Kligerman Seth, Bolster Ferdia, Mitchell Jason, Henry Travis, Jeudy Jean, White Charles S
*Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD †Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA.
J Thorac Imaging. 2017 Mar;32(2):127-135. doi: 10.1097/RTI.0000000000000257.
The aim of the study was to assess the detection of congenital cardiovascular anomalies (congenital heart disease) in neonates and infants using model-based iterative reconstruction (MBIR) algorithm compared with hybrid iterative reconstruction (HIR) and filtered back projection (FBP) reconstructions on axial computed tomography (CT) performed at minimum scanner dose.
Over 1 year, all CT angiographies performed in infants below 3 months of age with congenital heart disease were assessed retrospectively. All were scanned on a 256-slice CT (Brilliance iCT) using single axial rotation at minimum allowable scanner dose (80 kV/10 mAs), with patients free-breathing. Intravenous contrast was 1 mL/kg. Scan reconstruction was 0.9 mm/0.45 mm overlap, reconstructed with FBP, HIR (iDose5), and MBIR (IMR2). The 3 reconstructions per study were anonymized and randomized. Four cardiac radiologists (23, 9, 7, and 6 y experience) evaluated each reconstruction on a workstation for presence of an atrial septal defect, a ventricular septal defect, patent ductus arteriosus, and surgical shunt or anomalies of the aorta, pulmonary arteries, and pulmonary veins. Unevaluable structures were classified as nondiagnostic. Gold standard was surgery or both echocardiogram and cardiac catheterization. The sensitivity, specificity, and accuracy were determined for each reconstruction.
Fifteen scans in 14 infants met the inclusion criteria, with a total of 48 anomalies. Pooled sensitivity for MBIR of 0.82 (range, 0.75 to 0.9) was significantly better than those for FBP (0.58; range, 0.54 to 0.6; P<0.001) and HIR (0.67; range, 0.60 to 0.79; P<0.001). Pooled accuracy of MBIR, HIR, and FBP was 0.91, 0.84, and 0.81, respectively. Readers deemed 39 and 15 structures nondiagnostic with FBP and HIR, respectively, versus 2 with MBIR (MBIR-FBP, MBIR-HIR, P<0.0001). The CTDIvol, DLP, and estimated dose for all cases was 0.52 mGy, 4.2 mGy×cm, and 0.16 mSv.
MBIR significantly improves the detection of congenital anomalies in neonates and infants undergoing CT angiography at minimum allowable dose.
本研究旨在评估在最低扫描剂量下，使用基于模型的迭代重建（MBIR）算法与混合迭代重建（HIR）和滤波反投影（FBP）重建技术，在轴向计算机断层扫描（CT）上对新生儿和婴儿先天性心血管异常（先天性心脏病）的检测情况。
回顾性评估1年多来在3个月以下患有先天性心脏病的婴儿中进行的所有CT血管造影检查。所有检查均在256层CT（Brilliance iCT）上进行，采用单次轴向旋转，扫描剂量为最低允许剂量（80 kV/10 mAs），患者自由呼吸。静脉注射造影剂剂量为1 mL/kg。扫描重建层厚为0.9 mm/0.45 mm重叠，分别采用FBP、HIR（iDose5）和MBIR（IMR2）进行重建。每项研究的3种重建图像均进行了匿名化和随机化处理。4位心脏放射科医生（分别有23年、9年、7年和6年经验）在工作站上对每种重建图像进行评估，以检测房间隔缺损、室间隔缺损、动脉导管未闭以及手术分流情况，或主动脉、肺动脉和肺静脉的异常情况。无法评估的结构归类为非诊断性。金标准为手术或超声心动图和心导管检查结果。确定每种重建技术的敏感性、特异性和准确性。
14例婴儿的15次扫描符合纳入标准，共发现48处异常。MBIR的合并敏感性为0.82（范围为0.75至0.9），显著优于FBP（0.58；范围为0.54至0.6；P<0.001）和HIR（0.67；范围为0.60至0.79；P<0.001）。MBIR、HIR和FBP的合并准确性分别为0.91、0.84和0.81。放射科医生分别认为FBP和HIR有39处和15处结构为非诊断性结构，而MBIR仅有2处（MBIR与FBP、MBIR与HIR比较，P<0.0001）。所有病例的CTDIvol、DLP和估计剂量分别为0.52 mGy、4.2 mGy×cm和0.16 mSv。
在最低允许剂量下对接受CT血管造影检查的新生儿和婴儿，MBIR显著提高了先天性异常的检测率。
