PII: S0301-5629(02)00560-4

● Original Contribution

REAL-TIME THREE-DIMENSIONAL FETAL ECHOCARDIOGRAPHY -OPTIMAL IMAGING WINDOWS

JING DENG,*†‡ IAN D. SULLIVAN ,‡ ROBERT YATES,‡ MICHAEL VOGEL,‡ DAREN MCDONALD,†

ALFRED D. LINNEY,† CHARLES H. RODECK* and ROBERT H. ANDERSON‡

Departments of *Obstetrics and Gynaecology,†Medical Physics and Bioengineering; and‡Cardiac Unit of Institute

of Child Health & Great Ormond Street Hospital, University College, London, UK

(Received 7 March 2002; in final form 23 May 2002)

Abstract—A total of 15 fetuses were scanned using 2-D array volumetric ultrasound (US). Acquired cardiac datawere converted for rendering dynamic 3-D surface views and reformatting cross-sectional views. The imageusefulness was compared between the data obtained from subcostal/subxiphoid and other imaging windows; theformer are usually free of acoustic shadowing. Of 60 data sets recorded, 12 (20%) were acquired throughsubcostal windows in 6 (40%) patients. Subcostal windows were unavailable from the remaining patients due tounfavourable fetal positions. Of the 12 sets, 6 (50%) provided the dynamic 3-D and/or cross-sectional views ofeither the entire fetal heart or a great portion of it for sufficient assessments of its major structures and theirspatial relationships. Of 48 data sets from other windows, only 9 (19%) provided such 3-D and/or cross-sectionalviews; the lower rate being due to acoustic shadowing. Real-time 3-D US is a convenient method for volumetricdata acquisition. Through subcostal windows, useful information about the spatial relationships between majorcardiac structures can be acquired. However, to offer detailed information, considerable improvement in imagingquality is needed. (E-mail: jdeng@medphys.ucl.ac.uk) © 2002 World Federation for Ultrasound in Medicine &Biology.

Key Words: Three-dimensional and four-dimensional echocardiography, Real-time volumetric ultrasonography,Fetal heart, Cardiac (motion) gating.

INTRODUCTION

Three-dimensional (3-D) ultrasound (US) studies fo-cused on the fetal heart have been reported in recentyears (Chang et al. 1997; Chaoui et al. 2001; Deng et al.1996, 2002; Guerra et al. 2000; Meyer-Wittkopf et al.2001; Nelson et al. 1996; Scharf et al. 2000; Sklansky etal. 1999; Zosmer et al. 1996). Compared to 3-D imagingof many static, solid organs, the natural echogenic con-trast between the blood pool and the cardiac walls greatlyaids 3-D display of the cardiac internal surfaces andchamber shapes (Deng et al. 2001b, 2002; Meyer-Witt-kopf et al. 2001). On the other hand, the rapid heartbeatresults in motion artefacts with the conventional slice-reconstruction 3-D US method, making detailed struc-tural and accurate functional assessments impossible un-less cardiac motion is gated (Chaoui et al. 2001; Deng et

al. 1996, 2000b; Levental et al. 1998; Nelson et al. 1996;Sklansky et al. 1998).

The motion-gated slice-reconstruction method firstmoves a cross-sectional transducer, with both spatial andtemporal tracking, over the fetal heart to obtain a set oftomographic slices. A computer is used to sort the slicesat consistent time points throughout the recorded cardiaccycles (to remove the motion artefact), but from differentanatomical sections to form gated 3-D data sets. Gateddata sets are then rearranged in a cyclical order to forma dynamic 3-D data set that can be reconstructed into 3-Dimages of a beating heart. Although capable of revealingdetailed morphological and unique haemodynamic infor-mation (Deng et al. 2001b, 2002), the method involvesattaching freehand or mechanical tracking devices to theprobe or building a tracking mechanism in the probe and,in some approaches, using another probe from an addi-tional scanner (Deng et al. 2000b, 2001a). Because theacquisition complex is nonpurpose built, quality dynamic3-D data cannot conveniently be acquired with highsuccess rates.

