Automatic center point determination in two-dimensional short-axis echocardiographic images

Abstract This report indicates how “matched filter” techniques can be used to provide an accurate approximation of the center of the left ventricle (LV) in a two-dimensional (2D) short-axis echocardiographic image. These filters can be thought of as modifications of the well-known Hough transform. Since the epicardial-pericardial interface along the posterior wall is usually the most reproducible feature in a 2D short-axis image, a “circular arc” matched filter is used to estimate this border location. Once this posterior estimate has been made, a “coupled circular arc” matched filter is employed to estimate the position of the epicardial and endocardial borders of the anterior wall. The estimate of the center of the LV is the midpoint between the estimates of the two epicardial border estimates. The method is used to calculate center points in 36 normal subjects (36 frames at end diastole and 36 frames at end systole). At end diastole, 23 of the computer defined center points are found to be in excellent agreement with the center points defined independently by an expert observer; 9 are in good agreement and 4 are poor. Of the centers calculated in the 36 end systolic frames, 26 are in excellent agreement, 9 are good, and 1 is poor. Thus, these methods are capable of automatically providing accurate estimates of the center of the ventricle for a wide range of images.