Ultrasound Beam Orientation During Standard Two-dimensional Imaging: Assessment by Three-dimensional Echocardiography

Transcription

1 ORIGINAL ARTICLES Ultrasound Beam Orientation During Standard Two-dimensional Imaging: Assessment by Three-dimensional Donald L. King, MD, Michael R. Harrison, MD, Donald L. King, Jr., MS, Aasha S. Gopal, MD, Oi Ling Kwan, RDMS, and Anthony N. DeMaria, MD, New York, New York, and Lexington, Kentucky Standard two-dimensional echocardiographic image planes are defined by anatomic landmarks and assumptions regarding their orientation when these landmarks are visualized. However, variations of anatomy and technique may invalidate these assumptions and thus limit reproducibility and accuracy of cardiac dimensions recorded from these views. To overcome this problem, we have developed a three-dimensional echocardiograph consisting of a real-time scanner, three-dimensional spatiallocater, and personal computer. This system displays the line of intersection of a real-time image and an orthogonal reference image and may be used to assess actual image orientation during standardized two-dimensional imaging when the line-of-intersection display is not observed by the operator. Three hundred forty standard images were assessed from 85 examinations by 11 echocardiographers. Twenty-four percent of the unguided standard images were optimally positioned within ± 5 mm and ± 15 degrees of the standard. Of the optimal images, two thirds were parasternal long-axis views. A subsequent study with three-dimensional echocardiography and line-of-intersection guidance of image positioning showed 80% of the guided images to be optimally positioned, a threefold improvement (p < 0.001). Two-dimensional echocardiography does not achieve reasonably consistent optimal positioning of standard imaging views, suggesting that measurements taken from these views are likely to be suboptimal. Three-dimensional echocardiography that uses line-of-intersection guidance improves image positioning threefold and should therefore improve the accuracy and reproducibility of quantitative echocardiographic measurements derived from these images. (J AM Soc ECHOCARDIOGR 1992; ) From The Cardiology Division, Department of Medicine, College of Physicians and Surgeons, Columbia University, and The Division of Cardiology, College of Medicine, University of Kentucky. Presented in part at the First Annual Scientific Meeting, American Society of, Arlington, Virginia, June 13-15, Reprint requests: Dr. Donald L. King, Cardiology Division (P&S 9-441), Columbia University, 630 West 168th Sr., New York, NY /1/40389 Two-dimensional (2D) echocardiography is extensively used to measure cardiac chamber dimensions and to quantify function. As a means to yield accurate and reproducible measurements, standardized imagmg planes have evolved. I These planes have been defined primarily by visualized anatomic landmarks and are assumed to have a certain orientation when these landmarks are visualized. However, the heart is a three-dimensional structure of complex shape, and these assumptions may not always be valid because of anatomic variation as well as systematic and random variability of examination technique. Although the potential for these variations to influence image position and chamber measurements is recognized in a general way by most echocardiographers, few data are available regarding the frequency and magnitude of their effect on attaining standardized images. This lack of data has been the result of the lack of a practical means for measuring the position and orientation of images. 569

2 570 King et al. Journal of the American Society of We have overcome this deficiency by developing a three-dimensional (3D) scanner with a "line-ofintersection" display. It provides a means for observing the position and orientation of the real-time image with regard to its third, orthogonal, "nonvisualized" dimension. 2,3 The instrument computes and displays the line of intersection of a previously saved approximately orthogonal reference image and the real-time image. This line is displayed in each image and shows the geometric relationship between the two images, as well as the relationship of the realtime image to additional and otherwise nonvisualized anatomic landmarks shown in the reference image. The line of intersection is continuously and rapidly recomputed so that it moves appropriately in each image as the transducer is moved. Although the line of intersection display was created to guide 3D image positioning, it can also serve to assess standardized 2D imaging if unobserved by the operator. The purpose of this study is to perform an assessment of actual image position and orientation obtained during standardized 2D imaging by three groups of echocardiographers at three different institutions. In addition, a subsequent study of one of the three groups compares its performance with 3D line-of-intersection guided imaging to its previous performance with conventional, unguided 2D imaging. METHODS Study Groups Three groups of echocardiographers, two