## Background

The accuracy of the proximal isovelocity surface area (PISA) method for the quantification of mitral regurgitation (MR), in the case of multiple jets, is unknown. The aim of this study was to evaluate different two-dimensional (2D) and three-dimensional (3D) PISA methods using 3D color Doppler data sets.

## Methods

Several regurgitant volumes (Rvols) were simulated using a pulsatile pump connected to a phantom equipped with single and double regurgitant orifices of different sizes and interspaces. A flowmeter served as the reference method. Transthoracic (TTE) and transoesophageal echocardiography (TEE) were used to acquire the 3D data sets. Offline, Rvols were calculated by 2D PISA methods based on hemispheric and hemicylindric assumptions and by 3D integrated PISA.

## Results

A fusion of the PISA was observed in the setting of narrow-spaced regurgitant orifices; compared with flowmeter, Rvol was underestimated using the single hemispheric PISA model (TTE: Bland-Altman bias ± limit of agreement, −17.5 ± 8.9 mL; TEE: −15.9 ± 7.3 mL) and overestimated using the double hemispheric PISA model (TTE: +7.1 ± 14.6 mL; TEE: +10.4 ± 11.9 mL). The combined approach (hemisphere for single orifice, hemicylinder with two bases for nonfused PISAs, and hemicylinder with one base for fused PISAs) was more precise (TTE: −3.4 ± 6.3 mL; TEE: −1.9 ± 5.6 mL). Three-dimensional integrated PISA was the most accurate method to quantify Rvol (TTE: −2.1 ± 6.5 mL; TEE −3.2 ± 4.8 mL).

## Conclusions

In the setting of double MR orifices, the 2D combined approach and integrated 3D PISA appear to be superior as compared with the conventional hemispheric method, thus providing tools for the challenging quantification of MR with multiple jets.

The accurate assessment of mitral regurgitation (MR) directly influences the optimal timing for valve repair. The best prognosis after surgical treatment, for severe primary MR, is obtained at an asymptomatic stage without echocardiographic signs of left ventricular volume overload. The current guidelines recommend an integrated approach for the evaluation of the MR severity including, as the most robust two-dimensional (2D) color Doppler echocardiography (CDE) parameters, the regurgitant volume (Rvol) and the effective regurgitant area (EROA) obtained by the proximal isovelocity surface area (PISA) method, combined with a measurement of the vena contracta (VC). However, the optimal tools for the quantitative assessment of secondary MR remain to be defined, and MR evaluation is technically difficult in this context. Using 2D CDE, the recommended methods are based on geometric assumption of a circular orifice, which in the context of complex anatomical MR orifices often leads to inaccuracy. Moreover, there is to date no recommendation for quantifying MR with multiple jets, and the scientific literature on this topic is sparse. Since the development of three-dimensional (3D) CDE, several studies have shown the accuracy of direct PISA measurement in the assessment of MR. Brugger *et al.* recently validated a customized 3D PISA software and showed that the 3D PISA technique permits a more accurate MR assessment than conventional 2D PISA with cardiac magnetic resonance imaging (CMR) as the reference method. The aim of the present study was to evaluate different 2D and 3D PISA methods in the presence of multiple MR orifices, using 3D CDE data sets obtained by transthoracic (TTE) and transoesophageal (TEE) echocardiography, in vitro, against a flowmeter, in order to determine which simple geometric assumption could best estimate the PISA surface and to confirm the accuracy of the integrated 3D PISA method.

## Methods

## In Vitro Model

The in vitro model has been described elsewhere. The phantom was 3D printed using the plastic material WaterShed (DSM, Elgin, IL; Figure 1 ). It consisted of a “ventricle” and an “atrium” submerged in a commercially available “Doppler test fluid” solution (model 707, ATS Laboratories Inc., Bridgeport, CT). A pulsatile pump (ViVitro Systems Inc., Victoria, British Columbia, Canada), mimicking the cardiac cycle, was connected to the “ventricle,” using four connectors, to assure a symmetrical pressurization of the cavity (pump frequency at 60 beats per minute). Mitral valve models with one or two orifices, different interspaces between the orifices (10 or 20 mm), and different anatomic regurgitant orifice areas (AROA: 15, 30, 50, or 70 mm ^{2 }) were used ( Figure 2 ). All orifices were circular. A flowmeter (transit-time ultrasonic flow probe, Transonic Systems, Inc., Ithaca, NY) was placed on the tube connecting the ventricle to the pump. A second flowmeter was placed on the “outflow tract” of the ventricle, where a safety valve prevented pressure overload in the test bench ( Figure 1 ). The flow measured during one beat was then integrated in order to obtain the volume per beat. The difference between the volume passing through the entry tube and the volume passing through the exit tube gave the Rvol passing through the mitral valve, which served as reference method. Rvols acquired during five consecutive beats were averaged.

