Assessment of Left Ventricular Wall Motion Abnormalities with the Use of Color Kinesis: A Valuable Visual and Training Aid

Transcription

1 Assessment of Left Ventricular Wall Motion Abnormalities with the Use of Color Kinesis: A Valuable Visual and Training Aid Yung-Sang Lau, MBBS, MMed (Int. Med.), Josephine V. Puryear, RDMS, Sandra C. Gan, MD, Michael B. Fowler, MB MRCP, Randall H. Vagelos, MD, Richard L. Popp, MD, and Ingela Schnittger, MD, Stanford, California Accurate interpretation of left ventricular segmental wall motion by echocardiography is an important yet difficult skill to learn. Color-coded left ventricular wall motion (color kinesis) is a tool that potentially could aid in the interpretation and provide semiquantification. We studied the usefulness of color kinesis in 42 patients with a history of congestive cardiomyopathy who underwent two-dimensional echocardiograms and a color ldnesis study. The expert's reading of the two-dimensional wall motion served as a reference for comparison of color kinesis studies interpreted by the expert and a cardiovascular trainee. Correlation between two-dimensional echo- cardiography and the expert's and trainee's color coded wall motion scores were r = 0.83 and r = 0.67, respectively. Reproducibility between reviewers and between operators was also assessed. Interobserver variability for color-coded wall motion showed a correlation of r = Correlation between operators was also good: r = Color kinesis is reliable and appears promising as an adjunct in the assessment of wall motion abnormalities by echocardiography. It is both a valuable visual aid, as well as a training aid for the cardiovascular trainee. (J Am Soc Echocardiogr I997;10: ' The detection of left ventricular, segmental, wall motion abnormalities is the basis for recognition of ischemic heart disease by echocardiography. However, the appreciation of various degrees of abnormality is a learned skill and is difficult to master) The accuracy of echocardiography is good for identification of myocardial ischemia when the studies are viewed by an expert. Methods that facilitate recognition of, and quantify, segmental wall motion abnormalities would be beneficial to patients and to echocardiographers. Color ldnesis is a new method that provides realtime, color-coded display of the left ventricular endocardial motion on sequential frames of two-dimensional (2D) echocardiographic images. In each consecutive frame, a distinct color is applied to pixels that are identified as having changed from blood to tissue in systole. This results in color maps of sequential endocardial motion. 2,a This method may be helpful in the analysis of left ventricular wall motion From ttae Cardiovascular Medicine Division, Department of Medicine, Stanford University Medical Center. Reprint requests: Ingela Schnittger, MD, Cardiovascular Medicine Division, Stanford University Medical Center, 300 Pasteur Dr., Room H-2157, Stanford, CA Copyright 1997 by the American Society of Echocardiography /97 $ /1/80570 abnormalities 4 as the dimensions of the color maps, representative of endocardiai motion, provide semiquantitation of wall displacement. However, this method may be limited by factors that preclude endocardial border detection, as well as by translational motion of the heart and the reference system chosen for tracking endocardial motion. The objectives of our study were as follows: (1) evaluate the usefulness of the new color-coded method in the assessment of left ventricular wall motion abnormalities with particular attention to the less experienced cardiovascular trainee, (2) determine the reproducibility of the technique, and (3) assess the interobserver variability in the interpretation of the images. METHODS Forty-two patients (30 male and 12 female) aged 22 to 78 years (mean years) were enrolled in the study. Each patient had a history of heart failure and had been referred to the echocardiography laboratory for assessment of left ventricular function for clinical reasons. Of twenty-nine patients included in the final analysis, 12 patients had documented coronary artery disease, 14 patients had dilated cardiomyopathy, two patients had valvular heart disease, and one patient had a history of pulmonary embolism. 665

