Comparison of Biplane Area-Length Method and 3D Volume Quantification by Using Cardiac MRI for Assessment of Left Atrial Volume in Atrial Fibrillation

Abstract

Purpose

To compare maximum left atrial (LA) volume (LAV) from the routinely used biplane area-length (BAL) method with three-dimensional (3D)–based volumetry from late gadolinium-enhanced MRI (3D LGE MRI) and contrast-enhanced MR angiography (3D CE-MRA) in patients with atrial fibrillation (AF).

Materials and Methods

Sixty-four patients with AF (mean age, 63 years ± 9 [SD]; 40 male patients) were retrospectively included from a prospective cohort acquired between October 2018 and February 2021. All patients underwent a research MRI examination that included standard two- and four-chamber cine acquisitions, 3D CE-MRA, and 3D LGE MRI performed prior to the atrial kick. Contour delineation on cine imaging and LA 3D segmentations were performed by a radiologist. Maximum LAV (BALmax) was extracted from the BAL volume-time curve and compared with LAV from 3D CE-MRA and 3D LGE MRI. The Kruskal-Wallis test was performed, followed by the Dunn post hoc test and Bland-Altman analyses. Interobserver variability was assessed in 10 patients.

Results

BALmax underestimated LAV compared with 3D CE-MRA (bias: -23.5 mL ± 46.2, P < .001) and 3D LGE MRI (bias: -31.3 mL ± 58.3, P < .001), whereas 3D LGE MRI volumes showed no evidence of a difference from 3D CE-MRA (bias: 7.8 mL ± 45.7, P = .38). Interobserver variability yielded excellent agreement for each method (intraclass correlation coefficient, 0.96–0.98).

Conclusion

BALmax underestimated LAV in patients with AF compared with 3D LGE MRI and 3D CE-MRA, suggesting that the geometric assumption of an ellipsoidal LA shape in BAL does not reflect LA geometry in patients with AF.

Keywords: Left Atrial Volume, Biplane Area-Length, Late Gadolinium-enhanced 3D MRI, Contrast-enhanced 3D MR Angiography, Atrial Fibrillation

Supplemental material is available for this article.

© RSNA, 2023

Summary

The biplane area-length method underestimated left atrial volume in patients with atrial fibrillation compared with three-dimensional–based volumetry from late gadolinium-enhanced MRI and contrast-enhanced MR angiography.

Key Points

• ■ In patients with atrial fibrillation, the biplane area-length method significantly underestimated left atrial volume by 23.5 mL and 31.3 mL when compared with measurements derived from three-dimensional segmentations of late gadolinium-enhanced MRI (3D LGE MRI) (P < .001) and contrast-enhanced MR angiography (P < .001), respectively.

• ■ This offset between methods increased progressively with larger atria (r = 0.51, P < .001).

• ■ Interobserver analysis yielded excellent intraclass correlation for all methods (0.96–0.98), but only 3D LGE MRI showed no evidence of a difference between observers (bias: 0.5 mL; P = .87).

Introduction

Left atrial (LA) volume (LAV) is an important parameter for assessing cardiovascular morbidity and mortality in both the general population and in patients with atrial fibrillation (AF) (1,2). In fact, LA enlargement has been shown to predict stroke, myocardial infarction, and congestive heart failure in AF, as well as recurrent AF following antiarrhythmic drug therapy and ablation (3–6). Accordingly, accurate measurement of LAV is crucial for risk stratification. While echocardiography is the most widely used modality for LAV quantification, cardiac MRI is considered the reference standard for measuring cardiac volumes and function because it offers better accuracy and reproducibility (7).

Several approaches are used for quantifying LAV using cardiac MRI. The Simpson method, considered the reference method, measures the volume from a stack of cardiac-gated two-dimensional (2D) time-resolved (cine) short-axis series (8) to provide a three-dimensional (3D) time-resolved measurement of the LAV. However, it requires manual tracing of the LA contours in each contiguous slice and for every time frame, which limits its clinical application. The biplane area-length (BAL) method derived from 2D cine MRI in two- and four-chamber planes is a faster alternative and is used in clinical routine for LAV assessment (9–11). While the BAL method also allows for time-resolved estimation of LAVs for each time point along the cardiac cycle, it relies on the geometric assumption of an ellipsoidal LA shape for LAV estimation (12).