Address correspondence to: Dr. Jing Deng, Department of Med-ical Physics, 1st Floor, Shropshire House, 11-20 Capper Street, LondonWC1E 6JA UK. E-mail: jdeng@medphys.ucl.ac.uk

Ultrasound in Med. & Biol., Vol. 28, No. 9, pp. 1099–1105, 2002Copyright © 2002 World Federation for Ultrasound in Medicine & Biology

Printed in the USA. All rights reserved0301-5629/02/$–see front matter

1099

Real-time 3-D US uses 2-D array transducer tech-nology that can offer a pyramidal imaging volume (rath-er than a conventional imaging plane) at a rate of 20 to40 volumes per second (Light et al. 1998; Ota et al. 2001;Scharf et al. 2000; von Ramm 2000). This appears to bean ideal way to image motion targets in 3-D becausemotion artefacts may no longer be a problem. Mostrecently, it has been investigated in the fetal heart usingreformatted cross-sectional views (Scharf et al. 2000;Sklansky et al. 1999).

In this paper, we report our early experiences inapplying this technique to the prenatal heart via differentimaging windows for data acquisition, and in using dy-

namic 3-D surface displays, as well as reformatted cross-sectional views, for assessment of its clinical usefulness.

MATERIALS AND METHODS

SubjectsThe project was approved by the Ethics Committees

of University College London, and 15 volunteers wererecruited with fetuses aged between 20 and 37 weeks ofgestation. They were referred to the Cardiac Unit atGreat Ormond Street Hospital for detailed echocardio-graphic examinations (Table 1). Only patients with ade-quate cross-sectional images on routine grey-scale inter-

Table 1. Summary of data acquisition and imaging usefulness from different imaging windows

Patient n Gestation (weeks) Referralindications

Conventionalechocardiographic

diagnosis

Data setsacquired

Subcostal/subxiphoidwindows

Otherwindows,

imageusefulness

Real-time 3-Dechocardiographic

diagnosisObtained Image

usefulness

1 23 Echogenicfoci

Normal 3 0 0 0

2 27 CHD in asibling

Normal 2 0 0 0

3 24 Arrhythmia Super ventricularectopics

4 2 1 1 Irregularitynoticed, notclassified

4 35 Extracardiacabnormalities

3 0 0 0

5 21 Diaphragmatichernia

RV � LV 4 0 – –

6 26 Hydramnios Normal 8 4 3 47 25 Maternal MV

regurgitationNormal 4 2 2 2

8 22 CHD in asibling; LVechogenicfoci

Normal, but TVregurgitation

4 1 0 1

9 25 Familyhistories ofclef-lip

TOF 4 1 0 0 Unable todiagnose TOF

10 20 Abnormal4-chamberview

Normal, butenlarged RA

4 0 0 0

11 24 Paternal &siblingCHD

Normal 4 2 0 1

12 37 Arrhythmia II° AVB 4 0 0 0 Irregularity notclearly noticed

13 21 Abnormal4-chamberview

Normal 4 0 0 0

14 21 Down’ssyndrome

AVSD 4 0 0 0 Unable todiagnose AVSD

15 32 Abnormal4-chamberview

AVSD 4 0 0 0 Unable todiagnose AVSD

Total 60 12 of 60(20%)

6 of 12(50%)

9 of 48(19%)

CHD � congenital heart defects; AVSD � atrioventricular septal defect; RV & LV � right and left ventricles; RA � right atrium; TV & MV �tricuspid and mitral valves; TOF � Tetralogy of Fallot; AVB � atrioventricular block.

1100 Ultrasound in Medicine and Biology Volume 28, Number 9, 2002

rogation using a Sequoia C256 (Acuson, Mountainview,CA) were recruited for the purpose of this initial study.

