We conducted a retrospective review of all patients referred to the Morgan Stanley Children's Hospital of New York Presbyterian for clinical CMR from May 2008 to December 2008. CMR data from patients who had examinations performed with PC images using phantom correction were analyzed. The decision to perform PC images was based on clinical and technical considerations for each patient

To assess the effect of phantom correction on pediatric patients with structurally normal hearts, a subset of patients with no shunts and no valvular regurgitation was identified. To assess the impact of a given diagnosis on the magnitude of phantom correction, our largest single diagnostic group, TOF, was analyzed separately.

Examinations were performed on two identical GE TwinSpeed Signa HDx 1.5T v. 14 scanners using commercially available coils (GE Healthcare, Milwaukee, Wisconsin). Correction for concomitant gradient effect, built into the image reconstruction, was used throughout. Non-breath-hold PC images were acquired perpendicular to the vessel of interest using orthogonal long-axis views of the vessel with the commercially resident FastCine PC pulse sequence. The coil and patient were positioned so that the heart was at isocenter, and the couch move to z = 0 facility was used. Vessels were imaged near isocenter. The following parameters were used: the TwinSpeed gradient system was in the high speed mode, with small fields of view (FOV) of 25-35 cm, matrix 256 × 192, time per cine frame 33 msec, slice thickness 5 mm, TR 12.2 msec, TE min full (4.5 msec), flip angle 25°, bandwidth 15.63 Hz, and velocity encoded (VENC) 200-400 cm/s, based on clinical parameters. No additional turbo factor was used. The number of views per segment varied between 2 and 8 based on heart rate. The acquisition length was usually 90-120 seconds, with NEX = 3. Retrospective ECG gating was performed with continual gradient operation. Flow compensation was used throughout. Immediately after the patient was removed from the scanner, a stationary fluid filled phantom was moved into position, and in approximately one to two minutes the images were repeated using the same clinical parameters to establish a baseline of zero velocity. The phantom data was acquired within 15 to 20 minutes of obtaining the PC flow imaging. The size of the phantom was chosen based on patient size, and a heart rate simulator was set to 60 or 120 beats per minute. As imaging was acquired using retrospective ECG gating, the background offset was likely to be stable during cine frames, and setting the HR simulator at 60 vs. 120 beats per minute was unlikely to affect the velocity offset.

Based on the clinical protocol, PC images were obtained in the AAO, MPA, RPA, and/or LPA. PC images were analyzed using GE ReportCARD^© 3.6 both with and without phantom correction (Figures 1, 2, 3). A region of interest (ROI) was drawn for each vessel, and net flow rate (ml/min) was calculated for each. Flows were corrected in the AAO, MPA, RPA, and LPA by applying the corresponding baseline shift that zeroed the flow in the phantom using the reporting software. Subtraction was done based on an average baseline rather than on an image-by-image basis, to minimize signal-to-noise ratio loss. The Q_P/Q_S, percent flow to the RPA (Q_PR fraction), pulmonary regurgitation fraction (PR), and aortic regurgitation (AR) fraction were also calculated.

Changes in flow measurements with phantom correction were defined prior to analysis as follows, as agreed upon by consensus of two pediatric cardiologists experienced in CMR (WL, BP): (Table 1): a change in MPA or AAO flow ≥ 0.5 L/min/m², change in RPA or LPA flow ≥ 0.25 L/min/m², change in Q_P/Q_S ≥ 0.4, change in Q_PR fraction ≥ 10%, and change in PR or AR fraction ≥ 10%. These definitions were based on generally accepted normal ranges for the measured and calculated values. For example, a Q_P/Q_S cutoff of 0.4 was chosen as a clinically significant change because a Q_P/Q_S of 1.0 is generally considered to be normal while a value of 1.4 or greater is considered a potentially significant shunt. Marked changes in flow measurements for the purpose of this analysis were defined as double the amount of clinically significant change in each category--for example, a change in MPA flow ≥ 1.0 L/min/m² or PR fraction ≥ 20%.