## Echocardiographic Acquisition

TTE (X5-1 transducer, frequency range 1–5 MHz, harmonic imaging mode activated, imaging depth 9 cm, on “ventricular” side) and TEE (X7-2t transducer, frequency range 2–7 MHz, harmonic imaging mode deactivated, imaging depth 6 cm, on “atrial” side) echocardiography (iE33 Phillips Healthcare, Zurich, Switzerland) were performed after setting of the same stroke amplitudes. In order to render the in vitro model similar to the human heart, the distance between the TEE probe and the “mitral” valve was two thirds of the distance between the TTE probe and the valve. After optimization of the depth and narrowing of the image sectors on two perpendicular planes, electrocardiogram-triggered (six beats) 3D volumes focused on MR-PISA were acquired during three cardiac cycles. The peak velocity ( *V *max) and velocity time integral (VTI) were obtained with continuous wave (CW) Doppler.

## Quantification of MR

Color Doppler pyramidal volumes were analyzed on orthogonal planes using QLAB (Koninklijke Philips Electronics, Eindhoven, the Netherlands). Usually, a Nyquist velocity between 20 and 40 cm/sec (range, 11–51 cm/sec) permitted us to obtain an optimal shape of the PISA, and the visually biggest PISA was chosen at midsystole. Three groups were defined according to the presence of one or two MR orifices and according to the presence or absence of a fusion of the PISA shapes: single orifice, double orifices with PISA fusion, and double orifices without PISA fusion. For 2D analysis, the plane was placed across the middle of the regurgitant orifice(s) and manually optimized in order to obtain the visually biggest PISA. The radius ( *r *) and the horizontal length ( *h *) of each PISA were measured ( Figure 3 A ). To define the base of the PISA, the leading edge on the atrial side of the “mitral valve” was chosen. Once each PISA was characterized with *r *, *h *, and Nyquist velocity, several geometric assumptions were tested in order to define the most accurate 2D method for each MR group. We looked for simple formulae, only requiring measurements obtainable on one plane. The PISA surface ( *S *) was calculated according to the hemispheric assumption ( *S *hemisphere = 2π *r *^{2 }). EROA was calculated as EROA = ( *S *hemisphere) × (Nyquist velocity/ *V *max). The Rvol was calculated as Rvol = EROA × VTI. In the case of double orifices, the whole PISA surface was estimated by taking only the biggest hemispheric PISA (single hemisphere model) or by addition of each separated hemispheric PISAs (double hemisphere model; Figure 3 B). Alternatively, the PISA surface was calculated on the basis of a hemicylindric assumption ( *S *hemicylinder = π × *r *× *h *), with two bases ( *S *full hemicylinder = *S *hemicylinder + π *r *^{2 }), or with only one base ( *S *partial hemicylinder = *S *hemicylinder + 1/2π *r *^{2 }; Figure 3 C). We came to this formula observing the shape of the fused PISAs during the offline analysis. They were not really hemicylindric, given the rounded edges, but this geometrical assumption fitted better than the hemispheric assumption in this particular setting ( Figure 2 B). The horizontal lengths ( *h *) of each separated PISAs were added in order to obtain a unique hemicylinder length, and the radii of each observed PISA were averaged to obtain one hemicylinder radius.

The 3D PISA method has been described elsewhere. After suppression of the B-mode components, the PISA was cut into 16 slices from top to base and exported in our customized software (MATLAB, MathWorks Inc., Natick, MA). The first aliasing (red blue shift) border was automatically detected and marked with a red line on each slice. The pixels inside the PISA borders were converted into white-on-black surfaces, and the latter were stacked up to build a 3D object. A triangular mesh was computed onto this object, allowing the measurement of his surface ( Figure 2 ). All PISA shapes visible during the systole were extracted and reconstructed as described. For each PISA _{i }, the regurgitant flow rate ( *Qr *_{i }= PISA _{i }× Nyquist velocity _{i }) and the Rvol (Rvol _{i }= *Qr *_{i }× 1/volume rate) was computed. The integrated Rvol was calculated as integrated 3D PISA Rvol = <SPAN role=presentation tabIndex=0 id=MathJax-Element-1-Frame class=MathJax style="POSITION: relative" data-mathml='∑i=1nRvoli’>∑ni=1Rvoli∑i=1nRvoli

∑ i = 1 n Rvol i

( *n *= number of PISA extracted).