Data acquisitionAll patients were scanned using a real-time volu-

metric US system with a 3.5-MHz 2-D array transducerthat could scan 20 volumes/s (the former VolumetricsMedical Imaging, Durham, NC). Because the transducerwas about the same size as most conventional cardiacprobes, it could be moved freely on the maternal abdo-men to seek optimal imaging windows. This type oftransducer was so designed that it can be placed directlyon the intercostal acoustic windows of a paediatric oradult thorax, and its pyramidal image volume can covera big portion, if not all, of the heart. Because the fetalheart is some distance away from the maternal abdomi-nal wall, any ossified fetal skeletal frameworks (particu-larly in the thorax), if located between the probe and theheart, will cause acoustic shadowing from the late secondtrimester on (Fig. 1). More technical details about thereal-time volumetric imaging can be found in previouspublications (Light et al. 1998; Ota et al. 2001; vonRamm 2000).

To avoid the shadowing, we attempted to scan theheart from the front of the fetal lower thorax and/or upperabdomen. In the former area, most skeletal tissues are notwell-calcified cartilages during fetal life, reducing theacoustic shadowing. In the latter area, there are no bonystructures in the way. For simplicity, imaging accessthrough these paths is called subcostal/subxiphoid win-dows (or subcostal windows for short).

To assess the accessibility of the subcostal windowsin comparison with any other imaging windows, up to 16min were allocated for obtaining imaging windows andrecording dynamic 3-D data sets. Subcostal windowswere first sought and, if obtainable within the first 2 minor so, a subcostal data set was recorded. Then, one of anynonsubcostal windows was sought and a nonsubcostaldata set was recorded within the next 2 min or so. Afterthis, subcostal windows were re-sought and, if obtainablewithin 2 min, a further data set was recorded. In otherwords, the seeking subcostal and nonsubcostal windowswere carried out alternately and each lasted about 2 min.If subcostal could not be obtained within a 2-min period,nonsubcostal windows were located and data sets re-corded. Then, a new search for subcostal and otherwindows was repeated.

Because electrocardiography could not be reliablyperformed in the fetus to synchronise the recording toentire cardiac cycle(s), each recording time was set to1.5 s, covering about two to five cycles for a heart ratebetween 100 and 200 bpm.

Data conversion, visualisation and analysisThe volumetric imaging system only offered multi-

ple, reformatted dynamic cross-sectional views (Figs. 2a,b and 3a). To also obtain dynamic 3-D surface views,data sets recorded without fetal and probe movementswere converted, after patient scanning, into formats thatcould be rendered into dynamic 3-D images on a 4-DEcho Scan workstation (TomTec, Munich, Germany)(Figs. 2c, 3b) and on an MGI 3-D workstation (MedicalGraphics & Imaging Group, University College, London,UK) (Fig. 3c). Both workstations are commerciallyavailable 3-D/4-D graphics systems running on standardMicrosoft Windows NT 4.0 or higher; for more hardwareand software details, refer to previous publications (Denget al. 1996, 2000a, 2001a, 2001b; Meyer-Wittkopf et al.2000, 2001; Vogel et al. 2000; Wang et al. 1996) orcorresponding websites: www.tomtec.de and www.med-phys.ucl.ac.uk/mgi/workstat.htm.

The obtainable rate of subcostal data sets was cal-culated out of the total data sets recorded from subcostaland any other windows. Comparison was made of theusefulness of the data sets obtained between subcostaland any other windows. The usefulness was assessed onimage clarity and volume entirety. The clarity was di-

Fig. 1. An intersection between a plane in the pyramidal im-aging volume and the fetal thorax with the heart, aorta and ribs(small ellipses). Because the heart is some distance away fromthe tip of the pyramid on the maternal abdominal wall, thecalcified ribs will cause acoustic shadowing (dark grey areas)through most imaging windows unless through subcostal/sub-xiphoid windows. Because the 2-D array transducer can acquirevolumetric data in real-time (i.e., at 20 volumes/s), no cardiacgating is necessary for the structures except for those in veryrapid motion (such as the cardiac valves) (Deng 2002). If thetransducer has to be moved (say, tilted) to reveal structurespreviously in the shadows, tracking its spatial movement andgating the cardiac motion become essential for reliable volu-

metric assessment.

