Materials & methods
Population :
Healthy pediatric patients aged 0-18 years referred for murmurs, syncope
or chest pain with normal hearts identified by 2DE during their routine
clinical evaluation were recruited for an additional research
echocardiogram. Clinical data were taken from the medical records and
included age, sex, weight, height, and medical history. Anyone with
insufficient image quality precluding measurement using either software
was excluded. The study was approved by the Institutional Ethics Review
Board.
Real-Time Three-Dimensional Echocardiography :
The 3DE data sets were obtained using iE33 or Epic C7 machines (Philips,
Andover, MA) with a matrix transducer X5 or X7. Apical 4 chamber view
full volume acquisitions to include the entire LV were obtained. Full
volume acquisitions were over 4-7 consecutive beats with no significant
stitch artifact. The same process was completed for the LA.
3D volumetric data analysis :
Uncompressed 3D DICOM datasets focused on the LV or LA were imported
into Tomtec Image Arena. LV and LA end systolic (ES) and end diastolic
(ED) volumes and ejection fractions (EF) were measured using TomTec
Image Arena 3D LV analysis package. A single cardiac cycle was defined
using mitral valve (MV) closure. For the LV, the apex and aortic valve
(AoV) were defined, and automated tracking was adjusted visually against
the 2D imaging planes (four, three and two chamber apical views).
The same 3D DICOMs datasets were imported into VMS software, and
reference points from recreated four, three and two chamber view imaging
planes were manually placed at the AoV, MV, apex and LA and LV chamber
walls as per the VMS protocol at both end systole and end diastole. The
software generates a KBR-derived ESV and EDV and ejection fraction for
each chamber (Figure 1).
The time taken to complete LA and LV volumetric analysis via TomTec and
VMS were recorded.
Statistical Methods :
To adequately assess technique differences across a range of ages, sex
and BSA with a power of 0.8 and alpha of 0.05, and assuming a standard
deviation of 15% within techniques, 101 patients were included in this
study.
The relationship between end-diastolic and end-systolic values with body
surface area was assessed using linear regression modelling, with
Breusch-Pagan testing for heteroscedasticity. Logarithmic transformation
was used to reduce heteroskedasticity where present, and optimal
regression with BSA was identified using curve-fitting.
Analysis time between Tomtec and VMS was compared using a non-paired
Student’s t-test. Linear regression models were used to compare the two
software measurements of LV and LA ES and ED volumes and EF. Intraclass
coefficients (ICC) were calculated, and Bland-Altman plots constructed
for comparison of the two software algorithms. For interobserver
agreement (IOA) and intra-observer agreement (IAOA), we randomly
selected subjects for reanalysis by two of the investigators, LE & AA
(IOA – LE, IAOA AA), and a two-way agreement model with 95% confidence
interval and constructed Bland-Altman plots were used. All statistical
analysis was performed using StataIC 14 (College Station, Texas).