at large university hospitals and one at a SOO-bed community hospital, were studied performing standardized images on a series of subjects. The characteristics of the subjects are summarized in Table 1. Each examiner had greater than 2 years experience. In Group A the 12 subjects underwent 2S examinations, one five times, one four times, two three times, two two times and six were examined once by different echocardiographers. Two echocardiographers each performed seven examinations, and the remaining four each performed five, three, two and one examination, respectively. In Groups B and C each of three echocardiographers performed a single examination on 10 subjects. Each subject was examined only once. The examination procedure was explained and verbal informed consent was obtained from each subject. Instrumentation The 3D scanning system and its operation, including its image acquisition protocol and the line of intersection display, have been previously described 2,3 and shown to be highly accurate without introducing new measurement errors.4 Briefly, the system is composed of a real-time 2D echocardiograph (Model 77020AC, Hewlett-Packard Corp., Andover, Mass.), an acoustic spatial locater (Model GP8-3D, Science Accessories Corp., Stratford, Conn.), and a personal computer (Model AST Premium/286, AST Research, Irvine, Cal.). The acoustic locater is composed of a set of three sound emitters rigidly attached to the echocardiographic transducer and a set of four overhead microphones. The time of flight of 60 KHz sound waves from the emitters to the microphones is measured. The distances are then computed and used to determine the position and orientation of the transducer and its image within a three-dimensional spatial coordinate system defined by the microphone array. The digitized realtime image from the echocardiograph is also transmitted to the computer with its spatial coordinates and used to produce the line of intersection display. To create the display the 3D scanner acquires and saves in computer memory a reference image, for example, a parasternal long-axis image. A subsequent real-time short-axis image will intersect the reference long-axis image creating a line of intersection. By pressing appropriate function keys on the computer keyboard each image can be displayed showing the line of intersection. As the real-time image is moved, the line of intersection moves in both images, showing their changing geometric relationship and the changing relationship of the real-time image plane to anatomic landmarks in its nonvisualized dimension now shown in the orthogonal reference image. This additional information provides the operator a means to guide and document image positioning. Study Protocol Subjects were screened according to quality of the routine echocardiographic study. All examinations were easily performed with the patient in the 45- degree left recumbent position and were of minimal technical difficulty. High-quality images with easy and complete identification of all structures were available in all cases. The 3D line-of-intersection display was not visible to the examiner, who viewed only the conventional 2D real-time image. Images were recorded at end-expiration and end-diastole and printed showing their lines of intersection. An image acquisition protocol was followed to create a reference image on which the position and orientation of intersecting real-time images could be evaluated. First, a parasternal long-axis image was obtained, followed by a parasternal chordal short-axis image, and then a short-axis image through the aortic valve. In

3 Volume 5 Number 6 November-December 1992 Image orientation assessed by 3D Echo 571 Figure 1 Optimal position of the four standard images: parasternal long-axis, parasternal chordal short-axis, apical two-chamber and apical four-chamber views. The white lines of intersection displayed in each image shows the optimal relationship of corresponding orthogonal images. The parasternal long axis-image (upper left) and the parasternal short-axis image (upper right) show their common line of intersection. The reference parasternal short-axis image (lower) shows the optimal position and angulation of the apical two- and four-chamber views (not shown). a second sequence, apical two- and four-chamber views were obtained and followed by a parasternal short-axis image at the chordal level. The purpose of the study and method of evaluation were not explained to the examiners. They were instructed and encouraged to obtain optimal images in each position. It was assumed that the operators knew what constituted an optimal image and coaching or instruction to this effect was not given to assess how well these images were obtained under typical examination conditions. Optimal Image Position Criteria The optimal long-axis image was defined as passing through the center of the aortic valve and the center of the left ventricle as shown on the short-axis images. As shown on the long-axis image, the optimal chordal short-axis image was defined as perpendicular to the left ventricular central long axis and intersecting it approximately 1 cm below the convergence of the mitral leaflets at initial apposition. The optimal position of the apical four- and two-chamber views was defined as being through the center of the ventricle and perpendicular to each other as shown on the chordal short-axis image. The lines of intersection in Figure 1 illustrate these relationships. Analysis The position and orientation of each image was then assessed on its corresponding orthogonal image with a ruler and protractor to measure the displacement and angulation of its line of intersection with respect to its defined optimal image position. Position was considered to be acceptable if the displacement of the