Statistical analysis was performed using SPSS 16.0 (SPSS Inc., Chicago, Illinois) and Microsoft Excel 2007 (Microsoft Corporation, Redmond, Washington). The relationship between blood vessel cross sectional area and change in PC flow measurement, and the relationship between heart rate and change in PC flow measurement, were both analyzed using linear regression. For this group, the absolute difference between the measured Q_P/Q_S and the assumed true Q_P/Q_S of 1.0 were compared with and without phantom correction using paired t-tests. To assess the impact of a given diagnosis on the magnitude of phantom correction, our largest single diagnostic group, tetralogy of Fallot, was analyzed separately. A chi-squared test was used to compare the subgroup with tetralogy of Fallot with the entire group of analyzed patients, and to compare the measurements that were increased or decreased following phantom correction. A p value of < 0.05 was considered statistically significant.

There were 149 patients identified who had clinical CMR examinations with PC imaging using phantom correction during the study period. From this patient population, 144 patients had PC images of the ascending aorta (AAO), 118 patients had PC images of the main pulmonary artery (MPA), 70 patients had PC images of the right pulmonary artery (RPA), and 61 patients had PC images of the left pulmonary artery (LPA). The patient characteristics and diagnoses are provided in Table 2.

The median effects of phantom correction, presented as absolute change from baseline, are shown in Table 3. Phantom correction resulted in clinically significant changes in 13% to 48% of overall PC measurements in patients with known or suspected heart disease (Table 4). This range of changes is across all measured and calculated quantities. For example, the aortic regurgitation fractions showed the smallest changes, with 13% of measurements changing by > 10% with phantom correction, and the left pulmonary artery flow showed the greatest changes, with 48% of measurements changing by > 0.25 L/min/m² with phantom correction. Overall, 640 measurements or calculated values were analyzed, and clinically significant changes were found in 197 (31%). As discussed above, cutoff values were assigned prior to analysis to assess how often phantom correction resulted in changes that would have significance on clinical decision-making. There were marked changes in up to 25% of PC measurements (Table 4). Clinically significant or marked changes were generally more common in measurements of output (L/min/m²) as compared to regurgitant fractions or flow ratios.

Flow measurements in our study increased with phantom correction in some instances and decreased in others (Table 5). AAO flow increased in a significant majority of patients, while LPA flow decreased. Calculated Q_P/Q_S ratio and AR fraction decreased more often than increased. For MPA flow, PR fraction, RPA flow, and Q_PR, there was no significant difference in the direction of change.

There was no significant difference in the effect of phantom correction for any of the measured flows or calculated values for the sub-group of patients with tetralogy of Fallot as compared to the group of analyzed patients as a whole (Table 6). Clinically significant changes in patients with tetralogy of Fallot were present in up to 53% of PC measurements and marked changes in up to 21%.

Within our patient population, 16 of the 149 patients had structurally normal hearts, based on no shunt lesion and no valvular regurgitation by both echocardiography and CMR. In these patients, Q_P/Q_S after phantom correction (1.0 ± 0.1) was closer to 1.0 than before phantom correction (1.1 ± 0.2) (Figure 4, p = 0.016).

To assess the effect of blood vessel cross sectional area on the changes in flow measurements with phantom correction, we created a scatter plot of all 392 magnitude flow measurements (AAO, MPA, RPA, and LPA) performed in all patients. There was a small but statistically significant relation between vessel cross sectional area and change in PC flow measurement with phantom correction. Linear regression analysis found that 22% of the changes in PC flow measurements could be attributed to the cross sectional area of the vessel being measured (Figure 5, p < 0.001).

Due to the possibility that a longer R-R interval may allow more time for the velocity offset to accumulate into the flow error, we assessed the effect of heart rate on the changes in flow measurements with phantom correction. Scatterplots were created comparing change in flow measurements in the AAO and MPA to the patient's heart rate during the scan. Using linear regression analysis, we found that heart rate did not contribute significantly to the changes in PC flow measurements in the AAO (R² = 0.018, p = NS) or MPA (R²< 0.001, p = NS).