LAV can also be measured directly, without the need for geometric assumptions, from 3D acquisitions such as late gadolinium-enhanced MRI (3D LGE MRI) or contrast-enhanced 3D MR angiography (3D CE-MRA). Electrocardiographically triggered 3D LGE MRI is performed during diastole, immediately before atrial contraction, for the assessment of atrial fibrosis (13), whereas 3D CE-MRA is an ungated technique that is typically performed for preprocedural pulmonary vein mapping and monitoring for stenosis after ablation (14). LAV can be measured by both methods natively via 3D-based LA segmentation. Although 3D imaging techniques do not rely on geometric assumptions, they reflect a time-averaged LAV for 3D CE-MRA and a single cardiac time-point or short-interval LAV for 3D LGE MRI, and hence do not provide time-resolved volume information along the cardiac cycle (15,16).

The purpose of this study was to conduct a systematic comparison of LAV assessments derived from 2D BAL versus 3D techniques, particularly 3D CE-MRA and 3D LGE MRI, in patients with AF.

Materials and Methods

Study Patients

Patients with AF scheduled for ablation were recruited to undergo cardiothoracic research MRI examination at a tertiary center between October 2018 and February 2021 as part of an ongoing prospective Health Insurance Portability and Accountability Act–compliant study. Institutional review board approval was granted, and all patients provided written informed consent. All patients in this study were retrospectively selected from the prospectively acquired cohort. Inclusion criteria included availability of all three imaging techniques in each patient. Information regarding AF type (paroxysmal or persistent) and time since AF diagnosis were available from study records.

Cardiac MRI Acquisition

Cardiac MRI scans were acquired with 1.5-T systems (Avanto or Aera; Siemens Healthineers). Retrospectively electrocardiographically gated balanced steady-state free precession cine images were acquired in the standard two- and four-chamber views, during breath hold, using the following parameters: image acquisition matrix size = 192 × 192, spatial resolution = 1.8 × 1.8 mm, slice thickness = 6 mm, echo time/repetition time = 1.9/3.2 msec, and flip angle = 51.1° ± 21.8 (SD). Three-dimensional CE-MRA was performed using noncardiac-gated time-resolved angiography with interleaved stochastic trajectories technique during administration of gadobutrol at a dose of 0.05 mmol per kilogram of body weight (Gadavist; Bayer Healthcare Pharmaceuticals). Technical parameters were the following: image acquisition matrix size = 320 × 130, spatial resolution = 1.3 × 1.3 × 1.5 mm, receiver bandwidth = 680 Hz/pixel, echo time/repetition time = 2.6/1.1 msec, and flip angle = 25°. Following 3D CE-MRA, 0.15 mmol of gadobutrol per kilogram of body weight was administered. Approximately 15–20 minutes later, cardiac-gated 3D LGE MRI pulse sequence using balanced steady-state free precession readout with stack-of-stars k-space sampling was conducted in mid to end diastole (16). The 3D sequence parameters were as follows: image acquisition matrix size = 192 × 192 × 52, field of view = 288 × 288 × 114.4 mm, nominal spatial resolution = 1.5 × 1.5 × 2.2 mm, receiver bandwidth = 704 Hz/pixel, echo time/repetition time = 1.8/3.5 msec, flip angle = 40°, and inversion time to null the normal myocardium = 220–350 msec.

Volumetric Assessment

BAL method.—LA contours on two- and four-chamber cine series were semiautomatically delineated by a cardiovascular radiologist (M.P.) with 3 years of experience using a commercial software program (cvi42, version 5.13.9; Circle Cardiovascular Imaging) (Fig 1A). LA appendage and pulmonary veins were excluded, and the mitral annular plane was defined as the border between the LA and left ventricle. LAVs were automatically calculated by the software for each time point of the cardiac cycle using the following equation (17), where A is LA area, L is LA length, 2C is two-chamber view, and 4C is four-chamber view:

Figure 1:

Illustrations of the studied left atrial (LA) volume (LAV) quantification techniques in a 54-year-old man. (A) Biplane area-length (BAL) method. The LA endocardium is contoured on two-chamber (2C) and four-chamber (4C) MRI views for all time frames of the cardiac cycle (left). A time-volume curve (right) demonstrates the extraction of the minimum LAV (BALmin) and maximum LAV (BALmax). (B) LA three-dimensional (3D) segmentation based on 3D contrast-enhanced MR angiography (3D CE-MRA) and (C) LA 3D segmentation based on 3D late gadolinium-enhanced MRI (3D LGE MRI).

Next, the LA minimal volume at left ventricular end diastole (ie, BALmin) and LA maximal volume at left ventricular early diastole before mitral valve opening (BALmax) were extracted. We used BALmax as the clinical routine standard and compared it with the volumes derived from 3D techniques.