4 572 King et al. Journal of the American Society of PARASTERNAL SHORT AXIS PARASTERNAL long AXIS go'± 15' APICAL TWO CHAMBER APICAL FOUR CHAMBER Figure 2 Schematic diagrams showing the acceptable limits for optimal positioning of images. Top right: Displacement of the parasternal long-axis image ± 5 mm with respect to the center of the ventricle. Top left: Displacement (± 5 mm) and angulation (:±: 15 ) of the parasternal chordal short-axis image with respect to the central axis of the left ventricle 1 cm below the leaflet tips. Lower: Displacement ( ± 5 mm) of the apical two-chamber and apical four-chamber views from the center of the ventricle on the chordal short axis image. Angulation between these two images (90 ± 15 ) is also shown on the chordal short-axis image. image was within ± 5 mm of optimal. Angulation was considered to be acceptable if it was within ± 15 degrees of optimal. The long-axis image was assessed for displacement with respect to the center of the ventricle on the chordal short-axis image. The chordal short-axis image was assessed for angulation and for displacement of its intersection with the left ventricular central axis toward the apex or base, as seen on the parasternal long-axis image. The apical twoand four-chamber views were assessed for displacement from the center of the left ventricle and for angulation as seen on the chordal short-axis image. Figure 2 illustrates the acceptable range of displacement and angulation for these images. Thus the four images in each examination were evaluated with regard to a total of six criteria of displacement and angulation. The chordal short-axis image was considered underrotated when the angle between it and the long axis was less than 90 degrees. The apical views were considered overrotated when the anterior angle between them was greater than 90 degrees. The combined data of Groups A, B, and C included a total of 85 examinations, 340 images, and 510 criteria for evaluation. Chi square analysis of the three groups as a whole and for each criterion was performed. Guided 3D Imaging Study An additional study was performed by the three Group B echocardiographers several months after completion of the initial study to compare their performance with guided 3D echocardiography to their

5 Volume 5 Number 6 November December 1992 Image orientation assessed by 3D Echo 573 Table 1 Subject characteristics LV, Leti: ventricle. Institution N Males Females Mean age Age range Valve disease Ischemia Hypertension Arrhythmia Other diagnosis Normal LV < 55 mm LV > mm A B C B (guided 3D) IS 17 8 IS Table 2 Unguided 2D standardized imaging: Number of criteria attained for optimal image displacement and angulation Displacement Angulation No. Group examinations PLA PSA 4c 2C PSA 2C-4C Total criteria (%) A /150 (51) B / 180 (43) C / 180 (47) Total /510 (47) % PIA, Parasternal long axis; PSA, parasternal short IDS; 4C, apical four chamber; 2C, apical two chamber. Table 3 Unguided 2D standardized imaging: Number of images optimal for both displacement and angulation Group A B C Total % Examinations No. of examinations Total no. of images Total optimal images PLA PSA C-2C* 4(8) 3(6) 3(6) 10(20) 12 Total optimal images % Optimal images PIA, Parasternal long axis; PSA, parasternal short axis; 4C, apical four chamber; 2C, apical two chamber. *Both images optimal in same examination (number of optimal images). previous performance with unguided 2D echocardiography. The criteria for optimal positioning of the four images were defined for them. They were instructed and trained in the purpose and use of the line-of-intersection display for positioning these images. Following the previously described protocol, they were instructed to obtain an image at each position representing the optimal of both "image quality" and correct position. Each of the three technicians performed a single examination on each of 10 patients. Examinations were performed in the same manner as the initial study with the same instrumentation except that the technician observed, and used the line-of-intersection display to guide image positioning. By use of the same criteria and method of analysis as in the initial study, the position and orientation of the images were assessed. Chi square analysis of the group's performance in positioning unguided 2D versus guided 3D images was carried out.