## Statistics

Statistical analysis was performed using SPSS 21 (IBM Corporation, Armonk, NY). Continuous variables were expressed as mean ± SD. Comparisons of mean Rvol (mRvol) between different methods among groups and between groups were performed with paired and unpaired *t *-test and analysis of variance for repeated measures with Bonferroni post hoc analysis. Linear regression analysis and Bland-Altman bias with limits of agreement (−1.96 SD to +1.96 SD) were performed to estimate the precision of the different imaging modalities. Receiver operating characteristic (ROC) analysis for accuracy analysis was used to determine the cutoff and the biggest area under the ROC curve (AUC).

## Results

In total, 69 different Rvols were simulated for TTE (Rvol range, 8–73 mL) and 72 different Rvols for TEE (Rvol range, 8–76 mL). For each modality (TTE and TEE), we classified the Rvol into three groups: single orifice ( *n *= 23 for TTE, *n *= 24 for TEE), double orifices with PISA fusion ( *n *= 24 for TTE, *n *= 21 for TEE), and double orifices without PISA fusion ( *n *= 23 for TTE, *n *= 27 for TEE; Table 1 ). The mRvol measured by flowmeter was statistically similar between the TTE and TEE groups ( Table 1 ).

TTE | All orifices ( N = 69) |
One orifice ( n = 23) |
Two orifices, fusion ( n = 24) |
Two orifices, no fusion ( n = 22) |
P value |
---|---|---|---|---|---|

Flowmeter | |||||

Rvol | 36.5 ± 18.0 | 36.2 ± 18.4 | 37.5 ± 18.2 | 35.8 ± 18.2 | .943 |

EROA | 0.22 ± 0.11 | 0.21 ± 0.11 | 0.22 ± 0.11 | 0.22 ± 0.11 | NS |

3D integrated PISA | |||||

Rvol | 34.4 ± 22.1 | 31.2 ± 22.1 | 38.0 ± 22.5 | 33.9 ± 22.2 | .572 |

EROA | 0.20 ± 0.13 | 0.18 ± 0.12 | 0.22 ± 0.13 | 0.20 ± 0.13 | .472 |

Hemisphere | |||||

Rvol | 24.4 ± 19.2 | 34.9 ± 24.0 | 23.8 ± 15.9 | 14.1 ± 9.2 | .001 |

EROA | 0.14 ± 0.11 | 0.20 ± 0.13 | 0.14 ± 0.10 | 0.08 ± 0.06 | .002 |

2D double hemisphere | |||||

Rvol | 35.5 ± 25.6 | 34.9 ± 24.0 | 44.6 ± 30.6 | 26.1 ± 17.5 | .047 |

EROA | 0.21 ± 0.16 | 0.20 ± 0.13 | 0.27 ± 0.20 | 0.16 ± 0.11 | .049 |

2D partial hemicylinder | |||||

Rvol | 32.6 ± 21.5 | 36.5 ± 25.7 | 34.7 ± 20.9 | 26.1 ± 16.0 | .223 |

EROA | 0.19 ± 0.13 | 0.21 ± 0.14 | 0.21 ± 0.13 | 0.16 ± 0.10 | .331 |

2D full hemicylinder | |||||

Rvol | 38.5 ± 26.0 | 45.3 ± 31.7 | 40.3 ± 24.6 | 29.4 ± 18.2 | .110 |

EROA | 0.23 ± 0.15 | 0.26 ± 0.18 | 0.24 ± 0.16 | 0.18 ± 0.11 | .178 |

2D combined PISA | |||||

Rvol | 33.1 ± 21.0 | 34.9 ± 24.0 | 34.7 ± 20.9 | 29.4 ± 18.2 | .611 |

EROA | 0.19 ± 0.12 | 0.20 ± 0.13 | 0.21 ± 0.13 | 0.18 ± 0.11 | .727 |

TTE | All orifices ( N = 72) |
One orifice ( n = 24) |
Two orifices, fusion ( n = 21) |
Two orifices, no fusion ( n = 27) |
P value |