Three-dimensional CE-MRA and 3D LGE MRI.—Three-dimensional CE-MRA was performed using time-resolved angiography with interleaved stochastic trajectories technique, which captures multiple arterial, mixed, and venous phase series during the passage of the contrast material. To improve vascular conspicuity, postcontrast series were subtracted from a precontrast series. The radiologist (M.P.) selected the subtracted image series showing the highest LA contrast. The radiologist performed LA segmentation on the selected 3D CE-MRA series (Fig 1B) and 3D LGE MRI series (Fig 1C) using two commercial software programs (Mimics [version 20.0] from Materialise and ADAS3D [version 2.11.1] from Galgo Medical, respectively). Similar to the BAL method, the radiologist excluded LA appendage and pulmonary veins and defined the mitral valve plane as the border between the LA and left ventricle for both techniques.

Rhythm Assessment

To determine whether patients were in sinus or AF rhythm at the time of the MRI examination, electrocardiographic tracings recorded during 3D LGE MRI acquisition were extracted and analyzed by the radiologist (M.P.). Scans had either regular R-R intervals, indicating that the patients were in sinus rhythm, or irregular intervals, suggesting absolute arrhythmia caused by AF during MRI acquisition.

Interreader Assessment

To evaluate interobserver variability, 10 patients were randomly selected and analyzed by a trainee with 1 year of experience (A.M.) who was blinded to the results of the first observer (M.P.).

Statistical Analysis

Statistical analysis was performed using SPSS Statistics software (version 26; IBM). Normality was checked with the Kolmogorov-Smirnov test, which revealed a nonnormal distribution. LAV measurements were then compared using the Kruskal-Wallis test, followed by the Dunn multiple comparison test with Bonferroni correction. This analysis was conducted initially for the entire sample, then for the sinus and AF groups separately to determine whether the underlying rhythm had any influence on the results. Pearson correlation and Bland-Altman plots were also calculated to evaluate the mean difference (bias) and limits of agreement. Pearson correlation coefficient was scored as follows: poor = 0, slight = 0.01–0.20, fair = 0.21–0.40, moderate = 0.41–0.60, good = 0.61–0.80, and excellent = 0.81–1.00. Interreader variability was assessed by calculating intraclass correlation coefficients and Bland-Altman analyses.

To investigate potential sources of disagreement between BALmax and LAV from 3D techniques, we compared the difference between the two methods with absolute 3D volume (by correlation analysis), AF type (logistic regression), and time since AF diagnosis (linear regression) in MATLAB software (version 2019b, The MathWorks). In general, a P value < .05 was considered significant (P < .017 after Bonferroni correction in post hoc pairwise comparisons between techniques).

Results

Patient Characteristics

We recruited 120 patients. Of these, 53 did not have all three acquisitions available and three had imaging artifacts (see flowchart in Fig 2). In total, 64 patients with AF were included in this study, of whom 41 were determined to have been in sinus and 23 in AF rhythm during MRI acquisition. Mean patient age was 63 years ± 9 [SD], and 40 patients were male (Table 1). The LAVs as calculated using BAL, 3D CE-MRA, and 3D LGE MRI techniques are summarized for the entire study sample in Table 2 and are separated according to sinus and AF groups in Table 3. In the entire sample, the Kruskal-Wallis test showed a statistically significant difference among the three techniques (P < .001), and the Dunn post hoc test yielded two significant pairwise comparisons (Fig 3).

Figure 2:

Flowchart shows patients and reasons for exclusion. AF = atrial fibrillation, 3D CE-MRA = three-dimensional contrast-enhanced MR angiography, 3D LGE MRI = three-dimensional late gadolinium-enhanced MRI.

Table 1:

Patient Demographics and Characteristics

Table 2:

Descriptive Statistics for Left Atrial Volume Quantification by Method in Entire Study Sample

Table 3:

Descriptive Statistics for Left Atrial Volume Quantification by Method in Subgroups of Patients in Sinus Rhythm and Atrial Fibrillation during MRI Acquisition

Figure 3:

Graph shows left atrial volume comparison between the maximum left atrial volume from the biplane area-length method (BALmax), left atrial volume from three-dimensional) contrast-enhanced MR angiography (3D CE-MRA), and left atrial volume from 3D late gadolinium-enhanced MRI (3D LGE MRI) in the entire sample. Significant differences were observed for BALmax versus 3D CE-MRA and BALmax versus 3D LGE MRI, whereas there was no evidence of a difference between the 3D methods. The horizontal line in each box plot denotes the median value (50th percentile), while the box contains the 25th to 75th percentiles of the data set. The black whiskers mark the fifth and 95th percentiles, and values beyond these upper and lower bounds represent outliers.

