Four-Dimensional Ultrasonography of the Fetal Heart using a Novel Tomographic Ultrasound Imaging Display

Abstract

Objective

The objective of this study was to investigate the feasibility of examining the fetal heart with Tomographic Ultrasound Imaging (TUI) using four-dimensional (4D) volume datasets acquired with spatiotemporal image correlation (STIC).

Material and Methods

One hundred and ninety-five fetuses underwent 4D ultrasonography (US) of the fetal heart with STIC. Volume datasets were acquired with B-mode (n=195) and color Doppler imaging (CDI) (n=168), and were reviewed offline using TUI, a new display modality that automatically slices 3D/4D volume datasets, providing simultaneous visualization of up to eight parallel planes in a single screen. Visualization rates for standard transverse planes used to examine the fetal heart were calculated and compared for volumes acquired with B-mode or CDI. Diagnoses by TUI were compared to postnatal diagnoses.

Results

1) The four- and five-chamber and the three-vessel and trachea views were visualized in 97.4% (190/195), 88.2% (172/195), and 79.5% (142/195), respectively, of the volume datasets acquired with B-mode; 2) these views were visualized in 98.2% (165/168), 97.0% (163/168), and 83.6% (145/168), respectively, of the volume datasets acquired with CDI; 3) CDI contributed additional diagnostic information to 12.5% (21/168), 14.2% (24/168) and 10.1% (17/168) of the four- and five-chamber and the three-vessel and trachea views; 4) cardiac anomalies other than isolated ventricular septal defects were identified by TUI in 16 of 195 fetuses (8.2%) and, among these, CDI provided additional diagnostic information in 5 (31.3%); 5) the sensitivity, specificity, positive- and negative-predictive values of TUI to diagnose congenital heart disease in cases where both B-mode and CDI volume datasets were acquired prenatally were 92.9%, 98.8%, 92.9% and 98.8%, respectively.

Conclusion

Standard transverse planes commonly used to examine the fetal heart can be automatically displayed with TUI in the majority of fetuses undergoing 4D US with STIC. Due to the retrospective nature of this study, the results should be interpreted with caution and independently confirmed before this methodology is introduced into clinical practice.

Introduction

Severe congenital heart disease is present in approximately 6 of every 1,000 live births and is the leading cause of death among infants with congenital anomalies.[38, 39] When trivial defects with minor or no clinical significance such as small muscular ventricular septal defects (VSDs) present at birth are included in the analysis, the incidence may be as high as 75/1,000 live births.[38, 57, 59] Prenatal diagnosis of congenital heart disease [e.g. transposition of the great arteries, hypoplastic left heart syndrome, and coarctation of the aorta] is associated with a decrease in neonatal morbidity and mortality.[9, 26, 47, 70] However, prenatal diagnosis of congenital heart disease by two-dimensional (2D) ultrasound relies heavily on operator skills, and the detection rates in population based studies range from 6 to 35%.[11, 15, 25, 27, 34, 40, 42, 58, 66, 67, 69] Since the dependency on operator skills is considered by many as the limiting factor in improving the detection rates for congenital heart disease,[3, 68] several investigators have explored the use of three-dimensional (3D) and four-dimensional (4D) ultrasound for the examination of the fetal heart.[1, 2, 4, 5, 7, 8, 10, 12, 13, 16–24, 28–30, 32, 33, 35–37, 41, 45, 46, 48–56, 60–65, 71, 72, 74, 76]

In 2002, Yagel et al.[73] proposed a method to streamline the examination of the fetal heart, based on the evaluation of five transverse planes: 1) the transverse view of the upper abdomen; 2) the four-chamber view; 3) the five-chamber view; 4) the three-vessel view; and 5) the three-vessel view and trachea view (Figure 1).[73] This method has the potential to simplify the examination of the fetal heart by minimizing the need to obtain several short and long axis views of the heart. These views may be difficult to obtain due to an unfavorable fetal position or lack of operator experience.[73] Subsequently, Chaoui and McEwing[14] proposed that the evaluation of only three planes of section with color Doppler (the four- and five-chamber and three-vessel views) would be sufficient to identify most cases of congenital heart disease.

Figure 1.

Tomographic Ultrasound Imaging of a normal fetal heart in systole (A) and diastole (B). The overview image on the left upper panel of each figure shows the orthogonal sagittal plane to the sections that are being displayed. Each line represents a slice. The center slice is marked with an asterisk (*) and each subsequent plane to the right or left is marked with numbers ranging from −4 to +4. The plane marked by the dotted line is not displayed. In this volume dataset, the five transverse planes of section proposed by Yagel et al. (Ultrasound Obstet Gynecol 2001;17:367–369) for the examination of the fetal heart are visualized. Please note that the five chamber view was better visualized during systole. Legends: PA: pulmonary artery; Ao: aorta; SVC: superior vena cava; LPA: left pulmonary artery; RV: right ventricle; LV: left ventricle; RA: right atrium; LA: left atrium; FO: foramen ovale; IVC: inferior vena cava; IVS: interventricular septum; 4CH: four-chamber; 5CH: five-chamber.

