全模型迭代重建技术在心脏CT成像中应用的实验研究

Application of iterative model reconstruction iterative reconstruction in cardiac CT imaging--an animal experimental study

目的：探讨全模型迭代重建（IMR）技术在心脏MSCT成像中降低辐射剂量及控制图像质量的应用价值。方法选取实验猪10只，每头猪均进行常规剂量（A组）及在此基础上降低管电流[分别降低30%（B组）、50%（C组）、70%（D组）]的256层前瞻性心电门控心脏MSCT扫描，采用滤波反投影（FBP）、高级混合迭代重建（iDose4）和IMR重建，测量升主动脉根部的噪声、对比噪声比（CNR），分别对总体图像质量和冠状动脉图像质量进行评分，并对3种重建方法处理后所得结果进行比较，采用单因素方差分析比较3种重建方法处理的客观指标的差异，采用Kruskal-Wallis非参数检验和χ2检验比较3种重建方法处理的主观指标的差异。结果 A、B、C、D组采用IMR技术所得图像噪声分别为（15.1±6.1）、（18.8±5.5）、（22.1±4.8）、（33.0±4.0）HU，较FBP和iDose4显著降低，差异具有统计学意义（F值分别为82.77、90.71、96.59、95.51，P值均〈0.01）。A、B、C、D组IMR技术所得图像CNR分别为42.0±11.1、37.2±10.4、31.4±8.7、23.7±7.0，总体图像质量评分分别为5.0、（4.8±0.4）、（4.6±0.5）、（4.5±0.5）分，均较FBP和iDose4显著提高，差异具有统计学意义（F值分别为50.65、53.55、76.60、57.36，H值分别为20.96、15.63、18.66、23.56；P值均〈0.01）。A、B、C、D组IMR技术所得近段冠状动脉可诊断率分别为100%（40/40）、100%（40/40）、100%（40/40）、92%（37/40），A、B组较FBP和iDose4相仿，差异无统计学意义（P值均〉0.05），C、D组较FBP和iDose4显著提高，差异有统计学意义（χ2值分别为20.05、45.72， P值均〈0.01）；远段冠状动脉IMR技术可诊断率分别为100%（50/50）、98%（49/50）、90%（45/50）、78%（39/50），均较FBP和iDose4显著提高（χ2值分别为7.39、16.75、34.62、81.33，P值均〈0.05）。结论应用IMR较iDose4可更显著降低图像噪声，提高图像CNR和图像质量；应用IMR重建可在降低辐射剂量的情况下，保证图像质量。

Objective To evaluate the value of iterative modal reconstruction （IMR） for reducing radiation dose and controlling image quality in cardiac CT. Methods Ten pigs were included. All pigs were scanned on a 256-slice prospectively ECG-gated cardiac CT utilizing routine dose （group A） and tube current reduced by 30%（group B）, 50%（group C） and 70%（group D）, respectively. Filtered back projection （FBP）, hybrid iterative reconstruction （iDose4） and IMR were used for all data, respectively. Image noise and contrast-to-noise ratio （CNR） of ascending aortic root were measured, while overall image quality and coronary artery image quality was rated （five point scale）. All results reconstructed by FBP, iDose 4 and IMR were compared. Objective measurements were compared with one-way analysis of variance, and subjective assessments were compared with Kruskal-Wallis H test andχ2 test. Results Compared with that of FBP and＆amp;nbsp;iDose4, image noise of IMR was（15.1 ± 6.1）,（18.8 ± 5.5）,（22.1 ± 4.8）and（33.0 ± 4.0）HU, respectively in group A, B, C and D with significant reduction （F=82.77, 90.71, 96.59, 95.51 respectively, all P〈0.01）. Using IMR, groups A, B, C, D had higher CNR （42.0±11.1, 37.2±10.4, 31.4±8.7, 23.7±7.0;F=50.65, 53.55, 76.60, 57.36, all P〈0.01） and overall image quality （5.0 ± 0.0, 4.8 ± 0.4, 4.6 ± 0.5, 4.5 ± 0.5;H=20.96, 15.63, 18.66, 23.56, all P〈0.01） than FBP and iDose4. Using IMR, group A （100%, 40/40） and group B （100%, 40/40） had no significant difference （P〉0.05） in the diagnosis rates of proximal coronary arteries compared with that using FBP and iDose4, while group C （100%, 40/40） and group D（92%, 37/40） had significantly increased diagnosis rates （χ2=20.05, 45.72, both P〈0.01）. The diagnosis rates of distal coronary arteries of IMR reconstruction which were 100%（50/50）, 98%（49/50）, 90%（45/50）, 78%（39/50）, respectively in groups A, B, C, D had significant increase compared with that of FBP and iDose4 reconstruction （χ2=7.39, 16.75, 34.62, 81.33, all P〈0.05）. Conclusions IMR can significantly reduce image noise, improve CNR and image quality compared with iDose4. Application of IMR can reduce radiation dose but without compromising image quality.