2 666 Lau et al. Journal of the AmericanSocietyof Echocardiography July-August1997 F i g u r e 1 Sequential frames A through C during systole, illustrating progressive color encoded display of the inward movement of the endocardium in a parasternal short-axis view. Echocardiography All patients underwent the standard complete echocardiography study performed in our laboratory, including 2D imaging, pulsed and continuous-wave Doppler velocity measurements and color-flow Doppler imaging, with the use o f commercial equipment (SONOS 2500 ultrasonograph; Hewlett-Packard Company, Andover, Mass.). After the completion of the standard study, color ldnesis studies were next performed by two independent operators. One was a cardiovascular trainee physician and the other was an experienced sonographer. In each patient, the highest possible quality conventional images were first obtained, with optimal delineation of the endocardial surfaces o f the left ventricle in the short-axis view at the papillary muscle level and in the apical four-chamber view. The automatic border detection (ABD) system was then activated. The displayed border was automatically superimposed, in real time, over the endocardial-blood interface in each image frame that was detected by the built-in proprietary algorithm, s# In most patients, adjustments in the time-gain compensation setting and lateral gain control were needed to ensure that the displayed border was properly superimposed on the endocardium. This was judged by eye while activating and deactivating the ABD system. A region of interest was drawn just outside the end-diastolic position of the border and along the level of the mitral anulus to encompass all portions of the left ventricular end-diastolic and end-systolic cavity areas throughout the cycle. The color coding system then was activated for systolic tracking, and the system provided a color-encoded display o f the progressive motion of the endocardial border on sequential frames through a single systolic contraction (Figure 1). On completion of patient enrollment, the echocardiographic studies were interpreted by two reviewers blinded to the patient number and other studies on the same patient. The first reviewer, an experienced echocardiographer, first recorded an assessment of the left ventricular wall motion based on the conventional 2D images, and then reassessed it on the basis of the color coded studies. The second reviewer, a less experienced echocardiography trainee physician, scored the wall motion only on the color coded studies. For the purpose of this study, the American Society of Echocardiography 16-segment models was modified slightly for simplicity, dividing the left ventricle into four segments in the parasternal short-axis view (segment 1: anterior and antero-septal wall; segment 2: inferior and infero-septal wall; segment 3: infero-lateral wall; and segment 4: anteroqateral wall). The apical four-chamber view was divided into three segments (segment 1: septum; segment 4: lateral wall; and segment 5: apex) (Figure 2). A wall motion score was assigned to each of the five segments based on the scoring: 1 = normal, 2 = hypokinetic, 3 = aldnetic, and 4 = dyskinetic.9,: Because segments 1 and 4 were visualized in two echocardiographic views, only the highest score obtained from either view was chosen. A wall motion score index for each patient was calculated as the sum of the scores for the rated segments divided by the number of segments rated. The wall motion score index for each patient was deter-

3 Journal of the American Society of Echocardiography Volume 10 Number 6 Lau et al. 667 up Lp Figure 2 Schematic illustration of a simplified version of the i6-segment model of wall motion classification by the American Society of Echocardiography used in this study, a, Parasternal short-axis scan plane, b, Apical four-chamber scan plane: 1 = antero-septal and anterior walls; 2 = infero-septal and inferior walls; 3 = infero-lateral wall; 4 = antero-lateral wall; 5 = apex. mined by the expert echocardiographer from the conventional 2D images (designated data set A), and this was used as the reference against which the color coded studies were compared. The wall motion score index was then determined by the expert echocardiographer frorn the color coded studies of the first operator (B1) and the other operator (C1). The wall motion score index was also determined by the less experienced echocardiography trainee from the color coded studies of the first operator (B2) and the second operator (C2). Statistical Analysis The segmental wall motion score index was calculated as mean -+ standard deviation. The wall motion score index calculated from review of the 2D images by the expert reader was compared with a wall motion score index obtained from color coded images interpreted by the expert reader, as well as by the cardiovascular trainee, using the Pearson's Product rank order coefficient correlation test. Likewise, r values are given for the interoperator and interobserver variability using the same Pearson's test. ~S~TS Of the 42 patients enrolled in the study, 13 were excluded from the final analysis because of inadequate image quality, which made color coded studies uninterpretable or incomplete. Uninterpretable studies included 10 recordings where the background 2D echocardiogram noise level was high, resulting in identification of less than 75% of the endocardium. Incomplete studics wcre recorded in three patients in whom only the apical four-chamber view could be obtained. Of the 29 remaining patients, 7 patients had only one set of color coded studies (i.e., single operator studies). The results of the wall motion scoring by the expert echocardiographer and by the cardiovascular trainee is shown in Table 1. The average segmental wall motion indexes were , , 2.6 _+ 0.4, 2.2 _+ 0.4, and 2.2 _+ 0.4 for data sets A, B1, C1, B2, and C2, respectively. Figures 3, 4, and 5 illustrate studies with hypokinetic, akinetic, and dysldnetic segmental wall motion. Interoperator Variability The reproducibility of the method is reflected in the coefficients of correlation between the expert's reading of wall motion by color ldnesis of the first and second operator (r = 0.84) and between the trainee's reading of wall motion by color ldnesis of the first and second operator (r = 0.74). Interobserver Variability The agreement in the interpretation of the color kinesis studies between the two reviewers is reflected in the coefficient of correlation between the expert's and the trainee's reading of first operator (r = 0.78) and of the second operator (r = 0.78). The cxpert echocardiographer showed a higher degree of accuracy, as defined by the 2D study, in the