Spatiotemporal image correlation (STIC) is a commercially available technology for 4D ultrasonographic examination of the fetal heart. Volume datasets are acquired over a period of 7.5 to 15 seconds and can be analyzed either in the presence of the patient or offline in a computer workstation. Several techniques have been proposed to evaluate the fetal heart by 4D ultrasound with STIC, including multiplanar display to visualize the planes of section required to conduct a basic and extended examination of the fetal heart, as well as rendering techniques to display 4D images of specific cardiac structures such as the atrioventricular valves, outflow tracts, aortic arch, ductal arch, and venous return to the heart.[8, 13, 21–24, 28–30, 32, 33, 71, 72, 74]

Recently, ultrasound manufacturers have developed software that automatically slices volume datasets acquired by 3D and 4D ultrasonography (Multislice View^™, Accuvix, Medison, Seoul, Korea; Tomographic Ultrasound Imaging, Voluson 730, General Electric Medical Systems, Kretztechnik, Zipf, Austria). Such software produces a series of tomographic images, akin to display methods used by diagnostic imaging professionals to review computerized tomography and magnetic resonance imaging studies. This novel display modality has been recently described for prenatal visualization of anatomic fetal structures and to diagnose congenital anomalies.[44] In the current study, we investigate the feasibility of examining the fetal heart with Tomographic Ultrasound Imaging (TUI) in 4D volume datasets acquired with STIC. Specifically, we sought to determine whether the five planes of section proposed by Yagel et al.[73] and the three planes of section proposed by Chaoui and McEwing[14] could be automatically obtained using this new modality.

Material and Methods

Four-dimensional volume datasets of the fetal heart were acquired with transverse sweeps through the fetal chest in 195 patients examined at our ultrasound unit between December 1^st 2003 and December 31^st 2004. Examinations were performed with STIC (Voluson 730 Expert, release BTO4, General Electric Medical Systems, Kretztechnik, Zipf, Austria) using hybrid mechanical and curved array transducers (RAB 4-8P, RAB 4-8L, RAB 2-5P, RAB 2-5L). Patients were examined between 14 and 41 weeks of gestation (median 24 1/7 weeks; interquartile range: 15 3/7 to 32 6/7 weeks). All patients were enrolled in research protocols approved by the Institutional Review Board of the National Institute of Child Health and Human Development (NICHD/NIH/DHHS) and by the Human Investigation Committee of Wayne State University (Detroit, Michigan); all signed a written informed consent prior to participation in the study.

After removal of patient identifiers, examinations were retrospectively reviewed offline on a personal computer using the 4D View software version 5.0 (4D VIEW 5.0, General Electric Medical Systems, Kretztechnik, Zipf, Austria). Even though some of the patients were examined more than once during pregnancy and multiple volume datasets were acquired in the course of each scanning session, a maximum of two volume datasets per patient (one acquired with B-mode imaging and one with color Doppler) were included in the study. Preference was given to volume datasets acquired between 16 and 24 weeks of gestation, when available, and the volume dataset considered by the investigator to be of highest quality was selected. Volume datasets with the following characteristics were considered to be of high quality: 1) the acquisition sweep included planes of section from the upper mediastinum and through the upper abdomen; 2) the fetal spine was positioned between 9 and 3 o’clock, minimizing the possibility of shadowing from ribs or spine; and 3) minimum or no motion artifact was observed on the sagittal orthogonal plane to the original plane of acquisition (panel B). Regardless of the perceived quality, at least one volume dataset per patient was included in the study.

Automatic slicing of volume datasets with TUI

Cross-sectional planes to the original plane of acquisition were automatically displayed with TUI, a novel display modality available in version 5.0 of the 4D View Software (General Electric Medical Systems, Kretztechnik, Zipf, Austria) (Figures 1 and 2, video clips 1 and 2). This display modality allows examiners to automatically slice a volume dataset and simultaneously visualize up to eight parallel planes of section on the same screen. Importantly, motion information in volume datasets acquired using 4D ultrasonography techniques is not lost and, therefore, multiple sections of a beating heart can be analyzed at the same time. An “overview image” is shown on the upper left corner. This view depicts a plane orthogonal to the slices, and parallel lines demarcate the position of the slices within the volume dataset (Figure 1). The user can adjust the number and position of the slices with specific software controls. Hue, brightness and contrast controls can also be adjusted to optimize image quality.

Figure 2.

Tomographic Ultrasound Imaging (TUI) of a normal fetal heart in systole (A) and diastole (B). The volume datasets were acquired using B-mode and color Doppler imaging. The overview image on the left upper panel of each figure shows the orthogonal sagittal plane to the sections that are being displayed. Each line represents a slice. The center slice is marked with an asterisk (*) and each subsequent plane to the right or left are marked with numbers ranging from −4 to +4. The plane marked by the dotted line is not displayed. In this volume dataset, the three planes of section proposed by Chaoui et al. (Ultrasound Obstet Gynecol 2003;21:81–93) for the examination of the fetal heart are visualized.. Legends: PA: pulmonary artery; Ao: aorta; RV: right ventricle; LV: left ventricle; IVS: interventricular septum; 4CH: four-chamber; 5CH: five-chamber.