3-D fetal echocardiography ● J. DENG et al. 1101

vided, at a gross anatomical level, into recognisable andunrecognisable structures (i.e., whether the four cham-bers and great vessels could be identified). The entirety isclassified by whether or not most (over two thirds) oreven the whole heart could be visualised, by either se-quential cross-sectional reformatting or dynamic 3-Dsurface displaying, to appreciate the spatial relationshipsbetween the major structures. Thus, a data set withrecognisable major structures and appreciable spatial re-lationships was considered a useful one.

RESULTS

A total of 60 data sets were collected from the 15patients. It took from about 10 up to the allowed 16 minto locate optimal windows and acquire two to eight datasets in each patient.

Table 1 summarises the findings. Subcostal win-dows were obtained in 12 (20%) of the 60 data sets in 6(40%) patients. Subcostal windows were unavailablefrom the remaining 9 patients due to unfavourable fetalpositions within the allowed scanning period. Of the 12sets, 6 (50%) provided useful information with recogni-sable major structures and appreciable spatial relation-ships (Figs. 2 and 3, with corresponding dynamic 3-Dmovies on www.medphys.ucl.ac.uk/mgi/jdeng under Fe-tal Heart entry). The remaining 6 data sets failed toprovide useful information due to young gestational agein 3, long distance of the heart from the transducer in 2and fetal (skeletal) extremities shadowing the imagingwindow in 1.

Of 48 datasets from other imaging windows, only 9sets (19%) provided useful information. The low ratewas mainly due to acoustic shadowing of the thoracicskeleton obscuring many parts of the heart.

Fig. 2. Images of a 22-week fetal heart from a dynamic 3-D(4-D) data set from the right subcostal window. Two simulta-neously imaged planes in (a) and (b) are perpendicular to eachother along the dotted lines, but both parallel to sound direction(arrows). Due to better axial resolution, major structures, suchas the right and left atria (RA, LA), right and left ventricles(RV, LV), are clearly visualised. Due to poorer lateral resolu-tion, simultaneously imaged planes perpendicular to the sounddirection (not included here) show no identifiable structures. (c)A 3-D surface display of the chest transversely ‘cut’ open at thefour-chamber view level, showing recognisable major struc-tures in the upper half of the heart (dotted circle). The normalfine structures, such as the foramen valve (FV), interatrial andinterventricular septa (IAS, IVS), and the tricuspid and mitralvalves along with their corresponding valvular complex (TV�,MV�), appear to be much thicker than normal because ofinsufficient spatial resolution and because the structures areperpendicular to the sound direction. Therefore, both ventriclesare not properly visualised as shallow chambers (but a corre-sponding 4-D movie on our website gives a slightly moreaccurate impression). DAO � descending aorta.

1102 Ultrasound in Medicine and Biology Volume 28, Number 9, 2002

Anatomical and functional details were only occa-sionally visualised without certainty. The uncertain ana-tomical features included the number of valvular leaflets,the difference in the insertion to the septa between themitral and tricuspid valves, the morphological patterns oftrabeculations and papillary muscles and the great arte-rial branches. Therefore, none of the cardiac malforma-tions, despite being diagnosed on conventional cross-sectional scans, was reliably identified.

The uncertain functional features included the dif-ferentiation of atrioventricular contractions and thechamber volumetric changes. Therefore, although therhythm irregularity was noticed on dynamic cross-sec-tional reformatted views, the reformatted M-mode wave-forms were unable to depict the temporal relationshipsbetween the atrial and ventricular contractions for clas-sification of the types of the two arrhythmias. The over-thickened structures, as shown in Fig. 2c, also made itimpossible to quantify the chamber volumes reliably.