6 574 King et ai. Journal of the American Society of Table 4 Unguided 2D standardized imaging: summary of nonoptimai images Group A GroupB Groupe Mean Displacement - in millimeters (number of images) PLA Medial 8.5 (4) Lateral 8.4 (5) PSA Apical 12.9 (13) Basal 4C Anterior 8.3 (6) Posterior 9.6 (6) 2C Medial 8.2 (II) Lateral 8.6 (3) Angulation-in degrees (number of images) PSA Under 20.3 (IS) Over 4C-2C Under 28.9 (8) Over 33.0 (3) 13.5 (6) 12.5 (4) 13.1 (13) 9.0 (2) 13.1 (10) 11.3 (7) 10.6 (17) 10.7 (3) 26.7 (23) 31.2 (18) 9.5 (4) ll.8 (6) 11.2 (ll) 11.2 (5) 11.7 (9) 14.3 (3) II.5 (8) 17.5 (2) 34.8 (20) 26.5 (2) 45.7 (24) 10.5 (14) 10.9 (IS) 12.5 (37) 10.6 (7) 11.4 (25) 11.2 (16) 10.1 (36) 11.6 (8) 27.8 (58) 28.4 (10) 39.1 (45) PLA, Parasternal long axis; PSA, parasternal short axis; 4C, apical four chamber; 2C, apical two chamber. Table 5 Group B unguided 2D standardized imaging versus guided 3D standardized imaging No. and (%) optimal No. and (%) optimal displacement angulation No. and (%) optimal images 2D 3D 2D 3D 2D 3D PLA 20/30 (67) 29/30 (97) 20/30 (67) 29/30 (97) PSA 15/30 (50) 28/30 (93) 7/30 (23) 25/30 (83) 2/30 (7) 23/30 (77) 4C 13/30 (43) 27/30 (90) 2C 10/30 (33) 26/30 (87) 2C-4C 12/30 (40) 28/30 (93) 6/60* (10) 44/60* (73) TOTAL 58/120 (48) llo/120 (92) 19/60 (32) 53/60 (88) 28/120 (23) 96/120 (80)t PLA, Parasternal long axis; PSA, parasternal short axis; 4C, apical four chamber; 2G, apical two chamber. *Both images optimal in same examination. tchi square statistic = 77.2 (P < 0.001). RESULTS The number of times criteria for optimal image angulation and displacement were attained by Groups A, B, and C for each category are summarized in Table 2. Of a total of 510 criteria, 239 (47%) were attained. The number of images that were optimal for both displacement and angulation are summarized in Table 3. Of the 340 images assessed, 82 (24%) were optimal. Parasternal long-axis images were optimal in 66% of examinations. Parasternal short-axis images were optimal in 7% of examinations. The short-axis image was underangulated «90 degrees) an average of 28 degrees in 68% (58/85) of examinations. The apical two-chamber/four-chamber combination was optimal in 12% of examinations. Fifty-three percent (45/85) of these views were overrotated (>90 degrees) for an average of39 degrees. In 29% of the views (25/85) the fourchamber view was displaced anteriorly and in 19% (16/85) it was displaced posteriorly. In 42% of the views (36/85) the two-chamber view was displaced medially and in 9% (8/85) it was displaced laterally. In 64% of views (54/85) the four-chamber view passed through the posterior septum, whereas in 33% (28/85) it passed through the middle septum. Further, in not one of the 85 examinations were all four images judged as optimal. Table 4 summarizes the mean displacement, angulation, and number of examinations by category for all of the non-optimal

7 Volume 5 Number 6 November-December 1992 Image orientation assessed by 3D Echo 575 images. Chi square analysis of the three groups as a whole showed no significant difference between them. Analysis for each criterion showed no difference between the groups except for rotation of the apical views. For that criterion one group achieved only 13% optimal rotation while the other two achieved 40% and 56%. The results of the subsequent guided 3D imaging study performed by Group B are summarized in Table 5 along with their prior unguided 2D results. Overall, in the guided 3D study 91 % of the criteria for optimal displacement and angulation were fulfilled and 80% of images were judged as optimal. The parasternal long-axis image was optimal in 97% of the examinations, the parasternal short-axis image was optimal in 77% of the examinations, and the apical two-chamber / four-chamber combination was optimal for both displacement and angular rotation in 73% of them. In 53% of the examinations all four of the images were judged as optimal. Chi square analysis demonstrated that the improvement in the number of optimal images achieved by guided 3D echocardiography to be highly significant (p < 0.001). DISCUSSION From these results it is clear that all three groups performing unguided 2D echocardiography did not achieve a high level of optimal positioning of standard imaging views in a "routine" setting. There was no statistically significant difference between the three groups as a whole. Visualized anatomic landmarks alone appear to be inadequate to permit reasonably consistent, optimal