Accuracy of three-dimensional volume measurement using biplane transesophageal echocardiographic probe: in vitro...

Two phased-array scanning methods can be used for volumetric transesophageal echocardiographic imaging: (1) pull-back "breadloaf" reconstruction, and (2) rotation "fan-like" reconstruction. The purpose of this study was to (1) test accuracy and precision of pull-back versus rotational geometries for three-dimensional volume determination, and (2) test accuracy of the resulting surface/volume rendered images. The endoscope shaft was inserted into a tube with the handle connected to a stepper device. Seventeen balloons (61 to 471 ml) were put into a water bath consecutively. Two scans were performed: (1) pull-back: the probe was withdraw in 1 mm steps to obtain parallel "breadloaf" slices, and (2) rotational: the probe was rotated in increments of 1.8 degrees, 3.6 degrees, or 5.4 degrees to obtain "fan-like" slices. Each image was digitized for computer analysis. The data were interpolated into 128 x 128 x 128 voxels for three-dimensional reconstruction. Volume measurement was done using a stereometric random marking method. Volumes obtained from the reconstructed images were compared with the true volume (weight) by linear regression analysis. Excellent correlation between measured and actual volumes was obtained from rotation scans as follows: for 1.8 degrees steps (r = 0.9987, SEE = 6.5 ml), for 3.6 degrees steps (r = 0.9959, SEE = 11.5 ml), and for 5.4 degrees steps (r = 0.9943, SEE = 13.5 ml). The pull-back scans showed r = 0.9990, SEE = 5.8 ml. Three-dimensional surface/volume rendered images of the balloons indicate that 1.8 degrees rotation scans are almost as good as 1 mm pull-back scans. We conclude that volume measurements from rotation scans in vivo will not be hindered by scan geometry or software interpolation.