Three-Dimensional Echocardiography Can Simulate Intraoperative Visualization of Congenitally Malformed Hearts Michael Vogel, MD, Siew Yen Ho, PhD, Christopher Lincoln, FRCS, Magdi H. Yacoub, FRCS, and Robert H. Anderson, MD National Heart and Lung Institute, Royal Brornpton Hospital, London, United Kingdom

Background. The new technique of three-dimensional echocardiography can display the studied anatomy in any desired view plane. We sought to establish whether the reconstructions produced could provide views of the heart comparable with those obtained by the surgeon in the operating room.

Methods. Ninety-four patients, aged 1 day to 19 years (mean, 4.3 years), were examined. The ultrasound probe was placed either on the chest or subcostally and ac- quired 80 to 100 perpendicular parallel images of the heart with electrocardiographic and respiratory gating. Any plane, in particular oblique planes, within the data set can be analyzed. Whenever possible, the arrangement as seen by the surgeon was photographed, or heart specimens with similar intracardiac lesions were cut to simulate the view of the surgeons, to validate the echo- cardiographic reconstructions.

Results. Three-dimensional reconstruction of peri- membranous ventricular septal defects, atrial septal de-

fects, or anomalies of the atrioventricular valves could be displayed as viewed through an atriotomy. In similar fashion, reconstructions of muscular or doubly commit- ted ventricular septal defects, along with obstruction of the right ventricular outflow tract, could be prepared as seen through a right ventriculotomy. Obstruction of the left ventricular outflow tract was shown as viewed through an aortotomy. Transthoracic three-dimensional echocardiography provided additional information in the prospective diagnosis of supravalvar mitral mem- brane, doubly committed subarterial ventricular septal defect, and subaortic stenosis caused by a restrictive ventricular septal defect in double inlet left ventricle.

Conclusions. Three-dimensional echocardiography can simulate the display of the heart as seen by the surgeon in the operating room, and therefore can aid in better planning of surgical repair.

(Ann Thorac Surg 1995;60:1282-8)

C onventional cross-sectional echocardiography is fre- quently used as the noninvasive diagnostic method

of choice for delineation of the presence and nature of congenital heart defects [1]. This technique has proved sufficiently successful that many patients now undergo operation based on echocardiographic findings alone [2]. Although the presence and the anatomy of many con- genital heart defects can be diagnosed readily by cross- sectional echocardiography, the anatomy is rarely dis- played in views that are similar to the ones encountered intraoperatively. Three-dimensional echocardiography by computer-controlled acquisition of multiple sequen- tial cross-sectional echocardiographic scans [3] has the potential of displaying the intracardiac anatomy in views that are similar to the ones encountered during opera- tions [4]. The purpose of this study was (1) to determine the usefulness of three-dimensional echocardiography in displaying the anatomy of congenital heart defects in

Accepted for publication June 15, 1995.

Doctor Vogel's present address is Deutsches Herzzentrum Berlin, Au- gustenburger Platz 1, D-13305 Berlin, Germany.

Address reprint requests to Dr Ho, Department of Paediatrics, National Heart and Lung Institute, Dovehouse St, London SW3 6LY, England.

views simulating intraoperative visualization and (2) to compare the echocardiographic views with anatomic specimens of similar heart lesions or direct photographs of the intraoperative arrangement.

Mate r i a l and M e t h o d s

A commercially available Vingmed 800 (Vingmed Inc, Horten, Norway) annular array sector scanner was used interfaced with a Tomtec (Tomtec Inc, Munich, Germany) computer to generate three-dimensional reconstructions derived from multiple two-dimensional echocardio- graphic cross-sectional images. The extra cost of the Tomtec system used as an add-on machine to the stan- dard ultrasound scanner is about $100,000 (US$). A stan- dard ultrasound transducer is mounted inside a scan frame and moved longitudinally over a distance of up to 5.9 cm in 0.5-mm increments. A cross-sectional echocar- diographic scan is performed at each step, thus generat- ing one tomographic slice of the data set (Fig 1). Electro- cardiographic and respiratory gating is achieved with standard electrodes [3]. The external scan frame contain- ing the transducer was positioned either in the subcostal area or on the chest [4]. After performing a complete