positioning of these views. The displacement of one third of the parasternal long-axis images is surprising and suggests the possibility of significant interobserver variation for atrial and ventricular measurements taken in this view. The high percentage of under-angulation of the short-axis view indicates that it should not be used for ventricular measurements because underangulation elongates the anteroposterior ventricular "diameter." Short-axis variability also raises questions about the accuracy and reliability of wall motion analysis in this view. Similarly, the variable angular rotation and displacement of the apical views to each other and to anatomic landmarks raises a serious question about the accuracy and reproducibility of wall motion analysis and volume calculation with these views. Although the assumptions made for the standard imaging views, and measurements taken from those views, may sometimes be valid under optimal circumstances, this study shows that anatomic and technical variation prevent their fulfillment in routine practice. The relatively poor performance of 2D echocardiography positioning images is primarily the result of a technological deficiency of conventional systems rather than any deficiency of the echocardiographer. Echocardiographers generally do their best to perform high-quality examinations. However, 2D echocardiography lacks a means for measuring, assessing, and guiding image position and orientation that is independent of the anatomic landmarks visualized within the real-time image. With the increasing need for accurate and reliable quantitative echocardiography this deficiency must be overcome. We believe that guided 3D echocardiography provides a means to do this and demonstrates its ability to do so in the follow-up study performed with the Group B echocardiographers. In that study guided 3D echocardiography achieved a threefold improvement of image plane positioning compared with the unguided 2D study. The greatest relative improvement was achieved in accurate angulation of the short-axis and apical views. This improvement is directly attributable to the ability of 3D echocardiography to provide an on-line visual method for the operator to observe and adjust the position of each real-time image with regard to its nonvisualized dimension. In addition to adjusting displacement, angulation, and rotation of the real-time image, the display also documents image position, thus leading to improved standardization. There is little in the prior literature concerning the positioning of two-dimensional images. Despite extensive use of 2D echocardiographic imaging for determination of cardiac chamber dimension and wall thickness, significant intraobserver and interobserver variation persists for these measurements. 5-9 Variability of image positioning, as shown in this study, certainly accounts for some, if not most, of this variation. A study by Erbel et al.1o compared apical 2D echocardiographic images obtained in a right anterior oblique equivalent view with simultaneous cineventriculograms also obtained in a right anterior oblique view. They found a systematic underestimation of left ventricular volumes by echocardiography caused by positioning of the ultrasound transducer anterior and superior to the apex in 44 of 46 patients. This position led to a foreshortening of echocardiographic measurement of left ventricular diameters and lengths resulting in an underestimation of volume. In another study Geiser et al., 11 using

8 576 King et ai. Journal of the American Society of a mechanical 3D scanner, reported that short-axis images are not perpendicular to the parasternal long axis. More recently Katz et al. 12 have reported in preliminary form similar data regarding image positioning using an axial rotation transducer. The present study and experience with our 3D scanner confirms these findings and demonstrates the urgent need for improvement if quantitative echocardiography is to provide "hard" numbers for assessing cardiac size and function in the future. 