DISCUSSION

Using real-time 3-D data sets obtained in this study,we have demonstrated the feasibility of assessing themajor cardiovascular structures and their spatial relation-ships on dynamic 3-D surface displays, as well as onmultiplanar reformatted cross-sectional displays. In twoprevious studies, only reformatted cross-sectional dis-plays were used to compare the success rate of obtainingstandard cross-sectional views between real-time 3-Dand conventional cross-sectional approaches (Scharf etal. 2000; Sklansky et al. 1999). Such a comparison wasnot attempted in this study for two reasons. First, thequality of individual real-time 3-D images was not yetcomparable to that obtained by a state-of-the-art, real-time cross-sectional scanner in terms of morphologicaldetails needed for reliable clinical analysis. Second, our

b

a

c

Fig. 3. A 26-week fetal heart from a 4-D data set from thesubxiphoid window. (a) An imaging plane parallel to the sounddirection (arrow), clearly depicting the major structures. (b) A3-D surface display of the chest obliquely ‘cut’ open at thefive-chamber-view level, revealing major structures in the up-per half of the heart. (c) A 3-D negative surface display (i.e.,cardiovascular cavities shown as solid) with artificial coloursmarking major structures. Viewed right posteriorly, it vividlydiscloses the spatial relationships between the major structures.Due to insufficient image resolution, fine structures cannot berendered here, and areas in grey are noncardiac structures thatcould not be removed reliably even with manual segmentation.AAO � ascending aorta; AD � arterial duct; Arch � aorticarch; ICV � inferior caval vein; PV � pulmonary vein; RVi &RVo � right ventricular inlet & outlet; VD � venous duct.

Also see Fig. 2 for abbreviations.

3-D fetal echocardiography ● J. DENG et al. 1103

slice-reconstruction 3-D studies have shown that the 3-Dsurface display and reformatted cross-sectional displayare the most useful ways to present 3-D data sets (Denget al. 1996, 2000b, 2001b, 2002). We, therefore, onlyused real-time 3-D data to compare the quality of 3-Dsurface displays and reformatted cross-sectional viewsbetween different imaging windows and only at the grossanatomical level.

Pioneering real-time cross-sectional studies have es-tablished several useful cardiac sectional images as stan-dard views (Allan et al. 1980; DeVore et al. 1982; Langeet al. 1980). This was more or less an expedient measurethen to compensate for the lack of direct 3-D informationfrom cross-sectional scans and for ease of screening(such as the use of the four-chamber view) (Copel et al.1987; DeVore 1992; Fermont et al. 1986). Experiencedpathologists and echocardiologists have long been em-phasising the importance of obtaining sequential tomo-grams, whenever possible during a cross-sectional scan,to help form a mental 3-D picture. Fortunately, sequen-tial imaging using reformatted cross-sectional displayhas become available from both real-time and slice-reconstruction 3-D approaches, even after a patient scan.Therefore, our assessment of data set usefulness was notrestricted only to the availability of reformatted standardcross-sectional views. Instead, we used the sequentialcross-sectional reformatting and dynamic 3-D displayingto assess the success rate of obtaining the overall impres-sion of the continuity from and spatial relationship be-tween the major cardiovascular structures.