© 1995 by The Society of Thoracic Surgeons 0003-4975/95/$9.50 SSDI 0003-4975(95)00615-R

Ann Thorac Surg VOGEL ET AL 1283 1995;60:1282-8 3D-ECHOCARDIOGRAPHY DISPLAYS INTRAOPERATIVE VIEW

j r - -

Ns~SCANFRAME ~

I MO~

RR INTERVAL STEERING LOGIC

RESPIRATORY PHASE AND RATE

Fig 1. How multiple sequential parallel cross-sections of the heart are acquired suitable for three-dimensional reconstruction. The scan- frame with the transducer is positioned over the heart. The trans- ducer is pulled back from the apex to the outflow tract by the step- per motor in 0.5-mm increments. The steering logic of the stepper motor is fed with information on the heart rate (RR interval) and respiratory phase, ensuring that the motor moves the transducer only during expiration and when the RR intereal is within the pre- set limits.

conventional cross-sectional echocardiographic scan, in- cluding Doppler interrogation of all cardiac valves from various transthoracic positions, images suitable for three- dimensional reconstruction were acquired within 3 to 5 minutes. Two or three data acquisitions were usually performed in each patient. The temporally and spatially aligned cross-sectional images were digitally reformatted and subsequently, stored as a volume element or voxel [5] on a 486 33-MHz personal computer with a hard drive capacity of 500 MB.

In addition, during acquisition of data each of the multiple sequential cross-sections of the heart is dis- played in real-time on the monitor of the ultrasound scanner as with a conventional cross-sectional echocar- diogram. All these cross-sections are recorded using the videorecorder of the ultrasound scanner and are re- viewed before reconstruction, as this has proven to facilitate analysis of the data set. Three-dimensional reconstruction, performed off-line, took between 20 and 240 minutes. The cross-sections of the three-dimensional data set, which were displayed on the computer monitor as two-dimensional images, could be rotated in any of the three orthogonal planes. Depending on the experi- ence of the operator, and the anatomy of the congenitally malformed heart, the selection of the optimal cross- sections to display the anatomy took about 15 minutes for a relatively simple lesion, such as isolated discrete sub- aortic obstruction, approximately I hour for atrioventric- ular valvar malformations, and from 2 to 3 hours for complex anatomy, such as tetralogy of Fallot, particularly

if searches were made for additional views such as the one displaying the stenosed pulmonary valve "en face." The selection of the optimal planes is dependent on the operator and requires both knowledge of the anatomy as well as expertise in cross-sectional echocardiography. This process of selection is the most important step in the three-dimensional reconstruction and is the one that largely determines the time required for reconstruction. When a desired cross-section has been found, the com- puter begins three-dimensional reconstruction, which takes between 5 and 30 minutes depending largely on how many views are selected of the cut plane and these can be rotated around any axis. Three-dimensional ef- fects were achieved by a combination of distance, gradi- ent, and texture shading [6]. Distance shading assigns gray-scale values according to the distance of the object from the cut plane. Gradient shading assigns gray values by comparison with surrounding surfaces, and their angle to each other. Texture shading assigns the gray scale according to the intensity of the reflected signal. In all patients with anatomy suitable for surgical correction, view planes were then reconstructed to resemble the perspective of the surgeon in the operating room. The three-dimensional reconstructions can be displayed dy- namically and subsequently stored on magneto-optical disk. The malformations studied in the 94 patients are as follows: ventricular septal defect (VSD), 28 patients (peri- membranous VSD, 12; perimembranous VSD with pul- monary stenosis, 1; muscular VSD, 9; doubly committed juxtaarterial VSD with pulmonary stenosis, 1; multiple VSDs, 5); subaortic stenosis, 21; atrioventricular septal defect, 20; tetralogy of Fallot, 13; Ebstein's malformation, 6; parachute mitral valve, 2; supramitral membrane, 2; and double inlet left ventricle, 2 patients. Thus far, 75 of the 94 patients examined have undergone operative repair. Autopsied examples of selected lesions were cut to correspond to the surgical view plane obtained. In 2 patients, the intraoperative arrangement was photo- graphed directly during operative repair. Restless pa- tients were sedated with 70 mg/kg of chloralhydrate. In 33 patients who had undergone general anesthesia for operation and 16 who had undergone routine general anesthesia for a cardiac catheterization, the technique was performed while the patient remained anesthetized.

