were comparable between groups 1 and 2 (p > 0.676) ( Table 3).

Semi-quantitative Evaluation of Heterogeneous Enhancement
In the internal jugular veins, the homogeneities of contrast enhancement were signi cantly higher in group 3 than in group 1 (p < 0.006) ( Table 3), while groups 2 and 3 demonstrated comparable homogeneities (p > 0.244) ( Table 3). In the SVCs, the homogeneities of contrast enhancement were signi cantly different among the three groups (p < 0.043): the highest in group 3, followed by group 2 and then group 1 (Table 3). In the IVC, inferior cavopulmonary connection, and pulmonary arteries, the homogeneities of contrast enhancement were signi cantly higher in group 3 than in groups 1 and 2 (p < 0.001) ( Table 3), while groups 1 and 2 demonstrated comparable homogeneities (p > 0.086) ( Table 3). In the ascending aorta, ventricle, and atrium, there were no signi cant differences in the homogeneities of contrast enhancement among the three groups (p > 0.055) and they were 100% in all, except for the atrium in group 2 (87.8%) ( Table 3). Notably, group 3 frequently showed 100% homogeneity in all cardiovascular structures, except for the IVC (80.6%) and inferior cavopulmonary connection (91.7%) ( Table 3).
In the three groups, contrast enhancement was homogeneous in 863 cardiovascular structures and heterogeneous in 284 cardiovascular structures. The standard deviations in the cases showing heterogeneous enhancement were signi cantly higher than those in the cases showing homogeneous enhancement (196 ± 110.4 HU vs. 25.4 ± 16.9 HU, p < 0.001). ROC curve analysis demonstrated that the standard deviation measured in the cardiovascular structures was an excellent diagnostic test in determining homogeneity of contrast enhancement (area under the curve = 0.991; 95% con dence interval, 0.988-0.994; p < 0.001) (Fig. 5). With a cut-off value of 105.1 HU, higher standard deviations could be considered to indicate heterogeneous enhancement with 73.2% sensitivity and 100.0% speci city.

Discussion
Based on the quantitative and semi-quantitative results of this study, the 3-minute delayed scanning (group 3) demonstrated the most homogeneous contrast enhancement in the Fontan pathway compared to the other contrast enhancement protocols (groups 1 and 2). In contrast, contrast enhancement, SNR, and CNR were lowest in group 3 compared to groups 1 and 2. These ndings are similar to those reported in a previous study evaluating 88 CT Table 2).
In addition, the authors of the previous study [8] considered that the mean CT densities of 160 to 170 HU were in the subjectively acceptable range of contrast enhancement to evaluate the Fontan pathway, whereas the acceptable level (200 HU) of contrast enhancement was slightly higher in the current study. Even though the authors stated that all images using the 3-minute-delay scan achieved acceptable enhancement of all three vascular structures including the aorta, pulmonary artery, and Fontan tract, a substantial proportion of the CT images obtained at the 3-minute delay might show insu cient contrast enhancement based on the mean attenuations in the range of 160.1-166.2 HU and the standard deviations in the range of 25.1-42.3 HU, except for the aorta in cases using a foot vein root [8]. Moreover, the normal myocardium also showed contrast enhancement on delayed CT scans [25]. The mean myocardial attenuation measured on the delayed (1 minute to 3 minutes) CT scans was 142.1-151.8 HU (Table 3). Therefore, the CNR of the 3-minute-delay scans that was not calculated in the previous study [8] was expected to be insu ciently low for evaluating intracardiac morphology accurately. In the current study, 94.4% of the cases obtained using the 3-minute delay (group 3) showed suboptimal CNR (< 20) (Table 3).
Consequently, the CNR of the 3-minute-delay scans needs to be further improved. One way of increasing the CNR is by using a shorter scan delay, but the 60-second delay seemed to lead to inhomogeneous enhancement in the Fontan pathway based on the results of this study. Some researchers suggested a 70-second delay [1,12] and others suggested an 80-100 second delay [10]. However, it should be carefully considered that heterogeneity of contrast enhancement in the Fontan pathway seems to be increased proportional to the shortening of the scan delay. Therefore, further studies are warranted to determine how the scan delay may be shortened from 3 minutes without compromising homogeneous enhancement in the Fontan pathway. Another potential method of increasing the CNR of the 3-minutedelay scans is by using virtual monoenergetic image reconstruction at low keV using dual-energy CT scanning [26,27].
Objective evaluation of homogeneous contrast enhancement in the Fontan pathway is challenging. For example, visual evaluation is considerably affected by the window setting of the CT images as was demonstrated in Figure 1 of the current study. Nevertheless, subjective evaluation of homogeneous enhancement in the Fontan pathway has been frequently used [8,9]. To avoid visual perception errors, the standard deviation of the measured CT attenuation in the cardiovascular structures by placing a region of interest is infrequently used to represent a quantitative measure of homogeneity in contrast enhancement. Among seven different contrast enhancement protocols, the standard deviations of the aorta, pulmonary artery, and Fontan tract were less than 30 HU in the 1-minute-delay scan with both antecubital routes and the 3-minute-delay scan, irrespective of the intravenous injection sites [8]. Other protocols demonstrated higher standard deviations of the pulmonary artery and Fontan tract (up to 76 HU and 504 HU, respectively) [8]. However, the standard deviation (30 HU) used for the criteria of heterogeneous enhancement in their study [3] was rather arbitrary without any solid evidences. In addition, they used a xed size (30 mm 2 ) in the area of the regions of interest, irrespective of the sizes of the vessels, and therefore, they might have selected areas of relative homogeneous enhancement in the vessels for the measurements [8]. In contrast, the region of interest used in this study included the whole area of the vessels to avoid selection bias in the measurements. In addition, the patterns of the histogram of CT densitometry were used to minimize pitfalls of qualitative evaluation of homogeneous contrast enhancement solely depending on human eyes and substantial changes in enhancement pattern by changing window settings of the CT images. According to the ROC curve analysis, the standard deviation of 105.1 HU may be served as a useful cut-off value in determining homogeneity of contrast enhancement.
This study had several limitations. First, it is impossible to completely avoid selection bias in the patient population due to its retrospective nature. Second, the study population was relatively small.

Conclusions
The contrast enhancement protocol using a 3-minute scan delay is the best to achieve homogeneous contrast enhancement in the Fontan pathway on cardiothoracic CT among the tested protocols. However, low contrast enhancement, SNR, and CNR need to be improved.

Declarations
Compliance with ethical standards Con ict of interest All authors declare that they have no nancial or non-nancial con ict of interest.