Automatic Detection of End Systole within a Sequence of Left Ventricular Echocardiographic Images using Autocorrelation and Mitral Valve Motion Detection

• Published in: 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society Vol. 2007; pp. 4504 - 4507
• Main Authors: Kachenoura, N., Delouche, A., Herment, A., Frouin, F., Diebold, B.
• Format: Conference Proceeding Journal Article
• Language: English
• Published: United States IEEE 01.01.2007
• Subjects: Autocorrelation; DICOM; Echocardiography - methods; Electrocardiography; Female; Heart Ventricles - diagnostic imaging; Heart Ventricles - physiopathology; Humans; Image processing; Image Processing, Computer-Assisted - methods; Image segmentation; Image sequences; Male; Mitral Valve - diagnostic imaging; Mitral Valve - physiopathology; Motion detection; Motion estimation; Systole; Testing; Valves
• ISBN: 9781424407873; 1424407877
• ISSN: 1094-687X; 1557-170X; 1558-4615
• DOI: 10.1109/IEMBS.2007.4353340

The automatic detection of end diastole and end systole is the first step of any software developed for a fully automatic calculation of the ejection fraction. In this study, methods of image processing were applied to black and white echocardiographic image sequences corresponding to a cardiac cycle and the end systolic image number was automatically estimated. The first method took the advantage of the rapid mitral valve motion to estimate the end systole from the time signal intensity variation in a cavity region defined thanks to three landmarks usually used for the standard left ventricular segmentation. The second method was fully automatic; it was based on the left ventricular deformation during the cardiac cycle. The deformation curve was estimated using correlation and its minimal value was used to detect end systole. Method 3 was a combination of the two previous methods to overcome their limitations. The three methods were tested on a group of 37 patients (four chambers and two chambers apical views). The first image exhibiting the beginning of the mitral opening was considered as the end systolic on the visual readings. Compared with this visual reference reading, a linear regression led to a correlation coefficient r of 0.84 for the first method. This coefficient was improved to 0.87 for the second method and increased significantly to r= 0.93 for the third method.