Analysis of volume datasets displayed with TUI

All volume datasets were analyzed by a single investigator (LG). Those acquired with B-mode imaging were analyzed first, and visualization rates for cardiac structures in each of the five planes of section proposed by Yagel et al.[73] were determined as follows: 1) abnormal; 2) normal; 3) present in the volume dataset but inadequate for diagnosis; and 4) not present in the volume dataset. To be considered adequate for diagnosis, the following structures and relationships should have been visualized in each scanning plane:

1. Transverse view of the upper abdomen (Figure 1A): 1) stomach on the left side of the abdomen, 2) transverse section of the descending aorta in front and to the left of the spine, 3) inferior vena cava located on the right side of the spine.

2. Four-chamber view: evaluated according to the criteria proposed by the American Institute of Ultrasound in Medicine for the performance of the basic fetal cardiac screening examination (Table 1).[43]

3. Five-chamber view: 1) aortic root visualized and connected to the left ventricle; 2) the anterior wall of the aorta should be in continuity with the ventricular septum.[43]

4. Pulmonary artery view: pulmonary artery visualized leaving the right ventricle and at least one of the branches bifurcating at its distal end.[43, 73, 75]

5. Three-vessel and trachea view: 1) main pulmonary trunk in direct communication with the ductus arteriosus; 2) transverse section of the aortic arch between the main pulmonary trunk and superior vena cava; 3) cross section of the superior vena cava visualized on the right side of the chest; and 4) trachea visualized posterior to the superior vena cava.[73]

Table 1.

American Institute of Ultrasound in Medicine guidelines for the examination of the four-chamber view.

General	Four cardiac chambers are present
	Majority of heart located in left chest
	Heart occupies about one-third of the thoracic area
	Normal cardiac situs, axis, and position
	Pericardial effusion not seen
Atria	Atria appear approximately equal in size
	Foramen ovale flap lies in the left atrium
	Lower rim of atrial septum (septum primum) is present
Ventricles	Ventricles approximately equal in size
	Ventricular septum appears intact from apex to crux
AV Valves	Both atrioventricular valves open and move freely
	Tricuspid valve septal insertion more apical than mitral valve

Reprinted with permission from the American Institute of Ultrasound in Medicine (AIUM) (Lee, JUM, 1998)

Following the evaluation of B-mode imaging volume datasets, visualization rates for volume datasets acquired with color Doppler were determined according to the criteria proposed by Chaoui and McEwing:[14]

1. Four-chamber view: 1) diastolic perfusion across the AV valves documented with color Doppler, with no evidence of valve regurgitation; 2) no color flow observed crossing the ventricular septum.

2. Five-chamber view: 1) aortic root emerging from the left ventricle; 2) interventricular septum in continuity with the anterior wall of the ascending aorta. Abnormalities in this view included: a) turbulent flow across the aortic valve; b) shunting through a perimembranous VSD with a normal aortic connection; c) visualization of the origin of the pulmonary trunk emerging from the left ventricle identified by its bifurcation into the main pulmonary arteries; and d) overriding of the aorta over both ventricles, connected by a VSD.

3. Three-vessel view: 1) aorta and pulmonary trunk converging toward the left thorax with the trachea to their right; 2) pulmonary trunk with a slightly greater caliber than the aorta (ratio 1.2:1); 3) straight course of the vessels; and 4) antegrade flow through both great vessels throughout the cardiac cycle.

Volume datasets acquired with B-mode imaging and color Doppler were compared to determine if color Doppler provided more information than that obtained by examination with B-mode imaging alone. Visualization rates for the four- and five-chamber and three-vessel views were compared using the McNemar’s test. Agreement between diagnoses by the two-modalities was tested using the Kappa test.

Following determination of the visualization rates for each view, the diagnoses established by TUI were compared to postnatal diagnoses by neonatal imaging modalities, cardiac surgery or autopsy. Sensitivity, specificity, as well as positive- and negative-predictive values for the diagnosis of congenital heart disease were calculated first for volume datasets acquired with B-mode imaging and then for those acquired with the addition of color Doppler. Cases of isolated VSDs (either suspected prenatally or detected in the neonatal period) were excluded from this analysis because neonatal echocardiograms are not performed in every neonate in our institution in the absence of symptoms.

Statistical analysis was performed with SPSS 12.0 for Windows (SPSS, Chicago, IL).

Results

STIC volume datasets of the fetal heart were acquired with B-mode or color Doppler imaging in 195 and 168 cases, respectively.

Table II shows visualization rates for anatomical structures expected to be present in the five transverse scanning planes proposed by Yagel et al.,[73] after automatic slicing with TUI (Figure 1, video clip 1). The four- and five-chamber and three-vessel and trachea views were visualized in 97.4% (190/195), 88.2% (172/195), and 79.5% (142/195) of the cases, respectively. A three-vessel view showing bifurcation of the pulmonary artery was visualized in only 51.8% (101/195) of the cases. Transverse views of the upper fetal abdomen were present in 64.1% of the volume datasets (125/195).

Table 2.

