Reference value of left and right atrial size and phasic function by SSFP CMR at 3.0 T in healthy Chinese adults

The size and function of the left atrium (LA) and right atrium (RA) are related closely with the prognosis of cardiovascular diseases. However, their normal reference values, as measured by cardiac magnetic resonance (CMR), are not well established in Chinese populations. Healthy Chinese subjects (n = 135, 66 males, age 23–83 years) without cardiovascular risk factors were recruited. We imaged the LA and RA of all subjects using short axis and long axis slices by steady-state free precession (SSFP) sequences using a 3.0T scanner. The size and functional parameters were measured. Age and gender differences in LA were further explored. The normal reference values of atrial dimensions, volumes, and empty fractions (EFs) were provided by short axis (SAX) and area-length methods. Volumes and EFs derived by the area-length method showed correlated well with those derived by the by SAX method, but significantly underestimated the volumes (all P < 0.001) and overestimated the LA EFs (all P < 0.001). Atrial dimensions and volumes were generally larger in males. Conduit EFs and total EFs showed gender differences. Most atrial parameters correlated with age. In general, our results showed that gender and age have considerable impact on LA and RA size and function.

Image Analysis. LA Measurement. All CMR images were measured using a dedicated CMR post-processing software (Qmass 7.6, Medis, The Netherlands). LA dimensions were measured at the end of the systolic phase of the left ventricle (before the opening of the mitral valve) on 2-ch, 3-ch, and 4-ch SSFP cine images (Fig. 1). The LA volume was measured by two methods. First, the bi-plane area-length method with manually drawn endocardial contours in 2-and 4-ch views with exclusion of left atrial appendage and pulmonary veins [22][23][24][25] ; and second, the short axis (SAX) method using Simpson's method on the short axis slices of the atria. To calculate the atrial stoke volume and empty fraction (EF), the atrial volume at three phases during the cardiac cycle was measured. LA maximal volume (LAV max ) was defined at the end systole before the opening of the mitral valve. LA minimal volume (LAV min ) was defined at the end of diastole, just before the closure of the mitral valve. The pre-atrial contraction volume (LAV p−ac ) was defined at the beginning of left atrial active contraction phase at the mid diastole of the ventricle. Parameters for atrial emptying volume and emptying function were calculated as follows:  26 .
RA Measurement. The dimensions of the RA were measured on 4-ch SSFP images and RV 2-ch SSFP images (Fig. 2). The RA volume was measured by the area-length method and the SAX method, similar to the LA. Single plane area-length and bi-plane area length were both used to calculate the RA volume 27 . Similar to the measurement of the LA, maximal RA volume (RAV max ), minimal RA volume (RAV min ), and pre-active contraction RA volume (RAV p−ac ) were acquired at the same phases as the LA. RA phasic functions were defined the same as LA phasic function. RA total emptying fraction, RA passive emptying fraction, and RA active emptying fraction were calculated using the same formulas. Similarly, indexed dimension and volume values were also calculated for the RA.
Inter-observer and Intra-observer Variability. Subjects (20%, 24 cases) were selected randomly to test inter-and intra-observer variability. For inter-observer variability, two independent observers (WHL and KW), with more than 2 years experience and 500 cases of CMR image analysis, finished the post-processing for atrial dimensions and volumes blindly. For intra-observer variability, one observer (WHL) repeated the measurements for all parameters using the identical methods 8 weeks apart. Statistical Analyses. Statistical analyses were performed using SPSS (version 17.0; SPSS Inc., Chicago, IL, USA) and MedCalc (MedCalc Software version 13.0; Ostend, Belgium). The Kolmogorov-Smirnov test was used to check the normal distribution of the continuous variables. Independent-sample T tests were used to compare the mean values between men and women. Continuous data are presented as the mean ± SD. Non-normally distributed data were converted into log (normally distributed data), and then expressed as the mean ± SD. The normal reference range was calculated as the mean ± 2 SD. Linear regression was used to analyse the relationships between cardiac parameters and age. The inter-and intra-observer variability was assessed using the Bland-Altman method. A P value of <0.05 was considered statistically significant. Data Availability. The datasets is available from the corresponding author on reasonable request.

Parameters
Total (mean ± SD) Male (mean ± SD)  Table 2. Most dimensions showed no gender differences. The anterior-posterior dimension on the 3-ch view was greater in females than in males (31.1 ± 5.5 vs. 28.7 ± 5.3 mm, P = 0.011). The longitudinal diameter on 4-ch was shorter in females than in males (55.2 ± 6.0 vs. 57.9 ± 5.7 mm, P = 0.008). However, after indexing by BSA, the indexed LA diameters in females were slightly greater than those in males (All P < 0.001).
