A new integrated system for three-dimensional echocardiographic reconstruction: development and validation for ventricular volume with application in human subjects.

OBJECTIVES

The purpose of this study was to improve three-dimensional echocardiographic reconstruction by developing an automated mechanism for integrating spark gap locating data with corresponding images in real time and to validate use of this mechanism for the measurement of left ventricular volume.

BACKGROUND

Initial approaches to three-dimensional echocardiographic reconstruction were often limited by inefficient reconstructive processes requiring manual coordination of two-dimensional images and corresponding spatial locating data.

METHODS

In this system, a single computer overlays the binary-encoded positional data on the two-dimensional echocardiographic image, which is then recorded on videotape. The same system allows images to be digitized, traced, analyzed and displayed in three dimensions. This system was validated by using it to reconstruct 11 ventricular phantoms (19 to 271 ml) and 11 gel-filled excised ventricles (21 to 236 ml) imaged in intersecting long- and short-axis views and by apical rotation. To measure cavity volume, a surface was generated by an algorithm that takes advantage of the full three-dimensional data set.

RESULTS

Reconstructed cavity volumes agreed well with actual values: y = 0.96x + 2.2 for the ventricular phantoms in long- and short-axis views (r = 0.99, SEE = 2.7 ml); y = 0.95x + 2.9 for the phantoms, reconstructed by apical rotation (r = 0.99, SEE = 2.7 ml); and y = 0.99x + 0.11 ml for the excised ventricles (reconstructed in long- and short-axis views; r = 0.99, SEE = 5.9 ml). The mean difference between three-dimensional and actual volumes was 3% of the mean (3.0 ml) for the phantoms and 6% (4.6 ml) for the excised ventricles. Observer variability was 2.3% for the phantoms and 5.6% for the excised ventricles. Application to 14 normal subjects demonstrated feasibility of left ventricular reconstruction, which provided values for stroke volume that agreed well with an independent Doppler measure (y = 0.97x + 0.94; r = 0.95, SEE = 3.2 ml), with an observer variability of 4.9% (2.4 ml).

CONCLUSIONS

A system has therefore been developed that automatically integrates locating and imaging data in no more time than the component two-dimensional echocardiographic scans. This system can accurately reconstruct ventricular volumes in vitro over a wide range and is feasible in vivo, thus laying the foundation for further applications. It has increased the efficiency of three-dimensional reconstruction and enhanced our ability to address clinical and research questions with this technique.