Visualization rates for fetal heart structures displayed with TUI. All volume datasets acquired using STIC with B-mode imaging (n=195)

		Normal		Abnormal		Not diagnostic		Not present in the volume dataset
Cardiac view	Structures	n	%	n	%	n	%	n	%
Upper abdomen		123	63.1	2	1.0	70	35.9
	Stomach on the left side	129	66.2	2	1.0	-	-	64	32.8
	Aorta on the left side of spine	159	81.5	1	0.5	1	0.5	34	17.4
	IVC on the right side of spine	152	77.9	2	1.0	1	0.5	40	20.5
4CH view		166	85.1	24	12.3	5	2.6	-	-
	Four cardiac chambers present	188	96.4	3	1.5	4	2.1	3	1.5
	Majority of heart located in left chest	185	94.9	2	1.0	-	-	8	4.1
	Heart occupies about one-third of the thoracic area	185	94.9	-	-	-	-	10	5.1
	Normal cardiac situs, axis, and position	183	93.8	4	2.1	-	-	8	4.1
	Pericardial effusion not seen	192	98.5	1	0.5	2	1.0	-	-
	Atria appear approximately equal in size	183	93.8	4	2.1	8	4.1	-	-
	Foramen ovale flap lies in the left atrium	179	91.8	4	2.1	12	6.2	-	-
	Lower rim of atrial septum (septum primum) is present	183	93.8	4	2.1	8	4.1	-	-
	Ventricles approximately equal in size	183	93.8	9	4.6	3	1.5	-	-
	Ventricular septum appears intact from apex to crux	157	80.5	30	15.4	8	4.1	-	-
	Both atrioventricular valves open and move freely	185	94.9	4	2.1	6	3.1	-	-
	Tricuspid valve septal insertion more apical than mitral valve	177	90.8	6	3.1	4	2.1	-	-
5CH view		154	79.0	18	9.2	24	12.3
	Aortic root emerging from the left ventricle	161	82.6	13	6.7	21	10.8	-	-
	IVS in continuity with the anterior wall of the ascending aorta	153	78.5	18	9.2	24	12.3	-	-
PA view		93	47.7	8	4.1	2	1.0	92	47.2
	PA visualized leaving the right ventricle	94	48.2	8	4.1	2	1.0	91	46.7
	At least one of the branches of PA bifurcating at its distal end	93	47.7	7	3.6	2	1.0	93	47.7
3VTV		142	72.8	13	6.7	2	1.0	38	19.5
	PA communicates directly with ductus arteriosus	145	74.4	8	4.1	3	1.5	39	20
	Transverse aortic arch between PA and SVC	154	79.0	8	4.1	4	2.1	29	14.9
	SVC visualized on the right side of the chest	161	82.6	3	1.5	2	1	29	14.9
	Trachea visualized posterior to SVC	146	74.9	1	0.5	6	3.1	42	21.5

Legends: 4CH: four-chamber; 5CH: five-chamber; 3VTV: three-vessel and trachea view; IVS: interventricular septum; SVC: superior vena cava; PA: pulmonary artery; STIC: spatiotemporal image correlation; IVC: inferior vena cava; TUI: Tomographic Ultrasound Imaging.

In volume datasets acquired with color Doppler (Figure 2, ), the four- and five-chamber and three-vessel views were visualized in 98.2% (165/168), 97.0% (163/168), and 83.6% (145/168) of the cases, respectively (Table III). Table IV compares the visualization rates for the four-chamber, five-chamber and three-vessel views between volume datasets acquired with B-mode or color Doppler imaging. The additional contribution of color Doppler imaging was statistically significant for the five-chamber view only (McNemar test = 13.5, p = 0.004).

Table 3.

Visualization rates for fetal heart structures displayed with TUI. All volume datasets were acquired with STIC and color Doppler imaging (n=168)

		Normal		Abnormal		Not diagnostic		Not present in the volume dataset
Cardiac view	Structures	n	%	n	%	n	%	n	%
4CH view		132	66.7	33	16.9	3	1.5
	Diastolic perfusion across the atrioventricular valves documented with color Doppler, with no evidence of valve regurgitation	152	77.9	14	7.2	2	1.0	-	-
	No color flow observed crossing the ventricular septum.	134	68.7	31	15.9	3	1.5	-	-
5CH view		150	76.9	13	6.7	5	2.6
	Aortic root emerging from the left ventricle and the interventricular septum in continuity with the anterior wall of the ascending aorta	150	76.9	13	6.7	5	2.6	-	-
3VV		134	68.7	11	5.6	23	11.8
	Aorta and pulmonary trunk converging toward the left thorax with the trachea to the right	138	70.8	9	4.6	-	-	21	10.8
	Pulmonary trunk with a slightly greater caliber than the aorta (ratio 1.2:1)	137	70.3	9	4.6	-	-	22	11.3
	Straight course of the vessels	139	71.3	9	4.6	-	-	20	10.3
	Antegrade flow through both great vessels throughout the cardiac cycle	137	70.3	9	4.6	-	-	22	11.3

Legends: 4CH: four-chamber; 5CH: five-chamber; 3VV: three-vessel view; TUI: Tomographic Ultrasound Imaging; STIC: spatiotemporal image correlation.

Table 4.