LA Volume and Phasic Function. The LA volume parameters are shown in Table 3. Correlations between parameters measured by the SAX method and the bi-plane method were moderate. Compared with the SAX method, the bi-plane method underestimated the LA volumes and overestimated the phasic function (all P < 0.001). The  LA volumes in females were significantly lower than those in males, except for LAV max (P = 0.119 for SAX method and 0.090 for bi-plane area-length method, all others P < 0.05, namely 0.009 for the LAV p−ac SAX method, 0.008 for the LAV min SAX method, 0.004 for the LAV p−ac bi-plane area-length method, 0.020 for the LAV min bi-plane area-length method) (Table 4). However, after indexing by BSA, most of the volume parameters were similar between genders (P = 0.668 for the indexed LAV p−ac SAX method, P = 0.654 for the indexed LA min SAX method, P = 0.096 for the indexed LA max bi-plane area-length method, P = 0.755 for the indexed LAV p−ac bi-plane area-length method, and P = 0.949 for the indexed LAV min bi-plane area-length method), except for LAV max by the SAX method (female vs. male: 43.9 ± 8.2 mL/m 2 vs. 40.05 ± 8.3 mL/m 2 , P = 0.018). The LA conduit EF was greater in females than in males when measured by either the SAX method or the bi-plane method (P = 0.004 and 0.008, respectively), while there was no significant difference in booster pump EF and total LA EF (P = 0.984 for the booster pump EF by the SAX method, P = 0.372 for the booster pump EF by the bi-plane area-length method, P = 0.095 for the total EF by the bi-plane area-length method, and P = 0.654 for the total EF by the SAX method).
Normal Reference for RA. RA    longitudinal diameter in 2-ch, P = 0.581 for longitudinal diameter in 4-ch, and P = 0.127 for transverse diameter in 2-ch, except P < 0.001 for the transverse dimension in 4-ch), while the indexed diameters were higher in females than in males (P < 0.001 for the indexed longitudinal diameter in 2ch and 4ch, P = 0.020 for the indexed transverse diameter in 4ch), except for the indexed transverse diameter in 2ch, where P = 0.087).
RA Volume and Phasic Function. The RA volume and phasic function data are shown in Table 6. Compared with the SAX method, the absolute volume and indexed volume measured by either single plane or bi-plane area-length methods were much lower (all P < 0.001). Correlations between the SAX and the area-length methods were moderate. The phasic functions of the RA were similar when assessed by the two methods (all P > 0.05, namely P = 0.278 for the conduit EF by the bi-plane area-length method vs. the SAX method, P = 0.209 for the total EF by the bi-plane area-length method vs. the SAX method, P = 0.064 for the conduit EF area-length in the 4ch method vs. the SAX method, P = 0.073 for the booster pump EF area-length in the 4ch method vs. the SAX method, P = 0.369 for the total EF area-length in the 4ch method vs. the SAX method, except P = 0.002 for the booster pump EF by the bi-plane area-length method vs. the SAX method). The absolute RA volume was larger in males than in females (All P < 0.001), and this difference persisted for a number of methods after indexing by BSA (P = 0.021 for indexed RAV p−ac by the SAX method, P = 0.001 for indexed RAV p−ac by the bi-plane area-length method, P = 0.005 for indexed RAV min by the SAX method, P = 0.001 for indexed RAV min by the bi-plane area-length method, P = 0.011 for indexed RAV min area-length in the 4ch method, except for 0.070 for indexed RAV p−ac area-length in the 4ch method), except for the RV maximal volume index (P = 0.678 for indexed RAV max by the SAX method, P = 0.181 for indexed RAV max by the bi-plane area-length method, and P = 0.142 for indexed RAV max by the SAX method) ( Table 7). RA conduit EF and RA total EF were higher in females than in males by either method (for conduit EF, P = 0.003 by the SAX method, P < 0.001 by the bi-plane area-length method, and P < 0.001 by the area-length method in the 4-ch view; for total EF, P < 0.001 by the SAX method, P = 0.001 by the bi-plane area-length method, and P = 0.010 by the area-length method in the 4-ch view). The RA booster pump EF showed no gender difference (P = 0.092 by SAX, P = 0.152 by bi-plane area-length, and P = 0.660 by the area-length method in the 4-ch view).
