How accurate is the ultrasonic estimation of ventricular septal defect size?

The relationship between the size of ventricular septal defects estimated from ultrasonic images and the actual size of the ventricular septal defects is currently unknown. Our goal was to image simulated defects of known size under conditions as physiological as possible with several two-dimensional ultrasonic instruments. The study was not viewed as a contest between instruments. A static heart model we tested consisted of a simulated chest wall-right ventricular anterior wall, a human septum or muscular septal equivalent, and a simulated left ventricular posterior wall. Model tissues were placed in holders in a water bath for imaging at 90 degrees and 180 degrees with respect to the ventricular septal defect. These angles tested the lateral and axial resolution necessary for imaging simulated ventricular septal defects. Actual ventricular septal defect sizes ranged from 5 to 17.5 mm. Images were obtained with a 30 degrees, 3.5-MHz, mechanical sector scanner, a 2.4-MHz range focused phased array, and a prototype transmit and dynamically focused, 3.5-MHz phased array. Image quality of the defect varied at 90 degrees but was generally poor at 180 degrees. For studies conducted at 180 degrees, phased arrays imaged sharp defect edges for larger holes, but the mechanical sector scanner did not. Signal attenuation appeared to be the major limiting factor for axial resolution. For imaging at 90 degrees, the mechanical sector scanner approximated the actual defect size less 4 mm. The prototype phased array scanner approximated the defect size less 4 to 6 mm. With the range focused, phased array, imaged size was always less than one half actual size. The smallest defect imaged with our simulation with any instrument was 7 mm. These data or their regression equation should not be used to compute the actual size of a clinical ventricular septal defect, since different machines, transducers, and situations might alter the exact relationship between imaged and actual defect size. However, the data led to the conclusion that imaged size of a defect is instrument and transducer dependent and is always smaller than actual defect size.