Compared with complicated motion-gated slice-re-construction 3-D approaches (Deng et al. 2000b, 2001b,2002; Kwon et al. 1996; Meyer-Wittkopf et al. 2000,2001; Nelson et al. 1996), the main advantages of thereal-time 3-D approach are the simplicity of dynamic3-D data acquisition with just a single transducer and theabsence of artificial distortion caused by inaccurate spa-tial and/or temporal registration. The 2-D array had atemporal resolution of 20 volumes/s in this study andwas cited as offering 40 volumes/s at a depth of 8 cm inanother study (Scharf et al. 2000). When an acousticwindow to image the entire fetal heart or a considerableportion of it is found, the transducer only has to be keptimmobile for one cardiac cycle (about 0.5 s) to acquireabout 10 to 20 volumes. Without taking into account thespatial resolution, these temporal resolutions would befairly adequate for general assessment of the cardiac wallmovement (at 20 volumes/s), and even the cardiac val-vular movement (at 40 volumes/s) (Deng 2002). Becauseelectrocardiographic timing is unavailable during a pre-natal volumetric scan, the minimal acquisition time perdata set needs at least two cardiac cycles. During thistime, an M-mode sampling line can be placed throughcyclically informative structures, such as cardiac valves,

producing M-mode waveforms. If the probe is not movedduring the two cycles, the M-mode can trace two iden-tical waveforms, allowing the length of one cardiac cycleto be identified. The data for one cardiac cycle can thenbe saved, minimising the use of computing resources,such as in postprocessing, and increasing the speed ofIntranet echocardiographic consultation (Michailidis etal. 2001).

The theoretical benefit of the ability of a 2-D array toform an imaging volume in an instant is offset by theprenatal reality that only restricted imaging windows areavailable to the heart, which is usually caged in thethoracic skeleton and not positioned in the proximity ofthe transducer. The study has shown that the shadowing-free subcostal windows were only obtainable in aboutone fifth of attempts and in 6 of the 15 patients (Table 1).It is, therefore, necessary to direct the 2-D array trans-ducer through the fetal narrow intercostal spaces whenusing other imaging windows to cover a sufficient car-diac volume. As soon as a probe (no matter whethercross-sectional or volumetric) is moved during a dy-namic 3-D scan, spatial and temporal tracking (subse-quently cardiac gating) become necessary (Deng et al.2001b). Nevertheless, the real-time 3-D and other rapidvolumetric imaging techniques, such as Philips andKretztechnik’s real-time 4-D, may potentially reduce thisnecessity. But this is when, and only when, a movingorgan is well exposed to be imaged volumetrically.

The main problem with current real-time 3-D sys-tems is the inadequate spatial resolution for detailedanatomic and functional assessment. This was confirmedby its inability to diagnose cardiac malformations andclassify arrhythmia types in this study. In fact, only halfof the data sets obtained from the ideal windows offereduseful information about cardiac structures and theirspatial relationships at the gross anatomical level. Thefraction was as low as one fifth for the data sets obtainedfrom the other windows. Nonetheless, these findingsshould not discourage further development of this orother 3-D techniques for dynamic 3-D fetal cardiac im-aging, considering that over 40 years have been spent indeveloping fetal echocardiography from a simple M-mode tracing (Deng 2001) to the current state-of-the-artreal-time cross-sectional imaging (Allan 2000; Chaoui2001). Also, improved techniques have become availablefor fabricating more sophisticated transducers and pro-gramming more intelligent graphics software (Light et al.1998; Ota et al. 2001; von Ramm 2000).

Imaging the entire fetal heart in 3-D dynamically forreliable clinical diagnosis and physiologic study requiressufficient spatial and temporal resolution, and sufficientimaging window and operating practicality. Because nei-ther the real-time volumetric method nor the motion-gated slice-reconstruction method is able to meet all

1104 Ultrasound in Medicine and Biology Volume 28, Number 9, 2002

these requirements alone, a combined development ofthe two or more approaches into an integrated systemmay be a solution. In the future, it is most likely that thiswill be a real-time 3-D system with high spatial andtemporal resolution for convenient acquisition of dy-namic volumetric data, and also with spatial and tempo-ral tracking for free-hand transducer movement when-ever needed (e.g., to avoid acoustic shadowing) or fordealing with complex movements of body tissues, imag-ing probe and operating environment.