Results

Three-dimensional reconstruction was possible in 84 (89%) of the patients. Movement of the patient during acquisition of data (4 patients) or the finding of an unsuitable transthoracic echocardiographic window (6 patients), rendered 10 data sets unsuitable for three- dimensional reconstruction. The oldest patient in whom three-dimensional reconstruction proved possible was a 16-year-old boy with Ebstein's malformation of the tri- cuspid valve. Reconstruction of the heart to simulate the surgical views was not possible in 6 of the patients with atrioventricular valvar abnormalities. Gray structures produced by valvar reverberations obstructed the view of

1284 VOGEL ET AL Ann Thorac Surg 3D-ECHOCARDIOGRAPHY DISPLAYS INTRAOPERATIVE VIEW 1995;60:1282-8

Fig 2. (A) Cross-sectional image (right) and three- dimensional reconstruction (left) in a patient with supravalvar mitral membrane. The cross-section of the data set resembles a conventional four-chamber view. The arrows point to the membrane. (B) Intra- operative photography in the same patient. (ao = aorta; LA = left atrium; Iv = left ventricle; MV = mitral valve.)

A

the valve. Thus, reconstruction of intraoperative views proved possible in 78 of the 94 patients.

Specifically, the mitral valve could be presented as viewed through a left atriotomy by achieving "electronic removal" of the roof of the left atrium. The observer, like the surgeon, is then able to look down on the atrial aspect of the mitral valve (Fig 2). The three-dimensional recon- struction is compared to a photograph obtained during surgical repair in the same patient. In patients with subaortic stenosis, it proved possible to present the narrowing as might be viewed through an aortotomy. The "electronic cut" is made immediately beneath the hinge points of the valvar leaflets so as to provide a view of the subaortic area as would be obtained by the surgeon using retractors to move the leaflets of the aortic valve. The subaortic stenosis can then be seen from its superior aspect (Figs 3, 4). In this particular patient, the subaortic stenosis was caused by a restrictive VSD divided by a muscle bar in the setting of double inlet left ventricle, rudimentary right ventricle, and discordant ventriculoar- terial connections (transposition). This diagnosis had not

been made either by conventional cross-sectional echo- cardiography or by angiography. The three-dimensional echocardiographic findings were confirmed during sub- sequent operation.

Perimembranous VSDs can be displayed as if seen through the right atrium. The right atrial free wall and roof are electronically removed, thus permitting the ob- server to look down onto the atrial aspect of the tricuspid valve and see those parts of the defect that normally are partially hidden by the septal leaflet of the tricuspid valve (Fig 5). Muscular VSDs were displayed as seen through a right ventriculotomy with the observer looking from the right ventricular cavity onto the ventricular septum (Fig 6). In this patient, conventional cross-sectional echocar- diography and angiography had both suggested the presence of two VSDs. Three-dimensional reconstruction showed a singular large defect partitioned by the sep- tomarginal trabeculation. This finding was confirmed at operative repair, which required cutting away the sep- tomarginal trabeculation to expose the entirety of the VSD. In 1 patient, thought to have tetralogy of Fallot,

