Résumé
Tissue Doppler imaging is an echocardiographic useful method in the assessment of left ventricular myocardial function in the clinical condition. Pulsed Doppler interrogation measures the instantaneous velocities of the myocardium which passes through the sample volume during the cardiac cycle. The present study attempts to verify a computerized method to determine myocardial maximum and minimum velocities throughout the cardiac cycles using spectral pulsed-tissue Doppler imaging. The data of curves might be used to calculate myocardial physical and mechanical parameters throughout the cardiac cycle. Spectral pulsed-TDI was performed to evaluate longitudinal function in 23 healthy volunteers by using a sample volume placed in 170 left ventricular segments. The velocities were extracted automatically based on four common edge detection algorithms using Matlab software. Labeling of connected components in boundary of spectrum allowed comparing the methods. In addition to analysis of variance and t-test, linear correlation and Bland-Altman analysis were calculated to assess the relationships and agreements between the systolic and diastolic results of measurements before and after using the computed program. Comparison of the means of the four edge detection methods showed that there are statistically significant differences between methods [number of labels were 12 3 for Canny, 20 4 for Roberts, 31 4 for Sobel and 39 5 for Prewitt respectively, P<0.05]. There were not significant differences between measured velocities in the segments; before and after application of the Canny method. There was significant correlations [r=0.99 and r=0.96, P=0.01] at the base and mid segments, respectively with Bland-Altman analysis significant agreements between the measurements. It is concluded that the proposed method automatically extracts myocardial velocities using spectral pulsed images. Canny method showed relatively favorable results and seems to be a preferable option to extract velocities from the spectral images. Correlation study and Bland-Altman analysis confirmed a good agreement between the measurements