Age Related Changes in LA and RA parameters. Correlations between age and parameters of LA or RA are shown in Table 8. Age was mildly to moderately correlated with the size of the LA and RA (|R| from 0.074 to 0.559). Age also correlated positively with LA volume (R = 0.329, 0.518, 0.259 for LAV max , LAV p−ac , and LAV min , respectively, P < 0.001, <0.001, and =0.003, respectively), while it correlated only mildly with RA maximal volume (R = −0.220, 0.061, −0.092, and P = 0.011, 0.479, and 0.288, respectively). There was a negative correlation between age and atrial conduit EF, and a positive correlation between age and atrial booster pump EFs for both atria (All P < 0.001); however total EFs were not correlated with age (P = 0.568 for LA and P = 0.376 for RA).  Inter-and Intra-observer Variability. Inter-and intra-observer variability are shown in Tables 9, 10, and 11. Inter-and intra-observer variabilities in atrial dimensional parameters were moderate. Compared with the SAX method, variability was lower by the area-length method in RA 4-ch view, while it was greater by the bi-plane area-length method compared with other measuring methods.

Discussion
The present study provided comprehensive reference values for the atrial size and function by SSFP sequence in a population of healthy Chinese volunteers with a wide age range. In addition to providing normal reference standard values, we also found that the left or right atrial volume measured by area-length method was considerably lower than that produced by the SAX volume method, and gender and age have a considerable impact on atrial phasic function, especially on the conduit emptying function and booster pump function. CMR is an accurate quantitative tool for ventricular and atrial volume and function, based on multi-slice 2D volume acquisition. The SSFP sequence has high signal-to-noise ratio, good myocardium-to-blood pool contrast, and is used routinely in a clinical setting. SSFP at 3.0T further improved the signal-to-noise contrast and could potentially have high spatial resolution to delineate thin-walled chambers. In recent years, new techniques, such as GRE shimming or short TR, have been introduced to increase the robustness of SSFP at 3.0T 28 and decrease banding artefacts. The normal reference values for atrial size and volume have been investigated at 1.5T in past years in a population of European descent 20,21,29 . Only a recent study in a Singaporean Chinese population reported reference values for the left atrium volume, total left atrial ejection fraction, and right atrium area index measured by CMR at 3.0T 30 . To the best of our knowledge, the present study is the first to investigate both left and right atrial volume and phasic function systematically.
Reference LA dimensions and volume have been studied in normal populations before. However, data derived from earlier sequences, such as TSE or GRE, are not truly comparable to SSFP sequences. Also, data acquired by SSFP sequence with prospective ECG gating not fully covering diastole are not comparable with retrospective ECG gating, which is the current routine in clinical practice 31,32 . Therefore, very limited LA data could be comparable to our present study. We found the LA dimensions to be similar to those reported by Maceira, et al. 20 ; e.g., the upper limit for the LA antero-posterior dimension in the Chinese population was 41 mm, comparable to the 42 mm for those of European descent. The LA absolute maximal volume in our study was lower than that reported for people of European descent, but was accounted for by the BSA. However, the LA maximal volume in our study was lower than that in the Singaporean Chinese population, even after adjusting by BSA (LA maximal volume index: 50 ± 10 mL/m 2 vs. 38.2 ± 10.1 mL/m 2 ) 30 . The reason for this difference is unknown, as the sequence parameters, analysis methods used, and ethnicities of the study population are similar. Left atrial phasic function is a very interesting topic in cardiovascular disease 10,24,33 . However, few previous studies showed normal references in healthy populations. The total LAEF in our study was similar to the data presented in the study by Marceira et al.  Table 7. Gender specific reference values of the RA volume and phasic function assessed by the short axis or area-length method (n = 135). Lower/upper limits calculated as mean ± 2 SD; RAV max , maximal right atrial volume; RAV p−ac , right atrial volume before atrial contraction; RAV min , minimal right atrial volume; Indexed volumes are calculated by the corresponding volume in mL divided by BSA in m 2 ; Conduit EF, Conduit right atrial emptying fraction: 100% × (RAV max − RAV p−ac )/RAV max ; Booster pump EF, Booster pump right atrial emptying fraction: 100% × (RAV p−ac − RAV min )/RAV p−ac ; Total EF, total right atrial emptying fraction: 100% × (RAV max − RAV min )/RAV max .
