Donald A. Conetta

Computer-assisted edge detection in two-dimensional echocardiography: Comparison with anatomic data☆

Four methods of computer-based edge detection were evaluated for identification of endocardial and epicardial borders on 2-dimensional echocardiograms of excised hearts. A method was also evaluated for observer identification of cardiac borders in the same hearts. The accuracy of computer-derived borders and of observer-derived borders were determined by comparison to anatomic borders measured from photographs of slices of the excised hearts. Echocardiographic borders were compared with anatomic borders by calculation of segmental cavity areas and wall thickness. Each of the methods tested (computer and observer) identified endocardial borders accurately, resulting in cavity segment areas that correlated well with the corresponding anatomic data (r = 0.90 to 0.92). Regional wall thicknesses correlated less well with anatomic data (r = 0.74 to 0.80), suggesting that endocardial borders were identified more accurately than were epicardial borders. Thus, the methods of computer-assisted echocardiographic border detection tested identified the endocardium and epicardium as accurately as a trained observer evaluating unprocessed echocardiograms. Computer-based methods of border detection may be useful in the automated analysis of clinical echocardiograms.

A framework for three-dimensional time-varying reconstruction of the human left ventricle: sources of error and estimation of their magnitude.

Abstract This report summarizes development of a computer/ultrasound system to graphically reconstruct the contracting human left ventricle in three dimensions and calculate indices of cardiac performance. The three-dimensional reconstruction is performed by realignment of five cross-sectional two-dimensional echograms along a single longitudinal section. The spatial position of the cross sections is recorded by a specially developed indexing arm. Cross sections are then realigned perpendicular to the long axis and parallel to each other by the computer. Points at 30° intervals are chosen from the inner and outer muscle boundaries of a cross section and connected by straight lines. Thus areas can be calculated simply by summation of triangular areas with vertices at the origin. Volumes can then be calculated using a modified Simpsons rule. To test these computer programs three hypothetical time-varying computer-generated left ventricular models were developed. Effects of six major anticipated sources of error were determined by systemic introduction of simulated measurement errors into data for these hypothetical models. These results demonstrate that this system is practical for the three-dimensional reconstruction of the left ventricle, that volumes and derived indices of ventricular performance can be calculated, and that anticipated sources of error result in relatively small deviations from true values.

Evaluation of a three-dimensional reconstruction to compute left ventricular volume and mass☆

This study tests the accuracy of a model to calculate left ventricular volume (LVV) and muscle volume (MV) when optimal data were used. These volumes were calculated using endocardial and epicardial borders traced from photos of cross sections of 20 animal (dog, goat and pig) hearts. A pyramid summation algorithm was used to perform a 3-dimensional (3-D) reconstruction based on 5 short-axis views, thus providing computer volume estimates. These were compared with the true (T) ventricular volumes determined by water filling of the cavity and the true MV based on weight. Because each heart was sliced in 5 planes, the appropriateness of the algorithm for MV could be tested for 6 regions. The pyramid summation algorithm consistently underestimated MV at the base and apex, but was accurate from the midmitral valve to the inferior papillary muscle region. Consequently, the total MV was computed as the midventricular MV, plus base and apex volumes computed from regression equations. Results showed that 3-D reconstruction resulted in a regression of LVVT = 1.02LVV3D + 10.30 ml with r = 0.987 for the chamber of MVT = 1.05 MV3D - 9.78 ml, with r = 0.967. It is concluded that the pyramid summation algorithm can accurately estimate volumes from spatially registered short-axis data with 95% prediction limits about the mean of the data of +/- 10 ml for left ventricular chamber volume and +/- 17.6 ml for MV.

Reproducibility of left ventricular area and volume measurements using a computer endocardial edge-detection algorithm in normal subjects.

The variability of serially recorded 2-dimensional echocardiograms in normal subjects was determined. During a 2-week period, 10 normal subjects underwent echocardiography 5 times, in 2 laboratories, with use of different ultrasonographs. The video recordings were analyzed using a computer image analysis system (Quantic 1200) to provide standardized left ventricular short-axis areas and area ejection fraction (EF). Left ventricular volumes and volume EF were calculated. The 95% confidence limits of the percent difference for end-diastolic area and volume between 2 samples in a given subject were +/- 16.8 and +/- 16.7%, respectively. The limits for end-systolic area and volume were +/- 15.8 and +/- 17.0%, respectively. The 95% confidence limits for differences of area and volume EF between 2 recordings were +/- 12.8 and +/- 9.7%, respectively. No correlation was found between clinical grade (image quality) and the variability of area measurements. A good correlation (r = 0.98) was found between area and volume EF for any given subject over the 5 observations. These confidence limits are narrower than those previously recorded.