Acknowledgments—The authors are grateful to Dr. R. Richards of UCLMedical Physics for 3-D graphics computing, to Prof. J. Deanfield andthe Cardiac Unit staff, GOS Hospital for clinical arrangements, and toDr. L. B. Pauliks for assistance in data conversion. J. Deng and D.McDonald are supported by a WellBeing grant (Ref: 239) and anEPSRC/MRC IRC grant (GR/N14248/01), respectively. R. H. Ander-son is supported by the British Heart Foundation, together with theJoseph Levy Foundation. The volumetric scanner was supplied by agrant from the Deutsche Forschungsgemeinschaft.

REFERENCES

Allan L. Antenatal diagnosis of heart disease. Heart 2000;83:367.Allan LD, Tynan MJ, Campbell S, Wilkinson JL, Anderson RH.

Echocardiographic and anatomical correlates in the fetus. BrHeart J 1980;44:444–451.

Chang FM, Hsu KF, Ko HC, et al. Fetal heart volume assessment bythree-dimensional ultrasound. Ultrasound Obstet Gynecol 1997;9:42–48.

Chaoui R. Fetal echocardiography: State of the art of the state of theheart. Ultrasound Obstet Gynecol 2001;17:277–284.

Chaoui R, Kalache KD, Hartung J. Application of three-dimensionalpower Doppler ultrasound in prenatal diagnosis. Ultrasound ObstetGynecol 2001;17:22–29.

Copel JA, Pilu G, Green J, Hobbins JC, Kleinman CS. Fetal echocar-diographic screening for congenital heart disease: The importanceof the four-chamber view. Am J Obstet Gynecol 1987;157:648–655.

Deng J. Start of the art of the heart. Ultrasound Obstet Gynecol2001;18:405–406.

Deng J. Four-dimensional fetal echocardiography. In: Cosgrove D,Humphries K, ed.s. 3D ultrasound imaging: Principles and clinicalapplications. London: J Wiley & Sons, 2002(in press).

Deng J, Birkett AG, Kalache KD, et al. Conversion of umbilical arterialDoppler waveforms to cardiac cycle triggering signals: A prepara-tory study for online motion-gated three-dimensional fetal echocar-diography. Ultrasound Med Biol 2001a;27:51–59.

Deng J, Gardener JE, Rodeck CH, Lees WR. Fetal echocardiography inthree and four dimensions. Ultrasound Med Biol 1996;22:979–986.

Deng J, Newton NM, Hall-Craggs MA, et al. Novel technique forthree-dimensional visualisation and quantification of deformable,moving soft-tissue body parts. Lancet 2000a;356:127–131.

Deng J, Ruff CF, Linney AD, Lees WR, Hanson MA, Rodeck CH.Simultaneous use of two ultrasound scanners for motion-gatedthree-dimensional fetal echocardiography. Ultrasound Med Biol2000b;26:1021–1032.

Deng J, Yates R, Birkett AG, et al. Online motion-gated dynamicthree-dimensional echocardiography in the fetus-preliminary re-sults. Ultrasound Med Biol 2001b;27:43–50.

Deng J, Yates R, Sullivan ID, et al. Dynamic three-dimensional colourDoppler ultrasound of human fetal intracardiac flow. UltrasoundObstet Gynecol 2002;20:131–136.

DeVore GR. The aortic and pulmonary outflow tract screening exam-ination in the human fetus. J Ultrasound Med 1992;11:345–348.

DeVore GR, Donnerstein RL, Kleinman CS, Platt LD, Hobbins JC.Fetal echocardiography. I. Normal anatomy as determined by real-time directed M-mode ultrasound. Am J Obstet Gynecol 1982;144:249–260.