Scientific RepoRts | 7: 3196 | DOI:10.1038/s41598-017-03377-6 (60 ± 8% vs. 59 ± 8%) 20 . In addition, our study demonstrated gender specific LA phasic function systemically, which was not fully explored in previous studies. Few previous studies investigated RA size and volume. Accurate RA volume is difficult to estimate by 1D or 2D measurements. RA size, as measured by volume, was greater in males than in females, despite similar single dimension measurements in our study. Dimensions derived from the 4-ch view or RV 2-ch view were generally lower than those in previous data and the indexed dimensions were slightly higher than the indexed dimensions in people of European descent 21 . In our study, absolute RA maximum volumes measured on short axis slices were lower than those measured in people of European descent; however, the indexed values were similar (51.6 ± 11.0 mL/m 2 ; versus Sievers's 52.8 ± 16.3 mL/m 2 , and Maceira's 54 ± 10.3 mL/m 2 ) 20, 29 . This was the first study to demonstrate the phasic function of RA in a normal population. A recent study showed that the RA   emptying fraction was an independent and robust indicator for mortality in patients with pulmonary hypertension 16 . This study suggested the potential importance of RA phasic function evaluation in future studies. While SAX method is considered the gold standard for measuring atrial volume without geometric assumption, the area-length method is a simple alternative. Previous comparisons of these two methods based on small normal populations showed good correlation with the LA volume 34 . Our study validated the area-length method further in a Chinese population and demonstrated that the area-length method gives a reasonable estimation of LA volume, although the absolute volume is lower than the true volume, as measured by the SAX method. Left atrial conduit function estimated by the SAX method was significantly higher than that estimated by the area-length method. Therefore, the LA volume and function derived by the area-length method should be interpreted cautiously, especially when these parameters are the main indications for the CMR examination in patients with cardiac remodelling 23 . In contrast to the LA, area-length methods for estimating the RA volume have not been studied in depth. In our study, neither the area-length of the 4-ch nor the bi-plane area-length from 4-ch and RV 2-ch could estimate the RA volume accurately. The RA volume was underestimated remarkably by the area-length method comparing with the SAX method. Thus, if accurate RA volume measurement is necessary, the SAX method is preferred.
Generally, the absolute LA dimensions and volume were greater in males than in females; however, indexing by BSA reduced the differences. The LA maximal volume index was actually larger in females than in males. Our results were in accordance with previous studies 20,21,29,30,32,35 . Right atrial size and volume were greater in males than in females, and these differences also decreased after indexing by BSA, except the RA minimal volume and RA pre-contraction volume. Phasic function was also associated with gender. Males had lower conduit empty fractions (EFs) for both atria. This phenomenon could indicate a gender difference in diastolic ventricular function. The genders have different blood pressures; therefore, whether the conduit empty function is associated with blood pressure or just gender requires further study.
The age related changes in atrial size and function were inconsistent with previous reports. In Sievers's study 32 , LAV max did not correlate with age in healthy volunteers. Maceria et al. 20 also did not find an association between the LA volume with age, and age was not an independent indicator for the LA volume. However, the LA transverse and long dimensions were related with age. In a recently reported Singaporean Chinese population, the LA maximal volume did not correlate with age in either gender 30 . Meanwhile, another study in a large European population, including a younger age range, demonstrated a significant correlation between age and LA maximum volume 35 . The differences among these studies might be caused by a bias in population selection, inconsistent analysis methods for the LA volume (most of these were derived by the area-length method), and limited sample size with different age groups. In our study, we correlated age with LA volume by the SAX method and our population covered a wide age range. Our study confirmed there was a moderate positive correlation between age and the LA maximal volume or LA pre-contraction volume, while the correlation between age and RA maximal volume was only slightly negative. Interestingly, we found a very significant correlation between age and the LA phasic function, which was not observed in previous studies. Both left and right atrial conduit EFs correlated negatively with age, whereas booster pump EFs correlated positively with age. These data demonstrated the impact of age on ventricular diastolic function and atrial remodelling with aging.
In summary, in the present study, we investigated the reference values of the left and right atrial dimension, volume, and phasic function using the state of art SSFP sequence at 3.0T MRI in a healthy Chinese population. The SAX method provided more accurate values for the atrial volume and showed better reproducibility  Table 11. Inter-and intra-variability in right atrial volumes (n = 135). RAV max, maximal right atrial volume; RAV p−ac , right atrial volume before atrial contraction; RAV min, minimal right atrial volume.
compared with the area-length method, especially for the right atrial volume. Therefore, additional short axis slices for the atrium are necessary if the atrial volume is the main question in a clinical study. Indexing to BSA is important to account for certain gender differences. We also demonstrated that age is related to atrial geometry and atrial phasic function. These findings emphasized the potential utility of evaluating atrial phasic function in future studies.