Comparison of diagnostic information in TUI volume datasets acquired with B-mode imaging only or B-mode plus color Doppler imaging (n=168)

Plane of section	Perfect agreement between B-mode and color Doppler imaging in classifying views as normal, abnormal, or not diagnostic	Color Doppler modified the initial diagnostic impression by B-mode imaging only	No contribution of color Doppler to B-mode imaging	Kappa	P value by McNemar test
Four-chamber view	85.7% (145/168)	12.5% (21/168)*	1.8% (3/168)	0.537	0.135
Five-chamber view	83.9% (141/168)	14.2% (24/168)†	1.8% (3/168)	0.431	0.004
Three-vessel view	75% (126/168)	10.1% (17/168)‡	14.8% (25/168)	0.370	0.172

^*In 3 cases, non-diagnostic four-chamber views were adequately visualized with color Doppler; 4 false-positive and 14 false-negative diagnoses by B-mode imaging were corrected by color Doppler.

^†In 17 cases, a five-chamber view considered to be of no diagnostic value was adequately imaged by color Doppler; 6 false-positive and 1 false-negative diagnoses by B-mode imaging were corrected by color Doppler.

^‡In 13 cases, a three-vessel view considered to be of no diagnostic value was adequately imaged by color Doppler; 1 false-positive and 3 false-negative diagnoses by B-mode imaging were corrected by color Doppler.

TUI: Tomographic Ultrasound Imaging.

Isolated VSDs were suspected in 19 cases with B-mode imaging and 23 cases when the examination was performed with color Doppler. When isolated VSDs were excluded from the analysis, congenital heart anomalies were identified by TUI in 16 of the 195 fetuses (8.2%) (Table V). Figures 3, 4, 5 and 6 (video clips 3, 4, 5 and 6) illustrate abnormal findings in the various planes of sections displayed by TUI in cases of coarctation of the aorta, hypoplastic left heart, pulmonary atresia and transposition of the great arteries. Color Doppler provided additional diagnostic information in 5 of the 16 cases (31.3%). In case 5, retrograde perfusion of the pulmonary artery helped to correctly diagnose pulmonary atresia in a case of hypoplastic right ventricle (Figure 5). In case 6, a VSD was demonstrated in association with coarctation of the aorta. In case 8, disproportion in size between the right and left ventricles observed by B-mode imaging was correctly diagnosed as coarctation of the aorta (Figure 3). In case 13, color Doppler allowed the visualization of an interrupted IVC with azygous continuation in a case where volume datasets acquired by B-mode imaging were considered non-diagnostic. In case 16, what was initially suspected as transposition of the great arteries was diagnosed as tetralogy of Fallot.

Table 5.

Comparison of B-mode and color Doppler imaging in cardiac anomalies suspected by TUI, excluding isolated VSDs (n=16^*)

Case #	GA	B-Mode Findings	View	B-Mode Diagnosis	Color Doppler Findings	View	Color Doppler Diagnosis
		Abnormal axis	4CH
		VSD	4CH		VSD	4CH
1	23w5d	Overriding Ao	5CH	Tetralogy of Fallot	Overriding Ao	5CH	Tetralogy of Fallot
		Narrow PA	3VTV		Color flow only in transverse aortic arch	3VV
		Abnormal axis	4CH
		ASD	4CH		ASD	4CH
2	29w4d	VSD	4CH	DORV, TGA, ASD, VSD	VSD	4CH	DORV, TGA, ASD, VSD
		Hypoplastic RV	4CH		Hypoplastic RV	4CH
		Ao leaves RV transposed	5CH		Ao leaves RV transposed	5CH
		VSD	4CH
3	24w1d	Overriding Ao	5CH	Tetralogy of Fallot	poor imaging, not diagnostic		–
		PA not visualized	3VTV
4	25w2d	Transverse sections of Ao and azygous vein side by side	4CH	Interrupted IVC with azygous continuation	Transverse sections of aorta and azygous vein side by side	4CH	Interrupted IVC with azygous continuation
5	32w1d	Hypoplastic RV	4CH	Hypoplastic RV	Hypoplastic RV	4CH	Hypoplastic RV, pulmonary atresia
					PA not visualized	3VV
		RV/LV disproportion	4CH		RV/LV disproportion	4CH
6	25w3d	Narrow transverse aortic arch	3VTV	Coarctation of the Aorta	Narrow transverse aortic arch	3VTV	Coarctation of the Ao, VSD
					VSD	4CH
		Dextrocardia	Abdom
		Hypoplastic LV	4CH		Hypoplastic LV	4CH
7	19w3D	VSD	4CH	HLHS, DORV, TGA			HLHS, DORV, TGA
		DORV	3VTV		DORV	3VV
		TGA	5CH		TGA	5CH
		RV/LV disproportion	4CH		RV/LV disproportion	4CH
8	36w4d			RV/LV disproportion	Narrow transverse aortic arch	3VV	Coarctation of the Ao
					Reverse flow transverse aortic arch	3VV
		ASD	4CH		ASD	4CH
9	16w6d	VSD	4CH	AV canal	VSD	4CH	AV canal
		Single AV valve	4CH		Single AV valve	4CH
11	31w3d			HLHS	Hypoplastic LV	4CH	HLHS
		Hypoplastic LV	4CH		Tricuspid regurgitation	4CH
		Ao not visualized leaving LV	5CH		Ao not visualized leaving LV	5CH
		Ao not visualized in 3VTV	3VTV		Ao not visualized in 3VTV	3VTV
		VSD	4CH		VSD	4CH
12	22w6d	Overriding Ao	5CH	Tetralogy of Fallot	Overriding Ao	5CH	Tetralogy of Fallot
		PA not visualized	3VTV		PA smaller than Ao	3VV
		Abnormal axis	4CH
13	20w3d	Ao leaving RV	3VTV	TGA	Ao leaving RV	3VTV	TGA
		PA leaving LV	5CH		PA leaving LV	5CH
14	22w3d	Poor imaging, not diagnostic		—	Transverse sections of Ao and azygous vein side by side	4CH	Interrupted IVC with azygous continuation
		Hypoplastic RV	4CH		Hypoplastic RV	4CH
15	29w4d			Pulmonary atresia	Tricuspid regurgitation	4CH	Pulmonary atresia
		Absent PA	3VTV		Retrograde perfusion of PA	3VV
		Ao leaving RV	3VTV		Ao leaving RV	3VTV
16	28w0d	PA leaving LV	5CH	TGA	PA leaving LV	5CH	TGA
		VSD	4CH		VSD	4CH
17	19w3d	Parallel arteries leaving LV and RV	5CH	TGA	Overriding Ao	5CH	Tetralogy of Fallot
					PA not visualized	3VTV

