Consecutive patients with secondary MVR and reduced left ventricular ejection fraction (EF < 50%) evaluated for TEER between September 2021 and September 2024 underwent CMR. Patients were recruited prospectively to the Prediction of Reverse Remodeling and Outcome in Patients With Severe Secondary Mitral Valve Regurgitation Undergoing Transcatheter Edge-to-edge Mitral Valve Repair (PRE-MITRA) study (NCT04913727) after written informed consent. Exclusion criteria were age below 18 years, pregnancy or breastfeeding, severely impaired renal function (GFR < 15 ml/min), untreated severe concomitant valve diseases, such as severe tricuspid valve regurgitation, severe claustrophobia, or any other contraindication for CMR. The study was approved by the local ethics committees (KEK-BE 2021 − 00704) and was conducted following the Declaration of Helsinki.

CMR scans were performed on 1.5T (MAGNETOM Sola, Siemens Healthcare, Erlangen, Germany) (N = 28) and 3T (MAGNETOM Prisma, Siemens Healthcare, Erlangen, Germany) (N = 4) scanners. Cine steady-state free precession (bSSFP) sequences were employed with retrospective electrocardiographic (ECG) gating with breath-hold. Standard cardiac geometries were acquired, encompassing multiple short-axis slices with no interslice gaps, covering the entire left ventricle (LV) and long-axis two-, three-, and four-chambered views. The reconstructed in-plane spatial resolution was 2.1 × 2.1 mm2, with 8.0 mm slice thickness and 25 cardiac phases.

Additionally, different 2D-PC flow measurements were conducted: one in the ascending aorta, with the imaging plane positioned 10 mm above the aortic valve and perpendicular to the aortic flow direction, and the other across the mitral area valve. Velocity encoding (VENC) was set typically to 150 cm/s, adjusted individually if necessary. ECG gating was used for both acquisitions, with an in-plane spatial resolution of 2.0 × 2.0 mm2, temporal resolution of 25 phases per cardiac cycle, and slice thickness of 7 mm.

Out of the 32 patients enrolled in this study, only 15 were able to undergo additional 4D-flow acquisition. The remaining patients were excluded from this protocol extension due to their inability to tolerate prolonged scanning times, attributed to frailty, reduced compliance, or limitations in maintaining a stable position for the extended acquisition duration. For these patients, the 4D-flow data were acquired with a spatial resolution of 2.5 × 2.5 × 2.5 mm3 and a temporal resolution of 25 phases per cardiac cycle. VENC was set to 120 cm/s, with adjustments made as necessary to optimize data quality. All acquisition parameters are summarized in Table 1.

Image analysis was performed using CVI42 software (Circle Cardiovascular Imaging, Calgary, Canada, version 6.0.2) with two readers (authors YS and MB) blinded to the patient’s clinical characteristics and outcomes. Measurements were performed twice by the same reader (A1 and A2) and once by a different reader (B). Outliers with Z-scores greater than two were excluded to ensure data clarity and consistency.

Seven distinct methods for quantifying MVR volume using CMR were employed in this study (Fig. 1): (1) 2D-PC standard method (2D-PCstandard), which indirectly calculates MVR volume by subtracting the 2D-PC measured aortic outflow volume (AoPC), from the left ventricular stroke volume (LVSV) volumetrically obtained from standard 2D cine short-axis CMR images; (2) 2D-PC mitral valve method (2D-PCMVAAo), quantifying MVR volume by subtracting AoPC from the forward flow through the mitral valve (MVPC), both measured using 2D-PC planes; (3) 2D-PC mitral valve direct method (2D-PCMVdirect), directly measuring the regurgitation volume from the 2D-PC plane on the mitral valve; (4) 4D-flow standard method (4D-flowstandard ), which indirectly calculates MVR volume by subtracting AoPC, measured using 4D-flow CMR, from LVSV derived from standard 2D cine short-axis CMR images; (5) 4D-flow mitral valve (4D-flowMVAAo), quantifying MVR volume by subtracting AoPC from MVPC, both assessed using 4D-flow CMR; (6) 4D-flow mitral valve direct method (4D-flowMVdirect), directly measuring the regurgitation volume from the flow plane on the mitral valve using 4D-flow acquisition data; and (7) the Volumetric method that calculates MVR volume as the difference between LVSV and right ventricular stroke volume (RVSV), both obtained from standard 2D cine CMR images.

Illustration of MVR quantification methods. 2D-PCstandard, CMR flow gold standard (Left Ventricle Stroke Volume [LV SV]– Aortic Forward Flow Volume derived from 2D-PC plane [AoPC]); 2D-PCMVAAo, Forward Flow through Mitral Valve (MVPC) - AoPC; 2D-PCMVdirect, directly quantifying flow through Mitral Valve (Backward flow through Mitral Valve [RegurgVolMV]); Volumetric (LV SV– Right Ventricle Stroke Volume [RV SV]); 4D-flowstandard, similar to 2D-PCstandard using 4D-flow sequence instead of 2D-PC (LV SV– AoPC); 4D-flowMVAAo, similar to 2D-PCMVAAo using 4D-flow sequence instead of 2D-PC (MVPC– AoPC); 4D-flowMVdirect, directly quantifying flow through Mitral Valve (Backward flow through Mitral Valve [RegurgVolMV]); AoPC, Aortic Forward Flow (OutflowAAo); MVPC, Mitral Valve Forward Flow (InflowMV); EDV, End Diastolic Volume; ESV, End Systolic Volume

Box plots of Mitral Valve Regurgitation (MVR) volumes measured using various CMR flow quantification methods. Measurements were taken with seven distinct methods, based on a cohort comprising 26 patients with 2D-PC data, of whom 15 patients also underwent 4D-flow acquisition. Each method was analyzed by two readers: once by reader A and twice by the same reader (B1 and B2). Outliers (Z-score > 2) were removed from the data to enhance clarity. Each method is illustrated with its corresponding box plot, with individual measurement points overlayed in red

All 4D-flow measurements, including both aortic and mitral valve flow quantifications, were performed using valve tracking functionality in CVI42. For each valve, regions of interest (ROIs) were adapted throughout the cardiac cycle to account for in-plane motion and vessel dilation. For the direct mitral regurgitation measurements, the analysis plane was placed at the level of the mitral valve annulus, consistent with the corresponding 2D-PC acquisition planes, ensuring cross-method consistency. The regurgitant jet was also tracked throughout the cardiac cycle using the retrospective valve tracking feature to ensure accurate capture of flow dynamics across all time points.

Inter- and intra-reader agreement analyses were performed to evaluate the consistency and reliability of MVR quantification methods by calculating intraclass correlation coefficients (ICC). ICC values less than 0.50 indicate poor agreement, 0.50 to 0.75 indicate moderate agreement, 0.75 to 0.90 indicate good agreement, and values greater than 0.90 indicate excellent agreement [10, 11].

Bland-Altman plots were generated to visually assess the agreement between different measurements for inter- and intra-reader analyses, highlighting any systematic biases or discrepancies between these measurements.

MVR severities were classified into three categories based on the quantified MVR volume, following established clinical thresholds [12]: mild (MVR volume less than 30 ml), moderate (MVR volume between 30 and 59 ml), and severe (MVR volume greater than 60 ml).

Statistical analyses were conducted using Python (version 3.9.6). Statistical significance was denoted using stars: * indicates p < 0.05, ** indicates p < 0.01, and *** indicates p < 0.001.