Ann Thorac Surg VOGEL ET AL 1285 1995;60:1282-8 3D-ECHOCARDIOGRAPHY DISPLAYS INTRAOPERATWE VIEW

Fig 3. Display of one cross-section of the three-dimensional data set of a patient with double-inlet left ventricle, rudimentary right ven- tricle, discordant ventriculoarterial connection, and subaortic steno- sis. The cross-section is similar to a conventional four-chamber view. The black square surrounds the part of the data set that was subsequently reconstructed to produce Fig 4. (Ao = aorta; DILV = double-inlet left ventricle; rud RV = rudimentary right ventricle.)

reconstruction revealed a doubly committed juxtaarterial VSD and valvar pulmonary stenosis. The three-dimen- sional echocardiographic diagnosis was confirmed at operation (Fig 7).

Comment

Our study has demonstrated that processing of transtho- racically acquired multiple sequential echocardiographic

cross-sectional scans can provide three-dimensional im- ages of congenital cardiac malformations that simulate the perspective of the surgeon working in the operating room. Comparison either with autopsy specimens or direct intraoperative inspection confirmed that the re- constructions reflect accurately the anatomy present [7]. Indeed, the reconstructions of a heart under normal workload, which can be displayed dynamically on video- tape, may offer information that more accurately reflects the anatomy of the beating heart as compared to the surgical inspection of the nonbeating heart during oper- ation. Structures that may potentially interfere with the field of vision can be removed electronically during the reconstruction of the data set [8]. The relevant anatomic features may also be displayed in views offering even more information than intraoperative inspection. Acqui- sition of echocardiographic tomographic data suitable for reconstruction was feasible in most infants and children either through the transthoracic window or using a subcostal approach [4, 7, 9].

Although we used this technique in some patients undergoing general anesthesia for reasons other than acquisition of the echocardiographic data, three- dimensional echocardiography remains essentially a noninvasive technique. Nonetheless, to provide data suitable for three-dimensional reconstruction, the cross- sectional echocardiographic data must be of the highest quality. It cannot be used if there is significant motion of the patient during acquisition [4, 5, 7]. To provide optimal reconstruction of the echocardiographic scans, we rou- tinely performed at least two, and usually three, acquisi- tions of data, each of 3 to 5 minutes duration. The main drawback of our present method is not the time needed to acquire the multiple sequential cross-sections that form a three-dimensional data set, but the extensive time needed for reconstruction. Most of this time is spent on

A B

Fig 4. (A) Cross-section (right) and three-dimensional reconstruction (left). The arrows indicate the position of the observer who looks down through the aortic root into the rudimentary right ventricle and onto the ventricular septal defecL The division of the defect (VSD) by a muscle bar can be appreciated. (B) Anatomic specimen of a heart with similar anatomy. The aortic root and the free wall of the rudimentary right ventricle has been opened to expose the subaortic area and permit viewing of the ventricular septal defect from the rudimentary right ventricle (RV). The stars indicate the two parts of this ventncular septal defect divided by a muscle bar.

1286 VOGEL ET AL Ann Thorac Surg 3D-ECHOCARDIOGRAPHY DISPLAYS INTRAOPERATIVE VIEW 1995;60:1282-8

A B Fig 5. (A) Display of one cross-section of a small perimembranous ventricular septal defect (right) and a three-dimensional reconstruction (left). The arrow inside the left ventricle (LV) points to the ventricular septal defect. The three-dimensional reconstruction demonstrates that the view of the ventricular septal defect is partially occluded by the septal tricuspid valve leaflet (STVL). (B) Anatomic specimen of a heart with similar anatomy. The septal valve leaflet is indicated by the closed black arrow. The open arrow points to the small part of the peri- membranous ventricular septal defect that can be seen from the right side of the heart. (Ao = aorta; LA = left atrium; RA = right atrium; RV = right ventricle; VS = ventricular septum.)

choosing the adequate plane for reconstruction. Al- though increasing experience with three-dimensional echocardiography [10, 11] and recent advances in the software packaging have decreased the time needed for reconstruction, currently we still need hours rather than minutes to obtain satisfactory reconstructions. We re- main far from on-line three-dimensional echocardiogra- phy [12].