Legend: DORV: double outlet right ventricle; TGA: transposition of the great arteries; ASD: atrial septal defect; VSD: ventricular septal defect; IVC: inferior vena cava; RV: right ventricle; LV: left ventricle; HLHS: hypoplastic left heart syndrome; AV: atrioventricular; PA: pulmonary artery; Ao: aorta; 3VTV: three-vessel and treachea view; 3VV: three-vessel view; 4CH: four-chamber; 5CH: five-chamber; TUI: Tomographic Ultrasound Imaging.

^*Case 10 (displayed in Table 6) was a rhabdomioma that was not visualized by TUI, either with B-mode or color Doppler imaging.

Figure 3.

Tomographic Ultrasound Imaging of a fetus with coarctation of the aorta. A) Volume dataset acquired with B-mode: (1) the three-vessel and trachea view shows a narrow transverse section of the aortic arch; (2) the four-chamber view shows disproportion between the right and left ventricles. B) Volume dataset acquired with color Doppler: (1) the three vessel view confirms the narrow transverse aortic arch and shows aliasing; (2) disproportion between the right and left ventricles is confirmed in the 4-chamber view. Legends: PA = pulmonary artery; Ao = aorta; RV = right ventricle; LV = left ventricle; 3V = three vessel; 3VT = three-vessel and trachea; 4CH = four-chamber; 5CH = five-chamber; Trv = transverse.

Figure 4.

Tomographic Ultrasound Imaging of a fetus with hypoplastic left heart syndrome, double outlet right ventricle and transposition of the great arteries. A) Volume dataset acquired with B-mode: (1) the three-vessel and trachea view shows the aorta leaving the right ventricle; (2) the five-chamber view shows the pulmonary artery leaving the right ventricle as well; (3) the four-chamber view shows the right ventricle and atrium only – the left ventricle is not visualized; 4) the transverse view of the fetal abdomen shows that the stomach is located on the right. B) Bolume dataset acquired with color Doppler: (1) the three-vessel view confirms that the aorta leaves the right ventricle; (2) the five-chamber view shows the pulmonary artery leaving the right ventricle as well. Legends: PA = pulmonary artery; Ao = aorta; RV = right ventricle; LV = left ventricle; 3V = three vessel; 3VT = three-vessel and trachea; 4CH = four-chamber; 5CH = five-chamber; Trv = transverse.

Figure 5.

Tomographic Ultrasound Imaging in a fetus with pulmonary atresia. A) Systole. (1) In the three-vessel view, retrograde perfusion (in red) of a narrow pulmonary artery is observed; (2) severe tricuspid regurgitation is observed in the four-chamber view. B) Diastole: (1) the 5-chamber view shows the aorta connecting normally to the left ventricle; (2) the four-chamber view shows normal ventricular filling only on the left ventricle, with tricuspid regurgitation still observed in blue. Legends: PA = pulmonary artery; Ao = aorta; RV = right ventricle; LV = left ventricle; 3V = three vessel; 4CH = four-chamber; 5CH = five-chamber; Trv = transverse; TR: tricuspid regurgitation.

Figure 6.