Comparison of Volumetric Assessment by Method

In the entire sample, BALmax underestimated LAV by 21% compared with 3D CE-MRA (bias: -23.5 mL; limits of agreement: -69.7 to 22.8 mL; P < .001; Fig 4) and by 26% compared with 3D LGE MRI (bias: -31.1 mL; limits of agreement: -89.6 to 27.0 mL; P < .001). However, Pearson analysis showed excellent correlation between BALmax and 3D CE-MRA (r = 0.83; P < .001) and good correlation between BALmax and 3D LGE MRI (r = 0.79; P < .001).

Figure 4:

Pearson correlations (left) and Bland-Altman plots (right) show left atrial volume from three-dimensional contrast-enhanced MR angiography (3D CE-MRA) versus maximum left atrial volume from the biplane area-length method (BALmax) (top), left atrial volume from 3D late gadolinium-enhanced MRI (3D LGE MRI) versus BALmax (middle), and 3D CE-MRA versus 3D LGE MRI (bottom) in the entire sample. In the Bland-Altman plots, the red lines represent the mean (bias), and black lines represent the 95% limits of agreement.

We found no evidence of a difference in LAV quantification between 3D LGE MRI and 3D CE-MRA (bias: 7.8 mL; limits of agreement: -37.9 to 53.5 mL; P = .38), and these techniques achieved the highest correlation (r = 0.87; P < .001).

Subgroup analyses for patients in sinus and AF rhythm at MRI examination yielded the same results when comparing the three methods (see Appendix S1).

Impact of AF Type and Time Since AF Diagnosis on the Disagreement between Methods

We detected a significant moderate correlation (r = 0.51; P < .001) for the difference in 3D LGE MRI and BALmax volumes versus absolute 3D LGE MRI volumes (Fig 5), indicating that larger absolute volumes yielded higher disagreement between 2D and 3D assessments. In regression analysis, neither AF type nor time since AF diagnosis was associated with disagreement between methods (P = .81 and P = .27, respectively).

Figure 5:

Scatterplot shows the difference between left atrial volume from three-dimensional late gadolinium-enhanced MRI (3D LGE MRI) and maximum left atrial volume from the biplane area-length method (BALmax) versus 3D LGE MRI absolute volumes. The color bar shows the time since atrial fibrillation (AF) diagnosis. The symbols indicate persistent (triangle) and paroxysmal (circle) AF. Larger absolute volumes based on 3D LGE MRI yielded higher disagreement (= larger differences) between BALmax and 3D LGE MRI. This trend was not associated with AF type or time since AF diagnosis.

Interobserver Analysis

Table 4 shows the interobserver variability for the three techniques. All methods showed excellent interobserver correlation (intraclass correlation coefficient = 0.96–0.98). Bland-Altman analysis, however, yielded a significant bias between the two observers for BALmax and 3D CE-MRA (-3.5 mL, P = .01; and 8.8 mL, P < .01, respectively). Three-dimensional LGE MRI assessment did not show a significant bias (P = .87).

Table 4:

Interobserver Variability of Left Atrial Volume Quantification Methods

Discussion

LAV is an independent risk factor for adverse cardiac events and, therefore, must be clinically assessed in patients with AF. In this study, we examined the agreement of three MRI methods of determining LAV and found that the BAL method significantly underestimated LAV compared with 3D CE-MRA (-23.5 mL, P < .001) and 3D LGE MRI (-31.1 mL, P < .001), and we found no evidence of a difference between the two 3D techniques (7.8 mL, P = .38). Furthermore, we found that the volume difference between the BAL and 3D LGE MRI methods increased with larger absolute LA volumes (r = 0.51, P < .001).

The BAL method is widely used for estimating maximum LAV because it is fast and requires little user input compared with 3D methods (11). However, the BAL method has multiple inherent limitations: (a) it relies on the geometric assumption of an ellipsoidal LA shape, which renders it inaccurate when LA enlargement does not occur in an ellipsoidal form (18); (b) it depends on the accurate placement of two- and four-chamber planes following a series of scout images, so inaccurate plane placement may cause foreshortening of the LA (19); and (c) typical two- and four-chamber views are oriented along the left ventricular long axis and do not necessarily represent the respective LA axes, potentially introducing a source of error (20).