Sublingual nifedipine: Acute effects in severe chronic congestive heart failure secondary to idiopathic dilated cardiomyopathy

Nine patients with chronic, severe (New York Heart Association class III to IV) congestive heart failure were studied to determine the acute effects of 10 mg of sublingual nifedipine on left ventricular (LV) function. Hemodynamic and echocardiographic data were obtained at rest and 30 minutes, 1, 2, 4 and 6 hours after nifedipine. Measurements at rest reflected LV dysfunction with elevation of end-diastolic volume index (102 +/- 46 ml/m2), pulmonary capillary wedge pressure (17 +/- 8 mm Hg), systemic vascular resistance (1,547 +/- 439 dynes s cm-5) and reduction of cardiac index (2.8 +/- 0.5 liters/min/m2). There were no adverse effects noted with administration of sublingual nifedipine. Initial changes through 1 hour reflected an unloading effect of nifedipine with reduction in pulmonary capillary wedge pressure (11 +/- 5 mm Hg) (p less than 0.05), systemic vascular resistance (1,179 +/- 289 dynes s cm-5) (p less than 0.01), end-diastolic volume index (91 +/- 37 ml/m2 [difference not significant]) and an increase in cardiac index (3.6 +/- 0.7 ml liters/min/m2) (p less than 0.01). Subsequently the cardiac index, systemic vascular resistance and end-diastolic volume index returned toward baseline. Only the pulmonary capillary wedge and pulmonary artery pressures demonstrated a sustained reduction through the 6-hour study period suggesting an effect of nifedipine on LV relaxation. Thus, sublingual nifedipine administered acutely to patients with clinical congestive heart failure is a safe and efficacious vasodilator.

Autonomous Epicardial and Endocardial Boundary Detection in Echocardiographic Short-Axis Images

An autonomous endocardial and epicardial boundary detection (ABD) method is reported. One hundred ten cycles from 55 clinical studies were selected retrospectively. Image sequences were digitized at 512 x 480 pixel resolution. The point-by-point boundary positions of the ABD and the areas enclosed were compared with positions and enclosed areas drawn by three independent observers. Correlation coefficients for epicardial end-diastolic (ED) and end-systolic (ES) areas, endocardial ED and ES areas, muscle area, and fractional area change were 0.970, 0.976, 0.951, 0.985, 0.887, and 0.878, respectively. Bland-Altman analysis showed negligible biases with standard deviations comparable to those of the observers. The mean difference between the ABD border and the consensus observer border positions in 64 directions falls within the mean range of interobserver border positions, suggesting that shape is also well defined by the ABD.

A Method for Evaluation of Enhancement Operations in Two-dimensional Echocardiographic Images

A means of estimating the degree of enhancement of structure and suppression of background noise in filtered two-dimensional echocardiographic images is described. The method is termed the peak-to-background ratio. To test the method, two-dimensional short-axis echocardiographic images were enhanced with Laplacian operations of increasing mask size. There was excellent correlation between the calculated peak-to-background ratio and the subjective opinion of trained echocardiographers. Furthermore, radial length measurements made from images that were thought to be optimally enhanced by the peak-to-background ratio calculation showed the lowest interobserver mean differences. We conclude that the peak-to-background ratio does reflect improvement in characteristics of the image that favor more precise measurement (amplification of peaks and suppression of background) and can be used to help guide a dynamic approach to image processing.

In vitro analysis of boundary identification techniques used in quantification of two-dimensional echocardiograms

Echocardiographic cross-sectional areas of 10 formalin-fixed animal left ventricles were determined by 5 independent observers using black-white (B-W) and mid-spot (M-S) endocardial boundary identification techniques. The echocardiographic cross-sectional areas were compared with the true anatomic cross-sectional areas of the same 10 hearts to determine the accuracy, variability and reliability of each technique. The results of these comparisons revealed that the M-S technique was more accurate than the B-W technique (3.3 +/- 7.2 vs 34.9 +/- 8.6% error). However, the B-W technique was more reliable in that it had a smaller interobserver and estimated intraobserver variability. The M-S technique had a 6% greater intraobserver variability.

Automatic identification of papillary muscles in left-ventricular short-axis echocardiographic images

An automatic method for identifying the location of the papillary muscles in two-dimensional (2-D) short-axis echocardiographic images is described. The technique uses both spatial and temporal information to identify the presence and track the location of the muscles in the left ventricle from end-diastole to end-systole. The three main steps of the method are spatial preprocessing, spatial processing, and temporal processing. The spatial preprocessing step includes a region of search estimation. The spatial processing step includes a papillary muscle existence test and an initial approximation of the papillary muscle points. The temporal processing includes motion-pattern evaluation and final papillary muscle location, The estimates of existence and position for the automatic method were compared with estimates made by an independent expert observer. Two hundred and ten frames, three taken from each of 70 image sequences, were evaluated. Since two regions of search were processed for each frame (one for the posterior-inferior and one for the anterior-lateral papillary muscle), a total of 420 approximations were made. Of this total, 340 automatic estimates were judged to be in close agreement with estimates made by the expert. Of the remaining 80 approximations, 54 estimates were made by the expert when the computer determined that no papillary muscle was present, 17 estimates provided poor results, and nine estimates were made by the computer when the observer concluded that no papillary muscle was present.

Correlation of paradoxical atrial septal motion and an interatrial pressure gradient in severe tricuspid regurgitation

Abstract The anatomic and hemodynamic sequelae of tricuspid regurgitation (TR) have been evaluated by M-mode and 2-dimensional (2-D) echocardiography. 1 Tei et al 2 observed abnormal (paradoxical) motion of the atrial septum toward the left atrium during systole on 2-D and M-mode echocardiograms. 2 These investigators postulated that the paradoxical motion of the atrial septum was a result of the TR-induced volume overload of the right atrium and an abnormal right atrial-left atrial (RA-LA) pressure gradient. However, Lin et al, 3 in a study of atrial septal motion in LA and RA volume overload states, contended that normal and abnormal atrial septal motion can be explained wholly by LA volume change. The patient described herein, who had severe TR and paradoxical motion of the atrial septum, documents an RA-LA pressure gradient and compares this gradient with RA-LA area changes (reflecting volume changes) during the cardiac cycle.