Transthoracic three-dimensional echocardiography, because it depends on the data acquired, has exactly the same limitations as cross-sectional echocardiography. Thus, extracardiac structures or cardiac structures hid- den in part by lung tissue, such as conduits placed from the right ventricle to pulmonary arteries, cannot be imaged properly [13]. Modification of the hardware used sequentially to steer the transducer and the use of rotational scanning with a transducer in a fixed position rotated 180 degrees around its vertical axis [14], have recently allowed for imaging through the suprasternal window to evaluate the aortic arch. The major advantage of three-dimensional reconstruction is that it displays the intracardiac anatomy in the fashion routinely used by the surgeon for viewing intracardiac structures. Therefore, the technique functions not only as a tool to improve communication between cardiologists and surgeons, but also as an aid for teaching cardiac anatomy. In the future, we anticipate that it will prove invaluable in planning surgical procedures. With our present computer, we can already remove electronically cardiac structures to im- prove visualization [11]. In the future, with the develop- ment of interactive software, we anticipate the potential to simulate operative procedures and view their effect on the cardiac anatomy. Display of the three-dimensional

structures of the heart as holograms may still further enhance our understanding of the cardiac anatomy [15].

The current acquisition of data as well as the three- dimensional reconstructions were performed by the same person who had expertise in cross-sectional echo- cardiography and was working full-time in applying and evaluating this new technique. As the cross-sections were displayed on the monitor during acquisition and were available for review, it proved impossible to blind the observer performing reconstructions from the raw data obtained by conventional cross-sectional scanning. As with all new technology in the initial phase of its application, the results are highly dependent on the skill of the operator and may not be universally reproducible. The current system of three-dimensional reconstruction from multiple cross-sectional echocardiographic scans is certainly still cumbersome and t ime-consuming to use, and the process of reconstruction requires a substantial leaming curve. Another limitation is the current high additional cost of the off-line computer, which is largely due to the high cost of development of the dedicated software. As with all new computer technology, this cost, today in a range similar to the additional costs of color Doppler when initially introduced about 10 years ago, will eventually decrease. The high expenditure is similar to, or less than, that of alternative imaging techniques such as magnetic resonance, which require a longer time for data acquisition and equally skilled technicians to obtain three-dimensional reconstructions. The full bene- fits of three-dimensional echocardiography can only be evaluated if this technique is used by many centers, which will lead to competition among software and

Ann Thorac Surg VOGEL ET AL 1287 1995;60:1282-8 3D-ECHOCARDIOGRAPHY DISPLAYS INTRAOPERATIVE VIEW

Fig 6. (A) Display of one cross-section of the three- dimensional data set (right) and three-dimensional reconstruction (left) of a large muscular ventricular septal defect (VSD) showing how the data set was electronically cut to generate a surgical view. The arrows indicate the position of the observer in the right ventricle (RV) and the direction of view toward the septum. The boundaries of the ventricular septaI defect are marked by arrows. The septomarginal trabeculation (smt) is partially obstructing the view onto the defect. (B) Anatomic specimen of a heart with similar anatomy. The rather pronzinent sep- tomarginal trabeculation (SMT) divides the ventric- ular septal defect (stars). (LV = left ventricle; RA = right atrium.)

A

u l t r a sound c o m p a n i e s to p roduce eas ier and m o r e rap id recons t ruc t ions .

There fore , we conc lude that this n e w technology , de-

spi te its cur ren t l imitat ions, war ran t s fu r the r use and d e v e l o p m e n t . It p rov ides the su rgeon wi th useful infor- ma t ion on the a n a t o m y of congen i ta l cardiac m a l f o r m a -

A B

Fig 7. (A) Intraoperative arrangement and (B) three-dimensional reconstruction of a heart with doubly committed juxtaarterial ventricular sep- tal defect (VSD). The three leaflets of the aortic valve can be seen in continuity with those of the pulmonary valve.