4 Journal of the American Society of Echocardiography 668 Lau et al. ~uly-august 1997 Table 1 Wall motion score index based on 2D study Name Age (yr) Gender Diagnosis A B1 C1 B2 C2! 48 M DCM F DCM M DCM F DCM M DCM M DCM F CAD M CAD F CAD M CAD M DCM F CAD M VHD M DCM M CAD M CAD M VHD M CAD M CAD M DCM M CAD F PE M CAD M DCM F DCM M DCM M CAD M DCM F DCM A, Expert echocardiographer's reading of the color coded study as recorded by the first (B1) and second (C1) operator. (Values by the echocardlography trainee's reading of the color coded study by the first (B2) and the second (C2) operator are given.) VHD, valvular heart disease; DCM, dilated cardiomyopathy; CAD, coronary artery disease; PE, pulmonary embolism. interpretation of the color coded studies than the echocardiography trainee. The expert's reading of the color ldnesis compared with the 2D study showed an r value of 0.83 with the trainee's reading of the color ldnesis compared with the 2D study showing an r value of DISCUSSION Background The assessment of left ventticular wall motion performance remains one of the most fundamental indications for an echocardiographic study. Owing to its subjective nature, however, the accuracy and reproducibility of,the findings will depend largely on the training and experience of the reader? The development of a color coded display of the endocardial motion may enhance the ability of less experienced echocardiography interpreters to detect wall motion abnormalities. Another exciting prospect is the possibility of an objective measurement of wall motion performance using a computer algorithm to quantify the degree of endocardial motion on the color maps. ~ a This may potentially enhance the accuracy of stress echocardiography by allowing quantification of wall motion changes, as well as provide a new method of assessing left ventricular filling patterns. Previous Studies Lang et al. t 1 found in a selected group of patients that color kinesis tracked endocardial motion accurately in more than 91% of left ventricular segments (range 69% to 100%). Their interobserver and intertechnique variability were similar to that found in this report. In another study by Schwartz et al.4 semiquantitative analysis of 178 wall segments were identically graded by 2D echocardiography and color kinesis 74% of the time. All segments imaged were analyzed, but 4 of 22 patients were excluded prospectively because of poor image quality. Our study on the clinical application of color coded endocardial motion detection with emphasis on color kinesis as a training tool also shows that it is readily