Flowmeter | |||||

Rvol | 38.5 ± 19.7 | 38.3 ± 19.8 | 41.9 ± 18.8 | 36.1 ± 20.6 | .598 |

EROA | 0.23 ± 0.11 | 0.22 ± 0.10 | 0.25 ± 0.11 | 0.23 ± 0.13 | .563 |

3D integrated PISA | |||||

Rvol | 35.3 ± 22.2 | 34.5 ± 20.7 | 40.9 ± 21.4 | 31.6 ± 23.9 | .340 |

EROA | 0.21 ± .012 | 0.19 ± 0.11 | 0.24 ± 0.11 | 0.20 ± 0.14 | .337 |

Hemisphere | |||||

Rvol | 27.0 ± 18.0 | 35.6 ± 19.9 | 28.1 ± 15.7 | 18.5 ± 14.3 | .002 |

EROA | 0.16 ± 0.10 | 0.20 ± 0.10 | 0.17 ± 0.09 | 0.12 ± 0.09 | .010 |

2D double hemisphere | |||||

Rvol | 39.8 ± 26.3 | 35.6 ± 19.9 | 52.3 ± 29.8 | 33.9 ± 8.3 | .032 |

EROA | 0.24 ± 0.11 | 0.20 ± 0.10 | 0.31 ± 0.17 | 0.22 ± 0.17 | .025 |

2D partial hemicylinder | |||||

Rvol | 34.8 ± 21.5 | 34.9 ± 20.2 | 42.7 ± 21.0 | 28.7 ± 21.7 | .078 |

EROA | 0.21 ± 0.12 | 0.20 ± 0.11 | 0.26 ± 0.11 | 0.18 ± 0.14 | .103 |

2D full hemicylinder | |||||

Rvol | 41.3 ± 25.5 | 43.8 ± 25.1 | 49.2 ± 24.6 | 32.9 ± 24.9 | .072 |

EROA | 0.25 ± 0.14 | 0.25 ± 0.13 | 0.29 ± 0.13 | 0.21 ± 0.16 | .138 |

2D combined PISA | |||||

Rvol | 36.6 ± 22.3 | 35.6 ± 19.9 | 42.7 ± 21.0 | 32.9 ± 24.9 | .307 |

EROA | 0.22 ± 0.13 | 0.20 ± 0.10 | 0.26 ± 0.11 | 0.21 ± 0.16 | .308 |

The volume rates of the loops acquired by TTE and TEE were similar (TTE: mean, 43 Hz, range, 39–47 Hz; TEE: mean, 42.2 Hz, range, 27–47 Hz).

The 2D hemispheric PISA assumption correlated well with the reference method in the presence of one MR orifice ( *y *= 1.26 *x *− 10.79, *r *^{2 }= 0.93, *P *< .001 for TTE; *y *= 0.99 *x *− 2.35, *r *^{2 }= 0.97, *P *< .001 for TEE), with no significant or only minimal underestimation of the mRvol (bias, −1.2 ± 7.8 mL, *P *= .454 for TTE; bias, −2.7 ± 3.2 mL, *P *< .001 for TEE). In the presence of double MR orifices, measuring only the biggest hemispheric PISA (single hemisphere model) led to a significant underestimation of mRvol (bias, −17.5 ± 8.9 mL for TTE, *P *< .001; bias, −15.9 ± 7.3 mL for TEE, *P *< .001). Moreover, the bias toward underestimation increased at larger simulated Rvol. In the setting of double MR orifices and fused PISAs, adding two hemispheric PISAs (double hemisphere model) overestimated significantly the mRvol (bias, 7.1 ± 14.6 mL, *P *= .025 for TTE; bias, 10.4 ± 11.9 mL, *P *= .001 for TEE). Here again, the bias toward overestimation increased at larger simulated Rvol ( Figure 4 ). In the case of double MR orifices and nonfused PISAs, Rvol obtained by the double hemisphere method was associated with reference Rvol as *y *= 0.94 *x *− 7.42, *r *^{2 }= 0.95, *P *< .001 for TTE; *y *= 1.22 *x *− 9.94, *r *^{2 }= 0.93, *P *< .001 for TEE, with a significant underestimation of the mRvol by TTE (bias, −9.7 ± 4.2 mL, *P *< .001) and nonsignificant underestimation by TEE (bias, −2.2 ± 8.3 mL, *P *= .181). In the presence of double MR orifices and fused PISAs, we found that a hemicylindric assumption of the PISA, completed with the addition of one hemicircular basis (partial hemicylinder model), was well associated with reference method ( *y *= 1.12 *x *− 7.26, *r *^{2 }= 0.96, *P *< .001 for TTE; *y *= 1.08 *x *− 2.77, *r *^{2 }= 0.95, *P *< .001 for TEE) with only minimal underestimation of the mRvol by TTE (bias, −2.8 ± 4.8 mL, *P *= .009), and no significant underestimation of the mRvol by TEE (bias, 0.8 ± 5.0 mL, *P *= .485; Figure 5 ). In the case of double MR orifices and nonfused PISAs, Rvol obtained by hemicylindric assumption, completed with two hemicircular bases (full hemicylinder model), was associated with reference Rvol as *y *= 0.96 *x *− 5.1, *r *^{2 }= 0.93, *P *< .001 for TTE; *y *= 1.17 *x *− 9.33, *r *^{2 }= 0.94, *P *< .001 for TEE ( Figure 5 ). The full hemicylinder model underestimated slightly the mRvol as compared with the reference (bias, −6.4 ± 5.0 mL, *P *< .001 for TTE; bias, −3.2 ± 7.0 mL, *P *= .025 for TEE).