Materials
Six latex balloons were filled with water to represent volumes of LA both in normal physiological conditions and in left atrial enlargement. The balloons were relatively spherical in shape and their walls provided easily distinguishable ultrasound echo for analysis.
Seven human cadaveric LA casts were prepared in the Department of Forensic Medicine, Helsinki University, Finland. No fixation of cardiac tissues was used. Mitral valves and ventricular apices were removed, and the pulmonary veins were clamped. Hearts were suspended from apical portion and left hearts were filled with silicone rubber without extra filling pressure. After hardening of the silicone rubber, the casts were removed from the hearts. LA parts of the casts were separated from ventricular parts at the mitral annular level. These casts were previously used in magnetic resonance imaging studies by Järvinen and Jauhiainen in the 1990s [18, 19]. At that time, no approval of ethics committee was required and the study was approved by head of department.
For this study the silicone casts were transformed to casts made of agar-agar, as casts of silicone rubber are not permeable to ultrasound. Molds for agar-agar casts were made from latex rubber, which was applied on the primary silicone casts while they were stabilized at their mitral planes to a level surface. Four layers of latex rubber was applied on each silicone cast. Left atrial appendices were excluded at this point from the latex molds, as they are usually not included in the volumetric assessment of left atrium by transthoracic ultrasound. The latex molds were peeled from silicone casts and two further layers of latex rubber were used to shut the open left atrial appendix orifices. The latex molds were then positioned so that the mitral openings were facing upwards. A 1,5% agar-agar and water solution was prepared by heating until boiling and then poured into the latex molds at the temperature of 60 degrees centigrade. The molds were cast until the level of mitral annulus. The agar-agar casts were refrigerated overnight and then they were carefully removed from the molds (Fig. 1a).
Determination of cast and balloon volumes
True volumes of the balloons were determined by weighing the water-filled balloons, assuming one gram of weight representing one milliliter of volume. The weight of the balloons before filling with water were measured to be insignificantly low.
True volumes of the agar-agar casts were determined by volume displacement method. A vessel with an opening on its side was filled with water up to the lower level of the opening. The casts were carefully submerged in to the water and the displaced water was collected through the opening into a 100-ml measuring glass. The volume of the displaced water was assumed to be the true volume of the cast. Volumes were determined by this method before and right after imaging to detect whether the cast volumes had been affected by the immersion into water during imaging.
Imaging
GE Vivid E9 machine with 4 V probe (GE Vingmed Ultrasound AS, Horten, Norway) was used for both 3DE and 2DE imaging. The agar-agar casts were immersed in a twenty-liter tank for imaging. Water mixture with dried and crushed seeds of Plantago ovata was used as the imaging medium, as enhanced contrast of the water was necessary for the semi-automated volumetry softwares to function properly. Higher signal intensity of the medium than that of casts represents the relative signal intensities in the in vivo measurements where signal intensity from myocardium and other surrounding tissues is higher than from blood in the atrial cavity. Coarse cloth was placed on the bottom of the tank to attenuate reverberation. The probe was supported above the tank by a tripod so that lens of the probe was 10 mm below the water surface and orienting downwards.
The casts were stationed on a thin (diameter 3 mm) wooden support attached to a pedestal on the bottom of the tank to stabilize the cast during imaging, and the casts were positioned to represent transthoracic apical view. Mitral annular level of the cast was horizontal and thus perpendicular to ultrasound wave propagation. The orifice of left atrial appendix was positioned 60 degrees counterclockwise from the 2DE view so that the first aspect of the imaging represented the apical four chamber view. The ultrasound beam was then electronically rotated 60 degrees counterclockwise to obtain two chamber view. Zoomed 4D view was used to collect 3DE volumetric data over 4 to 6 cardiac cycles, which were defined from electrocardiogram recorded from researcher at a heat rate of 65 to 70 bpm. The gain was optimized by eye for best possible delineation. Recorded volume size and probe frequency were adjusted so that the volume frame rate was 35 to 50 Hz which is the typical acquisition frequency when imaging dynamic volume in vivo.
Water balloons were stationed by a thread and weight to the bottom of the tank so that the center of the balloons was approximately at 100 mm distance from the lens of the probe. 3DE images were collected similarly. No contrast enhancement agent was used during balloon imaging.
Image analysis
Image analysis was performed offline with GE EchoPAC work station software version 112.1.1 (GE Vingmed Ultrasound AS, Horten, Norway) after imaging. GE EchoPAC 4D LVQ (Figs. 1c–f) and TomTec 4D LV-Function (TomTec Imaging Systems GmbH, Unterschleissheim, Germany) (Fig. 1b) softwares were used for volume analysis. Both softwares are designed for cardiac left ventricle volume analysis, but they can also be applied for LA volumetry, as we demonstrate in this study. The researcher was blinded to the weights of the balloons and the results of volume displacement representing true volumes of balloons and casts.
There was a distinct border echo in the ultrasound images at the interface of water as medium, balloon wall, and water inside the balloon. This border echo thickness in the images was clearly greater than that of the balloon true walls (< 1 mm). To assess the most appropriate approach of volumetric measurement, the borderline was placed on the inner (Fig. 1e) and outer rims (Fig. 1f), and to the middle of this border echo by GE 4D LVQ. Automation of TomTec software assumed the borderline to the inner aspect of the rim, so only this approach was used by TomTec software.
Measurements from the ultrasound images representing apical view during cast imaging were 1) 3DE volumetry by both softwares, 2) LA cast cross section areas from four and two chamber views and 3) the greatest length of these cross sections from middle of mitral orifice area to the roof of the atrial casts. The four and two chamber areas and their respective lengths were used to calculate approximations of volumes by biplane area-length method (A-L method) by equation \( \frac{8}{3\pi}\times \frac{4 ch\_ area\times 2 ch\_ area}{length} \), where length is the shorter of the two measured lengths. Repeatability was tested for 3DE volumetry and A-L method of left atrial casts by a time interval of two weeks between repeated measurements.
Statistical and data analyses
IBM SPSS for Macintosh version 24.0 (Armonk, NY, USA) and Microsoft Excel for Mac version 15.26 (Microsoft Corporation, Redmond, WA, USA) were used for statistical and data analyses. Mean and range of true and measured volumes of balloons were determined. Paired differences of measured volumes to true volumes were calculated. Mean of these differences were considered bias. 95% confidence intervals (95% CI) for bias and limits of agreement (LOA), defined as bias ±1.96 standard deviations, were calculated. LOA represent the range in which 95% of measured volumes differ from true volumes when normal distribution is assumed. Pearson correlation coefficients and their statistical significances were calculated. The same statistical methods were applied for repeated volumetric measurements in addition to intraclass correlation coefficients and their statistical significance. Two-way mixed testing for absolute agreement was used for intraclass correlation. Bland-Altman difference plots were used to visualize data.