PubMed2004
In clinical practice, interuser variability, high computational cost and low image quality are always big problems that puzzle the clinical application of computer-aided echocardiographic boundary detection. Star algorithm (StaA) is a new endocardial boundary detector that has been designed to overcome these problems. The purpose of the paper is to present the detection details of the algorithm and evaluate its clinical value. The main elements of StaA are radial search technique, cost function-driven system and self-designed edge detector. The algorithm has four main steps: image preprocessing, initial left ventricular chamber detection, left ventricular chamber center detection and left ventricular endocardial boundary detection. StaA was performed on 50 pairs of end-diastolic (ED) and end-systolic (ES) echocardiographic images, which were divided into high image-quality group (HImQ) and low image-quality group (LImQ). The results of the test were analyzed in two ways: 1. Compared with the manually-traced boundary, the mean relative radial error (MRRerr) of the computer-detected boundary was 12.07% and there is no significant difference of MRRerr between HImQ and LImQ. 2. The two-dimensional (2-D) ejection fraction calculated by the computer detected boundary (EFa) can be used interchangeably with that calculated by the manually-traced boundary (EFm). The test proves that simple and effective methods can also make the echocardiographic boundary detector automatic, quick and robust.