Tomographic Ultrasound Imaging in a fetus with transposition of the great arteries. A) B-mode imaging. (1) In the three-vessel and trachea view, only the aorta is visualized, leaving the right ventricle; (2) the five chamber view demonstrates a vessel that bifurcates (pulmonary artery) connected to the left ventricle; (3) the four-chamber view is normal. B) Color Doppler imaging. (1) the three vessel view shows the aorta connecting to the right ventricle; (2) the pulmonary artery leaves the left ventricle and bifurcates. Legends: PA = pulmonary artery; Ao = aorta; RV = right ventricle; LV = left ventricle; 3V = three vessel; 3VT = three-vessel and trachea; SVC = superior vena cava; 4CH = four-chamber; 5CH = five-chamber; Trv = transverse

Table VI compares diagnoses established by the review of volume datasets displayed with TUI and postnatal diagnoses (isolated VSDs were excluded). Among the 17 cases with a cardiac anomaly suspected by TUI or identified in the neonatal period, two were lost to follow-up. In one case, only a fetal echocardiogram performed by an independent pediatric cardiologist was available for comparison. Among the 15 patients for whom follow-up was available, there was perfect agreement between prenatal and postnatal diagnoses in 73.3% of the cases (11/15). The four discrepant diagnoses were: 1) in case 3, a fetus diagnosed with tetralogy of Fallot by TUI had a postnatal diagnosis of TGA associated with a subaortic VSD; 2) in case 7, coarctation of the aorta was suspected by both B-mode and color Doppler but was not confirmed in the neonate; 3) in case 10, a rhabdomioma was not visualized by both B-mode and color Doppler volume datasets displayed with TUI; and 4) in case 14, severe tricuspid regurgitation was detected in a fetus whose final diagnosis was Ebstein anomaly.

Table 6.

Comparison of prenatal diagnosis by TUI (B-mode imaging and color Doppler and postnatal diagnosis (n=17).

Case #	GA	TUI Diagnosis	Postnatal Diagnosis
1	23w5d	Tetralogy of Fallot	Tetralogy of Fallot
2	29w4d	DORV, TGA, ASD, VSD	DORV, TGA, ASD, VSD
3	24w1d	Tetralogy of Fallot	TGA
4	25w2d	Interrupted IVC with azygous continuation	Interrupted IVC with azygous continuation
5	25w3d	Coarctation of the aorta, VSD	Coarctation of the aorta, VSD
6	19w3d	HLHS, DORV, TGA	HLHS, DORV, TGA*
7	36w4d	Coarctation of the aorta	Normal
8	16w6d	AV canal	AV canal
9	31w3d	HLHS	HLHS
10	20w6d	Normal	Rhabdomioma
11	22w6d	Tetralogy of Fallot	Lost to follow-up
12	20w3D	TGA	TGA
13	22w3d	Interrupted IVC with azygous continuation	Interrupted IVC with azygous continuation
14	37w0d	Severe tricuspid regurgitation	Ebstein anomaly
15	29w4d	Pulmonary atresia	Pulmonary atresia
16	28w0d	TGA	TGA
17	19w3d	Tetralogy of Fallot	Lost to follow-up

Legend: DORV: double outlet right ventricle; TGA: transposition of the great arteries; ASD: atrial septal defect; VSD: ventricular septal defect; IVC: inferior vena cava; RV: right ventricle; LV: left ventricle; HLHS: hypoplastic left heart syndrome; AV: atrioventricular; TUI: Tomographic Ultrasound Imaging.

^*In case #7, the diagnosis was confirmed in the prenatal period independently by a pediatric cardiologist. The pregnancy was lost to follow-up and no neonatal information was available.

Sensitivity, specificity, positive- and negative-predictive values of TUI for the diagnosis of congenital heart disease, excluding cases of isolated VSDs, are displayed in Table VII. For this analysis, cases 3 and 14 (Table VI) were considered as true positive diagnoses because, although a specific diagnosis was not made, a significant cardiac anomaly was suspected by TUI.

Table 7.

Sensitivity, specificity, positive- and negative-predictive values of TUI to detect congenital heart disease.

	# of patients available for analysis	Prevalence	Sensitivity	Specificity	Positive predictive value	Negative predictive value
B-mode imaging	115†	12.2% (14/115)	85.7% (12/14)	100.0% (101/101)	100.0% (12/12)	98.0% (101/103)
Color Doppler	98‡	14.3% (14/98)	92.9% (13/14)	98.8% (83/84)	92.9% (13/14)	98.8% (83/84)

^†Complete follow-up was available in 131/172 patients scanned by B-mode imaging (76.2%) (excluding cases of isolated VSDs). Sixteen patients were excluded from this portion of the analysis: one patient whose images were interpreted as tetralogy of Fallot but who was found at the time of surgery to have a subaortic VSD associated with transposition of the great arteries, and 15 patients whose ultrasonographic images were considered to be of poor diagnostic quality.

^‡Complete follow-up was available in 105/142 patients scanned by color Doppler (61.0%) (excluding cases of isolated VSDs). Seven patients were excluded from this portion of the analysis: in all cases, the ultrasonographic images were considered to be of poor diagnostic quality.

TUI: Tomographic Ultrasound Imaging; VSF: ventricular septal defects.

Among fetuses with a VSD suspected by B-mode imaging and color Doppler (n=23), two cases (8.7%) were confirmed after delivery, one case (4.3%) was diagnosed as an AV canal, and another (4.3%) case was diagnosed as double inlet single ventricle. One of the three (33.3%) isolated VSDs detected after delivery was missed by prenatal examination of STIC volume datasets displayed with TUI.