5 lournal of the American Society of Echocardiography Volume l0 Number 6 Figure 3 Parasternal short-axis view at the level of the papillary muscles. The infero-septal, inferior, and inferolateral walls are hypoldnetic; the anterior and antero-septal segments are akinetic (arrows). reproducible and with little interoperator difference in the acquisition stage. The good correlation between the interpretation o f the two reviewers, as well as the correlation o f their individual interpretations with those based on the conventional 2D images, suggests the trainee's interpretation approaches that o f the more experienced interpreter. Limitations o f the Technique There are several limitations to the color kinesis technique. First, the automatic border tracldng system seems to become unreliable when the image quality is poor because o f either patient or operator factors. Color coded automatic border detection studies therefore would not be amenable to meaningful interpretation in such circumstances as verified by the 13 studies which had to be omitted from analysis because o f inadequate image quality. We noted that overenthusiastic use o f the time gain compensation and lateral gain compensation can create a "false L a u et al. 669 Figure 4 Apical four-chamber vicar of the left ventricle. Septum appears aldnetic (arrows), apex hypoldnetic and the lateral wall normal. La, Left atrium; Lv, left ventricle. border" that would lead to inaccurate color coded display studies. Second, translational motion o f the heart also affects the accuracy o f the method. When the translation is in the same direction as the endocardial motion, it tends to give the illusion o f a greater degree o f endocardial wall motion on the color display. Conversely, it creates the illusion o f a lesser degree of apparent wall motion when translation occurs in the direction opposite to the endocardial motion (Figure 6). These factors partially may account for the differences in the wall motion scores assigned on the conventional 2D images by the expert reviewer as compared with the color coded studies. This may possibly explain why the interpretation o f the trainee correlates better with the expert's interpretation o f color coded images than with the expert's interpretation o f the conventional 2D images. Third, a further limitation is an extension o f the concern expressed previously regarding translation. If there is rotation out o f the plane of imaging, such

6 670 Journal of the American Society of Echocardiography July-August 1997 L a u et al. a) b) Figure 5 Parasternal short-axis view at the level of the papillary muscles. The infero-septal and inferior walls were thin and akinetic as shown in the 2D image in a. When the "dysldnesis mode" of the color ldnesis was activated in b, a red outline of these segments illustrates dysldnetic motion. rotation would be difficult to comprehend on such a study. Sufficient translation o f the heart during contraction in the direction opposite myocardial motion may spuriously create what appears to be dysldnesis through the color coding. Dyskinesis is encoded with a red color on the image when pixels are determined to go from myocardium to blood on sequential images. Although we were not aware o f a specific problem to be mentioned during this study, we have noted such a problem in other cases not included in this series. Fourth, another potential limitation o f this method is the algorithm used for the implementation o f these studies. For example, the length o f systole is rather arbitrarily defined. Color coding occurs during the defined systolic interval. Although one could also set the system to do color coding during diastole only, the question o f timing o f the duration o f systole or diastole remains one for further study. One subjective advantage o f this color coding system is the impression that one can appreciate the rate o f displacement oft_he cndocardial border by observing thc thickness o f each color band on the endsystolic image. Because the image frames occur in sequence at equal time intervals, the width o f the color band represents the displacement per unit time. Thus, what some echocardiographers call "tardokinesis" or delayed motion appears to be discernible on these studies. Limitations o f Study Design Patients recruited to the study were referred to the Echocardiography Laboratory for clinical evaluation o f left vcntricular function. All had a history o f heart failure and congestive cardiomyopathy. This group o f patients may potentially have been easier to image than the average patients with segmental wall motion abnormalities. Although nearly half our patients had a history o f coronary artery disease, our study population could have affected the overall correlation favorably. The use o f a simplified five-segment model (using