We found that despite good to excellent correlation between methods, BALmax significantly underestimated the LAV compared with measurements derived from the two 3D techniques. We observed these findings in the entire study sample and additionally in subgroups of patients in sinus rhythm and AF rhythm during MRI acquisition, suggesting that the underlying heart rhythm during the MRI examination had no influence on the significance of results in this study. Also, this difference is likely not caused by interobserver variability because we observed lower biases and excellent correlations between our two observers. On the other hand, the two 3D techniques showed no evidence of a difference in LAV assessment and had the highest correlation between any two methods. We therefore suggest that the BAL method, which relies on geometric assumptions for a normal-shape LA, may not capture the entire LAV in patients with AF where asymmetric wall remodeling is expected (19). Hence, 3D techniques should be preferred. Our results were in agreement with three previous studies that reported an underestimation by the BAL method when comparing BALmax from transthoracic echocardiography to 3D CE-MRA, and BALmax to the Simpson technique (9,21,22). It is notable, however, that we observed a higher absolute bias between 2D and 3D methods compared with these prior reports, which included, on average, smaller LA than our study sample.

When investigating potential factors influencing the volume difference between BALmax and 3D LGE MRI, we found significantly higher disagreement with larger volumes. Prior studies suggested that LA enlargement occurs irregularly and asymmetrically across the three directions in space, which could explain this progressive increase in disagreement (19,23,24). Furthermore, given that some of these prior reports included smaller LA than our study sample, this finding might explain why the absolute biases between 2D and 3D methods were higher in our analysis (9,21). While studies have described the impact of AF type and AF duration on LA size, we did not find AF type or time since AF diagnosis to be associated with a larger volume difference between methods (25). On the other hand, the difference in the underlying atrial pathology of these patients (eg, valvular disease, infiltrative heart disease) could be associated with different atrial dilatation profiles but was not analyzed in this study.

Finally, interobserver analysis revealed significant interobserver bias for the BAL method and 3D CE-MRA and no evidence of a significant bias for 3D LGE MRI, despite the excellent correlation for all methods. This difference in interobserver agreement among the methods could be explained by the varying segmentation techniques used for each method. The 3D CE-MRA segmentation was performed using commercial software that required the delineation of LA contours on all slices, which likely increased interobserver discrepancies. In contrast, LA segmentations on cine MRI series and 3D LGE MRI scans were software assisted. However, because BALmax is calculated from the peak four-chamber and two-chamber LA areas and the maximum LA length derived from LA contouring, even slight variations in LA delineation of the cardiac phase that corresponds to peak LAV may result in large BALmax discrepancies between observers.

Our study had several limitations. First, this was a single-center study, and cardiac acquisitions were acquired with MRI systems from a single vendor. Second, the study focused on comparing the BAL method with 3D CE-MRA, which has no cardiac gating, and 3D LGE MRI, which is cardiac gated and acquired in mid diastole. Comparison with the Simpson technique, which provides time-resolved 3D-based LAV assessment, would be advantageous but was not performed in this study. However, it is important to consider that, in cardiac arrhythmias, time-resolved 2D (cine) series are hampered by reconstruction artifacts caused by inconsistent data from multiple heartbeats. Real-time MRI with high temporal resolution could reduce these problems and improve the assessment of LAV in patients with AF during MRI acquisition (26). The third limitation of this study was that no comparison with healthy controls was performed to determine whether BAL underestimation occurs exclusively in patients with AF; controls were not recruited in the ongoing prospective study given that 3D CE-MRA and 3D LGE MRI require gadobutrol administration. Fourth, all methods required some judgment by the interpreter, particularly for the delineation of LA borders at the mitral valve and exclusion of pulmonary veins and the LA appendage, which could have partially contributed to observed differences. However, interobserver variability was excellent for all methods. Finally, time since AF diagnosis might not be a reliable characteristic given that AF often starts as an asymptomatic disease.

In conclusion, the routinely used BAL method significantly underestimated maximum LAV compared with the volumes derived from 3D LGE MRI and 3D CE-MRA in patients with AF. This underestimation was accentuated in larger atria, suggesting that LA enlargement does not follow the shape of an ellipsoid. Hence, we suggest that 3D measurements be preferred for LAV assessment in AF. We recognize that clinically, however, the BAL method is a convenient method that offers physicians the possibility to obtain an LAV estimation in a time-efficient manner, using commercial software programs that are clinically available. The findings of this study should therefore motivate the development of automated segmentation tools that could measure LAV in a time-efficient manner, thus limiting user input and interobserver variability and contributing toward the clinical translation of 3D LAV assessment.