Discussion

The results of this study indicate that: 1) the four-and five-chamber and three-vessel and trachea views can be visualized in the majority of volume datasets automatically sliced with TUI; 2) color Doppler provides additional diagnostic information in 10% of the three-vessel views, 12% of the four-chamber views, and 14% of the five-chamber views displayed with this modality; 3) the additional information provided by color Doppler helped to establish a correct diagnosis in 31% of the fetuses with cardiac anomalies other than isolated VSDs; and 4) the diagnoses established by TUI agreed with the postnatal diagnoses in 73.3% of the fetuses, with good sensitivity, specificity, positive- and negative-predictive values for the diagnosis of congenital heart disease.

Technical aspects of 4D ultrasonography with STIC have been described by several investigators. [8, 13, 21–24, 28–30, 32, 33, 71, 72, 74] However, a systematic evaluation of the capability of obtaining specific planes of section with this diagnostic modality has been reported once.[72] Viñals et al.[72] studied 100 fetuses who were examined by 4D ultrasonography with STIC, and whose volume datasets were evaluated by a specialist in fetal echocardiography who was not involved in volume acquisition. Standard cardiac planes were obtained by scrolling through the volume datasets from the upper abdomen to the mediastinum. Visualization rates for the four-chamber view, left and right ventricular outflow tracts, three-vessel view, and three-vessel and trachea views ranged from 81 to 100%, with the lowest visualization rates observed for structures located in the abdomen or upper mediastinum. TUI provides an alternative approach to manually scrolling through the volume datasets to obtain these cardiac views. The difference between the two approaches is that the volume dataset is automatically sliced and, thus, it comes as no surprise to us that the visualization rates for cardiac structures reported by Viñals et al. [72] are similar to those reported in the current study.

The view with the poorest visualization rates in our study was the three-vessel view demonstrating bifurcation of the main pulmonary artery, the fourth plane of section proposed by Yagel et al.[73] This could reflect the short distance between this plane and the three-vessel and trachea view. Therefore, visualization of both planes using automatic slicing techniques with a fixed distance between the planes of section may have forced the display of one plane to the detriment of the other. Although the distance between each slice can be adjusted by the software and thus potentially correct this problem, we wanted to evaluate how this new technique would perform with minimum operator interference. It is important, however, to be aware of this limitation, since if this view is not automatically displayed, the examiner may need to use other techniques to explore the volume dataset in order to visualize the pulmonary artery and its bifurcation. Among these techniques, the examiner may choose to use a previously reported systematic approach for evaluation of the outflow tracts with STIC,[31, 32] whereby both the long axis view of the left ventricular outflow tract and the short axis view of the right outflow tract are displayed side-by-side on the same image, the “spin” technique reported by DeVore et al.,[22] or rendering techniques to visualize the relationship between both great arteries on the same image.[13, 24, 28, 29, 33]

The results of this study also corroborate the proposal of Chaoui and McEwing[14] that the examination of three-planes of section (four- and five-chamber and three-vessel views) with color Doppler may be sufficient to detect most cases of congenital heart disease. Indeed, these investigators subsequently reported that these three planes could be demonstrated in 31/35 healthy fetuses and in 24/27 fetuses with CHD in volume datasets acquired with color Doppler STIC.[13] In the current study, the four- and five-chamber and three-vessel views acquired with color Doppler were visualized in 98.2%, 97.0%, and 83.6% of the cases, respectively, and additional diagnostic information was observed in 12%, 14% and 10% of these views when compared to B-mode imaging only. More importantly, color Doppler helped to establish the correct diagnosis of pulmonary atresia and coarctation of the aorta in two cases in which the previous findings by B-mode imaging had shown only disproportional ventricular size.

The role of sonographic tomography in clinical practice remains to be determined. Benacerraf et al.[6] reported preliminary findings in 25 pregnancies scanned during the second trimester, in which five volume datasets encompassing the fetal head, face, chest, abdomen, and limbs were acquired for offline later analysis. Volume datasets were examined by physicians who were not involved in volume acquisition, and the visualization rates for fetal anatomical structures and time to complete the examination (including volume acquisition and review) were calculated. Complete structural surveys were obtained in 20 of the 25 fetuses. In one of the 5 incomplete surveys, a face was visualized by neither 3D nor 2D ultrasound. Portions of the hands and feet were not visualized in the other four cases. Importantly, the time required to complete the anatomical surveys was decreased by half using 3D ultrasonography (13.9 minutes vs. 6.6 minutes, p<0.001). With the availability of software to automatically slice the volume datasets, busy clinical practices may welcome this approach.

Leung et al.[44] have recently reported the use of automatic slicing of 3D volume datasets (Multi-Slice View, Accuvix, Medison, Seoul, Korea) for the examination of the fetal spine, face, brain and heart in a group of 35 fetuses scanned during the second trimester of pregnancy, and demonstrated the feasibility of diagnosing neural tube defects, facial clefts, and holoprosencephaly. The current study extends these observations to prenatal diagnosis of congenital heart disease. Although most cases of congenital heart disease other than VSDs were correctly diagnosed by exploring the volume datasets with the novel TUI display, this portion of our study needs to be interpreted with caution. Despite removing patient identifiers and including only cases examined at least 6 months before the conduction of this study, the possibility of recall bias exists. Therefore, our results need to be independently confirmed.