Fermont L, de Geeter B, Aubry MC, et al. A close collaborationbetween obstetricians and pediatric cardiologists allows antenataldetection of severe cardiac malformations by 2D echocardiography.In: Doyle EF, Engle ME, Gersony WM, et al., eds. Pediatriccardiology1986; Proceedings of the Second World Congress 1986.New York: Springer-Verlag, 1986:34–37.

Guerra FA, Isla AI, Aguilar RC, Fritz EG. Use of free-hand three-dimensional ultrasound software in the study of the fetal heart.Ultrasound Obstet Gynecol 2000;16:329–334.

Kwon J, Shaffer E, Shandas R, Knudson O, DeGroff C, Valdes-Cruz L.Acquisition of three-dimensional fetal echocardiograms using ex-ternal trigger source. Am J Soc Echocardiogr 1996;9:389.

Lange LW, Sahn DJ, Allen HD, et al. Qualitative real-time cross-sectional echocardiographic imaging of the human fetus during thesecond half of pregnancy. Circulation 1980;62:799–806.

Levental M, Pretorius DH, Sklansky MS, et al. Three-dimensionalultrasonography of normal fetal heart: Comparison with two-di-mensional imaging. J Ultrasound Med 1998;17:341–348.

Light ED, Davidsen RE, Fiering JO, Hruschka TA, Smith SW. Progressin two-dimensional arrays for real-time volumetric imaging. Ultra-son Imaging 1998;20:1–15.

Meyer-Wittkopf M, Cooper S, Vaughan J, Sholler G. Three-dimen-sional (3D) echocardiographic analysis of congenital heart diseasein the fetus: Comparison with cross-sectional (2D) fetal echocar-diography. Ultrasound Obstet Gynecol 2001;17:485–492.

Meyer-Wittkopf M, Rappe N, Sierra F, Barth H, Schmidt S. Three-dimensional (3-D) ultrasonography for obtaining the four and five-chamber view: Comparison with cross-sectional (2-D) fetal sono-graphic screening. Ultrasound Obstet Gynecol 2000;15:397–402.

Michailidis GD, Simpson JM, Karidas C, Economides DL. Detailedthree-dimensional fetal echocardiography facilitated by an internetlink. Ultrasound Obstet Gynecol 2001;18:325–328.

Nelson TR, Pretorius DH, Sklansky M, Hagen-Ansert S. Three-dimen-sional echocardiographic evaluation of fetal heart anatomy andfunction: Acquisition, analysis, and display. J Ultrasound Med1996;15:1–9.

Ota T, Kisslo J, von Ramm OT, Yoshikawa J. Real-time, volumetricechocardiography: Usefulness of volumetric scanning for the as-sessment of cardiac volume and function. J Cardiol 2001;37(Suppl.1):93–101.

Scharf A, Geka F, Steinborn A, et al. 3D real-time imaging of the fetalheart. Fetal Diagn Ther 2000;15:267–274.

Sklansky MS, Nelson TR, Pretorius DH. Three-dimensional fetal echo-cardiography: Gated versus nongated techniques. J Ultrasound Med1998;17:451–457.

Sklansky MS, Nelson T, Strachan M, Pretorius D. Real-time three-dimensional fetal echocardiography: Initial feasibility study. J Ul-trasound Med 1999;18:745–752.

Vogel M, Ho SY, Lincoln C, Anderson RH. Transthoracic three-dimensional echocardiography for the assessment of straddlingtricuspid or mitral valves. Cardiol Young 2000;10:603–609.

von Ramm OT. 2-D arrays. Ultrasound Med Biol 2000;26(Suppl.1):S10–S12.

Wang XF, Li ZA, Cheng TO, et al. Four-dimensional echocardiogra-phy: Methods and clinical application. Am Heart J 1996;132:672–684.

Zosmer N, Jurkovic D, Jauniaux E, et al. Selection and identification ofstandard cardiac views from three-dimensional volume scans of thefetal thorax. J Ultrasound Med 1996;15:25–32.

3-D fetal echocardiography ● J. DENG et al. 1105