7 Journal of the American Society of Echocardiography Volume 10 Number 6 a) L a u e t al, 671 b) Figure 6 Two pictures, a and b, from the apical four-chamber view of the same patient illustrating a case of translation artifact. In a, is shown correctly aldnesis of the septum and apex and near normal lateral wall motion. A few frames later, during respiration, the heart moved laterally resulting in a net motion of the septat endocardium inward, giving the appearance of improved septal motion while the lateral wall endocardium had a net decrease in inward motion and hence appeared aldnetic. L, Lateral wall; s5 septum. one parasternal short-axis view at the papillary level and the apical four-chamber view) instead o f the standard 16-segment model may potentially have improved the level o f agreement between 2D echocardiographic wall motion scoring and color encoded wall motion scoring. An apical four-chamber view was chosen instead o f the parasternal short-axis view o f the mitral valve because o f the sometimes challenging task o f color encoding wall motion in this latter view because o f interference from motion o f the mitral valve apparatus. Correlation coefficients between observers, operators, etc. were calculated on the basis o f averaged wall motion scores rather than comparing individual wall motion segments. Averaging could potentially yield a "falsely too high" correlation. However, when the statistical analysis was applied for each segment individually, some correlations came out better than the averaged ones and some worse. Because wall motion between segments in a given ventricle is not entirely independent o f each other, especially in this group o f stable, chronic cardiomyopathy patients, the averaged scores may be a more independent variable than the wall segments alone and more representative of the actual correlations. Because o f the design o f this study, one can only assess how well the color coded studies correlated with interpretation o f 2D images. I f the color coded automatic border detected studies were superior to, or more accurate than, interpretation of standard 2D images, one could not discern that from this study design. It is the impression of both the expert echocardiographer and the echocardiographic trainee that the color coded studies are easier to interpret than are the standard black and white 2D images. The color cues to decreased motion are especially useful. Conclusions O u r study shows that color coded automatic border detection o f the left ventricular e n d o c a r d i u m is a technique that seems reproducible and apparently reliable as c o m p a r e d with standard 2D studies.

8 Journal of the American Society of Echocardiography 672 Lau et al. July-August 1997 Color coded studies of this type appear to be promising as an adjunct in the assessment of wall motion abnormalities by echocardiography and is both a valuable visual aid and a training aid for the cardiovascular trainee. Nonetheless, its clinical applicability is still conditional on the ability of the operator to acquire high-quality 2D images with good endocardial definition. Although movement of the heart and suboptimal gain control settings may affect the accuracy of the method, they do not seem to have a significantly adverse effect on its overall accuracy and reliability. However, color kinesis cannot, in its present form, substitute for evaluating wall motion and wall thickening by standard gray-scale 2D echocardiography. We acknowledge the contributions of Byron W. Brown, PhD, regarding the biostatistical methods used in this study. We appreciate the expertise of Marcia Gibbs in preparation of the manuscript. REFERENCES 1. Gan SC, Lau S, Frank R, Schnittger I, Popp RL. Colorkinesis: a new method to enhance detection of wall motion abnormalities by echocardiography [abstract]. J Am Soc Echocardiogr 1995;8: Mor-Avi V, Weinert L, Korcarz C, et aft. Assessment of filling patterns in LV hypertrophy using color ldnesis [abstract]. Circulation 1995;92(Suppl):I Vandenberg BF, Oren RM, Lewis J, Burns TL, Keber RE. Color idnesis: evaluation of a new echocardiographic method for analyzing regional wall motion in patients with dilated cardiomyopathy [abstract]. Circulation 1995;92(Suppl):I Schwartz SL, Cao Q-L, Vannan M, Pandian NG. Automatic backscatter analysis of regional left ventricular systolic function using color ldnesis. Am J Cardiol 1996;77: Chenzbraun A, Pinto FJ, Popylisen S, Schnittger I, Popp ILL. Filling patterns in left ventricular hypertrophy: a combined acoustic quantification and Doppler study. I Am Coil Cardioi 1994;23: Gorscan III J, Romand JA, Mandarino WA, Deneault LG, Pinsky MR. Assessment of left ventricular performance by on-line pressure-area relations using echocardiographic automated border detection. J Am Coll Cardiol 1994;23: Barasch E, Wilansky S. Dobutamine stress echocardiography in clinical practice. Texas H Ins J 1994;21: Schiller NB, Shah PM, Crawford M, et al. Recommendations for quantitation of the left ventricle by two-dimensional echocardiography. J Am Soc Echocardiogr 1989;2: Armstrong WF. Stress echocardiography for detection of coronary disease. Circulation 1991;84(Suppi):I Harrison MR, Smith MD, Friedman BJ, DeMaria AN. Uses and limitations of exercise Doppler echocardiography in the diagnosis of ischemic heart disease. J Am Coll Cardiol 1987; 10: LangR,VignonP, Weinert L, et al. Echocardiographic quantitation of regional left ventricular wall motion using color ldnesis. Circulation 1996;93: