Evan A Nelson^{1}, Li-Yueh Hsu^{1}, Anna Noreuil^{1}, Sara E Berger^{2,3}, Marcus Y Chen^{1}, W. Patricia Bandettini^{1}, Sujata M Shanbhag^{1}, Shahryar G Saba^{4}, Giancarlo Serafini^{1}, Christine Mancini^{1}, Vandana Sachdev^{1}, and Andrew E Arai^{1}

The purpose of this study was to investigate the magnitude of ex vivo static phantom corrections affecting Qp, Qs, and the Qp-to-Qs ratio measurements in diagnosing patients with intracardiac shunts. In the patients with Cath as the reference standard, phantom correction improved both the bias as well as the limits of agreement for Qp and the Qp:Qs ratio. For Qs, phantom correction improved the bias but did not significantly change the limits of agreement. While phantom correction appears to improve phase contrast MRI measurements, the main benefit is limited to patients with Qp:Qs close to the important diagnostic threshold selected.

In the patients with Cath as the reference standard, phantom correction improved both the bias as well as the limits of agreement for Qp and the Qp:Qs ratio (Table 1). For Qs, phantom correction improved the bias but did not significantly change the limits of agreement (Table 1).

For the entire cohort of patients, the median uncorrected CMR Qs was 4.5 L/min [4.0, 5.5] and was significantly less than the corrected CMR Qs 5.0 l/min [4.3, 5.9] (p<0.001). The median uncorrected CMR Qp was 5.3 L/min [4.5, 6.7] and was significantly less than the corrected CMR Qp 5.5 l/min [4.7, 6.8] (p<0.001). The median phantom correction for Qs was -0.38 l/min [-0.57, -0.16] and was larger than the correction for Qp -0.27 l/min [-0.46, 0.02] (p<0.001). The median percent correction for Qs was -8.1% [-12.3, -3.7] and was greater than the median percent correction for Qp -4.8 [-8.4, 0.48] (p<0.001). However the phantom corrections were as large as 49.7% for Qs and 47.2% for Qp. Negative phantom corrections were more common than positive corrections for both Qs (145 vs 20) and Qp (120 vs 44, Figure 1).

Clinical significance was assessed in the entire cohort as the number of patients re-categorized as having shunts requiring surgical correction using two commonly used thresholds for the Qp:Qs ratio (1.5 and 2.0). Thirteen patients had an uncorrected Qp:Qs ratio between 1.5 and 2.0, and 7 of these patients (53.8%) were re-classified to a non-surgical shunt fraction (i.e. <1.5) as a result of the phantom correction. Fourteen patients had an initial Qp:Qs ratio >2.0, and phantom correction reclassified 3 patients to shunt fractions between 1.5 and 2.0 (21.4%) but none were re-classified to <1.5. Conversely, none of the 138 patients with an uncorrected Qp:Qs ratio <1.5 were re-classified to a shunt fraction >1.5 after phantom correction.

Of the 27 patients with an uncorrected Qp:Qs >1.5, four patients did not have an anatomic intracardiac shunt. After phantom correction, all 20 patients with a Qp:Qs >1.5 had an identifiable anatomic intracardiac shunt. The phantom correction also successfully reclassified those 4 patients that had no anatomic intracardiac shunt to a Qp:Qs ratio <1.5.

Factors favoring implementation of phantom corrections include the following: The magnitude of phantom correction was large relative to cardiac output in some patients. Phantom correction improved bias and limits of agreement with Cath as a reference standard. Phantom correction eliminated the misclassification of 4 patients who had no identifiable intracardiac shunt.

Factors against phantom correction include the following: Out of 165 patients, phantom correction only re-classified 3 patients at an uncorrected Qp:Qs threshold of 2.0 and 7 patients at a threshold of 1.5. No patient that had an uncorrected Qp:Qs ratio <1.5 were up-classified to a surgically significant shunt. The phantom correction process increases scanner time as well as the time to process and analyze studies. These extra steps appear to be of limited benefit in the patients that were below a Qp:Qs ratio of <1.5 who were being assessed for shunt fraction.