MDCT Evaluation of Thoracic Aortic Anomalies in Pediatric Patients and Young Adults: Comparison of Axial, Multiplanar, and 3D Images

OBJECTIVE. The objective of our study was to compare accuracies of axial, multiplanar, and 3D volume-rendered images in the diagnosis of thoracic aortic anomalies in pediatric patients and young adults. MATERIALS AND METHODS. Fourteen patients, 17 days to 20 years old, with thoracic aortic anomalies underwent MDCT using axial, multiplanar, and 3D volume-rendering imaging. All images were reviewed by three radiologists for position of the aortic arch, coarctation, vascular compression of the airway, collateral vessel formation, and aortopulmonary shunts (patent ductus arteriosus). Final diagnosis was determined by echocardiography, conventional angiography, bronchoscopy, or surgery. Diagnostic accuracy, sensitivity, and interobserver agreement were evaluated. RESULTS. Average accuracies (average of the three observers for a correct diagnosis) were greater than or equal to 96% for diagnoses of aortic position and airway narrowing on all image types. For the diagnosis of coarctation, average sensitivities (average of the three observers for a true diagnosis) were 73% for axial, 100% for multiplanar, and 100% for 3D volume-rendered images. For the diagnosis of patent ductus arteriosus, average sensitivities were 78% for axial, 94% for multiplanar, and 89% for 3D volume-rendered images. No patients in this study had collateral vessel formation. For the diagnosis of absence of collateral vessel formation, average sensitivities were 100% for axial, 100% for multiplanar, and 100% for 3D volume-rendered images. There were no significant statistical differences in diagnostic performances, agreement with truth, or confidence scores among observers or imaging formats (p > 0.05). CONCLUSION. Axial, multiplanar, and 3D volume-rendered images serve equally well as methods for assessing the side of the aorta to diagnose anomalies. For evaluation of coarctation and patent ductus arteriosus, multiplanar and 3D volume-rendered images perform slightly better than axial images.