13 It could be argued that the poor 2D positioning results recorded in this study were the result of overly strict criteria and analysis. We believe this is not the case. A tolerated range of position of I em and a range of rotation of 30 degrees does not seem to be an unreasonable or unattainable goal for optimal positioning. In this study minor variations in respiration and obliquity of reference images did occur, but their random variation was the same for all groups and most likely has no significant influence. Group A data were not independent because multiple examinations were done on several individuals. However, overall Group A performance was not statistically different from Group B or Group C. In groups Band C, each technician examined a different subject and each subject was examined only once, providing independent data. In the Group B guided imaging study a "perfect" examination was not always obtained because of limitations imposed by the cardiac windows. It is not always possible to obtain a satisfactory image at the correct position. Many images at an optimal position were diagnostic but of poorer quality than an image at a nearby suboptimal position. Compromise between anatomic representation and image position must be made in some instances. In this situation, when the 3D line-of-inters~ction display was used, the nature of the compromise is dear and documented. Furthermore, we have found that accurate, correctly positioned measurements can often be obtained even when anatomic representation is less than optimal. Because the line-of-intersection display presents a new source of information, the operator must learn new eye-hand coordination skills that require a short period of instruction and practice. When this skill level has been achieved the 3D scanner is easy to use. The addition of the locator device to the transducer has not interfered with its manipulation. Furthermore, 3D capability can be added to any echocardiographic system. Conclusions Standard imaging views obtained by routine 2D echocardiography do not consistently achieve a reasonable level of optimal positioning. Therefore, we infer that measurements taken from these standard 2D views also may be suboptimal. 3D echocardiography with line-of-intersection display guidance achieves a threefold improvement of image positioning and therefore should also significantly improve the accuracy and reproducibility of quantitative echocardiographic measurements. REFERENCES I. Henry WL, DeMaria A, Gramiak R, et ai. Report of the American Society of Committee on Nomenclature and Standards in Two-Dimensional echocardiography. Circulation 1980;62: King DL, King DL Jr, Shao MY. Three-dimensional spatial registration and interactive display of position and orientation of real-time ultrasound images. J Ultrasound Med 1990;9: Gopal AS, King DL, Katz 1, Boxt LM, King DL Jr, Shao MYC. Three-dimensional echocardiographic volume computation by polyhedral surface reconstruction: in vitro validation and comparison to magnetic resonance imaging. J AM Soc ECHOCARDIOGR 1992;5: King DL, King DL Jr, Shao MY. Evaluation of in vitro measurement accuracy of a three-dimensional ultrasound scanner. J Ultrasound Med 1991;10: Crawford MH, Grant D, O'Rourke RA, Starling MR, Groves BM. Accuracy and reproducibility of new M-mode echocardiographic recommendations for measuring left ventricular dimensions. Circulation 1980;61: Stefadouros MA, Canedo MI. Reproducibility of echocardiographic estimates of left ventricular dimensions. Br Heart J 1977;39: Bhatt DR, Isabel-Jones JB, Villoria G1, et ai. Accuracy of echocardiography in assessing left ventricular dimensions and volume. Circulation 1978;57: Abdulla AM, Frank M1, Canedo MI, Stefadouros MA. Limitations of echocardiography in the assessment of left ventricular size and function in aortic regurgitation. Circulation 1980;61: Schnittger I, Fitzgerald PI, Daughters GT. Limitations of comparing left ventricular volumes by two dimensional echocardiography, myocardial markers and cineagiography. Am J Cardiol 1982;50: Erbel R, Schweizer P, Lambertz H, et ai. Echoventriculography-a simultaneous analysis of two-dimensional echocardiography and cineventriculography. Circulation 1983; 67: II. Geiser EA, Lupkiewicz SM, Christie LG, Ariet M, Conetta DA, Conti CR. A framework for three-dimensional timevatying reconstruction of the human left ventricle: sources of error and estimation of their magnitude. Comput Biomed Res 1980;13: Katz AS, Wallerson DC, Pini R, Devereux RE. Visually determined long-axis and short-axis parasternal views and four chamber and two chamber apical views do not represent paired orthogonal projections [Abstract]. J Am Coli Cardiol 1990;15:94A. 13. Harrison MR, King DL, King DL Jr, Smith MD, Kwan OL, DeMaria AN. Ultrasound beam orientation during standardized imaging: assessment by 3-dimensional echocardiography [Abstract]. J AM Soc ECHOCARDIOGRAPHY 1990;3:228.