1288 VOGEL ET AL Ann Thorac Surg 3D-ECHOCARDIOGRAPHY DISPLAYS INTRAOPERATIVE VIEW 1995;60:1282-8

tions and may be an important step to achieving instant real-t ime three-dimensional display of cardiac anatomy.

Doctor Vogel was supported by a grant from the European Society of Cardiology.

References

1. Gutgesell HP, Huhta JC, Latson LA, Huffines D, McNamara DG. Accuracy of 2-dimensional echocardiography in the diagnosis of congenital heart defects. Am J Cardiol 1985:55: 514-8.

2. Huhta JC, Glasow P, Murphy DJ, et al. Surgery without catheterization for congenital heart defects: management of 100 patients. J Am Coll Cardiol 1987;9:823-9.

3. Wollschl~ger H, Zeiher AM, Klein HP, Kasper W, Geibel A, Wollschl/iger S. Transesophageal echo computer tomogra- phy: a new method for dynamic 3-D imaging of the heart [Abstract]. Circulation 1989;80(Suppl 2):569.

4. Vogel M, L6sch S. Dynamic three-dimensional echocardiog- raphy with a computed imaging probe (echo-CT). Initial clinical experience with transthoracic application in infants and children with congenital heart defects. Br Heart J 1994; 71:462-7.

5. Marx GR, Fulton Dr, Pandian NG, et al. Delineation of site, relative size and dynamic geometry of atrial septal defects by real-time three-dimensional echocardiography. J Am Coll Cardiol 1995;25:482-90.

6. Cao Q-L, Pandian NG, Azevedo J, et al. Enhanced compre- hension of dynamic cardiovascular anatomy by three- dimensional echocardiography with the use of mixed shad- ing techniques. Echocardiography 1994;11:627-33.

7. Vogel M, Ho SY, Anderson RH. Comparison of three- dimensional echocardiographic findings with anatomic

specimens of various congenitally malformed hearts. Br Heart J 1995;73:566-70.

8. Pandian N, Roelandt JR, Nanda NC, et al. Dynamic three- dimensional echocardiography: methods and clinical poten- tial. Echocardiography 1994;11:237-59.

9. Vogel M, L6sch S, Biihlmeyer K. The application of transtho- racic dynamic three-dimensional echocardiography by com- puter-controlled parallel slicing in patients with fixed sub- aortic obstruction. Cardiology in the Young 1994;4:7-14.

10. Schwartz SL, Cao QL, Azevedo J, Pandian N. Simulation of intraoperative visualization of cardiac structures and study of dynamic surgical anatomy with real-time three-dimensional echocardiography in patients. Am J Cardiol 1994;73:501-7.

11. Pandian NG, Cao QL, Erbel R, et al. A comprehensive approach for image segmentation, cutting planes and dis- play projections in three-dimensional echocardiography: suggested guidelines for clinically useful projections based on multicenter experience in 300 adults and pediatric pa- tients [Abstract]. J Am Coll Cardiol 1994;23(Suppl):9A.

12. Sheikh KH, Smith SW, von Ramm O, Kisslo J. Real-time, three-dimensional echocardiography: feasibility and initial use. Echocardiography 1991;8:119-25.

13. Kandah, T, Kimball TR, Daniels SR, et al. When is echocar- diography unreliable in patients undergoing catheterization for pediatric cardiovascular disease? J Am Soc Echocardiogr 1991;4:51-6.

14. Ludomirsky A, Vermillion R, Nesser J, et al. Transthoracic real time 3-dimensional echocardiography using rotational scanning approach for data acquisition. Echocardiography 1994;11:599-606.

15. Vannan M, Pandian N, Dalton M, et al. Ultrasound holog- raphy of the heart and great vessels: a new direction in 3-dimensional echocardiography using volumetric multi- plexed registration and transmission holography--methods and feasibility [Abstract]. Circulation 1993;88(Suppl 1):1837.