ULTRASONIC BACKSCATTER SYSTEM FOR AUTOMATED ONLINE ENDOCARDIAL BOUNDARY DETECTION - EVALUATION BY ULTRAFAST COMPUTED-TOMOGRAPHY

Objectives. The purpose of this study was to evaluate the accuracy of the recently developed echocardiographic on-line endocardial border de tection system using ultrafast computed tomography, an independent and proved tomographic imaging modality. Background. The automated system for on-line endocardial border detection identifies the blood-tissue interface by acoustic quantification of the ultrasonic backscatter sig nal. Methods. Eighteen subjects were screened by conventional echocard iography and acoustic quantification. Ten of these, with high quality echocardiographic images, were also examined by ultrafast computed tom ography. Comparable image planes at the midpapillary level were analyz ed. Measurements of left ventricular cavity area were compared at end- diastole and end-systole and time course analyses of cavity area durin g the cardiac cycle were performed. Results. There was good correlatio n between values for left ventricular end-diastolic area (r = 0.99), e nd-systolic area (r = 0.93) and fractional area change (r = 0.91) usin g the two methods. The on-line backscatter system underestimated end-d iastolic area (p < 0.001), but the negative bias was small (- 1.6 cm2 ) and the 95% confidence intervals were narrow (-3.6 CM2 to +0.4 CM2). In contrast, the backscatter system overestimated end-systolic area ( p < 0.02); the positive bias for this variable was also small (+2.6 Cm 2) but the confidence intervals were relatively wide (+7.9 to -2.8 cm2 ). The negative bias of backscatter values for cavity area was fairly constant during diastole and early systole (range -5% to -10%), but du ring the second half of systole, backscatter values increased progress ively relative to computed tomographic values. Real time values for fr actional area change measured by the backscatter system were 13% small er than those determined by ultrafast computed tomography (p < 0.001), with wide confidence intervals (+ 3 % to - 30%). Absolute peak rates of area change during systole and diastole were lower by 39% (p < 0.00 1) and 41% (p < 0.01), respectively, using the on-line ultrasonic back scatter system. Time course analyses revealed the errors to be consist ent with cardiac cycle-dependent alterations in gain sensitivity of th e ultrasonic backscatter system. Conclusions. The ultrasonic backscatt er system is associated with cyclic cavity area measurement errors tha t need to be addressed if its early promise for on-line assessment of ventricular function is to be fulfilled. Incorporation of an electroca rdiographically triggered time-varying gain control may improve accura cy for on-line analysis of ventricular performance.