不同迭代算法对冠状动脉CTA人工智能重建与分析的影响

Effects of Different Reconstruction Ratios of Iterative Modes on Artificial Intelligence Reconstruction and Analysis of Coronary CTA

目的 探究冠状动脉CTA数据按照不同迭代算法比例进行重建后,对冠状动脉CTA人工智能软件重建与分析的影响。方法 本研究回顾性收集200例使用冠状动脉CTA扫描的患者,平均体质指数为(BMI22.65±3.01Kg/M^(2))kg/m^(2),平均年龄为55.20±9.10岁,平均心率为70.09±8.03bpm。原始数据分别采用AIDR 3D Enhance、AIDR 3D Strong(75%)进行重建,分别编为A、B组。两组图像均导入数坤冠状动脉人工智能软件进行后处理和分析,不进行手动矫正。采用双盲法对两组冠状动脉人工智能曲面重建(AI CPR)进行4分法主观图像质量评估;对两组图像自动分段的准确性进行人工复检;在CPR图像上测量右冠状动脉近段、远段、左主干,前降支近、段,回旋支近、中段的管腔CT值和临近脂肪图像的噪声标准差(Standard Deviation,SD),并计算信噪比(Signal to Noise Ratio,SNR)。结果 A、B两组人工智能重建图像在主观质量评分中,A组比B组获得更多的优秀质量图像(P<0.05);两组图像自动分段准确率基本一致(P>0.05);在客观图像质量评价中,冠状动脉各节段CT值无明显差异(P>0.05),A组图像SD值显著低于B组且SNR值显著高于B组(P<0.001)。结论 AIDR 3D Enhance算法能够显著提高冠状动脉人工智能软件重建的图像质量,图像噪声值显著降低,自动分段准确性能稍提升,可以在日常临床诊断中推广使用。

Objective To explore the effect of coronary CTA data reconstruction according to different iterative algorithm ratios on the reconstruction and analysis of coronary CTA artificial intelligence software.Methods This study retrospectively collected 200 patients with coronary CTA scan,the average body mass index was(BMI22.65±3.01Kg/M^(2))kg/m^(2),the average age was 55.20±9.10 years old,and the average heart rate was 70.09±8.03bpm.The original data were reconstructed with AIDR 3D Enhance and AIDR 3D Strong(75%),and were divided into groups A and B,respectively.Both groups of images were imported into Shukun coronary artificial intelligence software for post-processing and analysis without manual correction.Double-blind method was used to evaluate the subjective image quality of two groups of coronary artificial surface reconstruction(AI CPR).The CT values of the lumen of the proximal,distal,left main,proximal and proximal segments of the anterior descending branch,and the proximal and middle segments of the circumflex branch,and the standard deviation(SD)of the adjacent fat images,and the Signal to Noise ratio(Signal to Noise)was calculated.Ratio,SNR).Results In the subjective quality score of artificial intelligence reconstructed images in groups A and B,group A obtained more excellent quality images than group B(P<0.05);the accuracy of automatic segmentation of the two groups of images was basically the same(P>0.05).In the objective image quality evaluation,there was no significant difference in the CT value of each segment of the coronary artery(P>0.05).The SD value of group A was significantly lower than that of group B and the SNR value was significantly higher than that of group B(P<0.001).Conclusion AIDR 3D Enhance algorithm can significantly improve the image quality of coronary artery artificial intelligence software reconstruction,the image noise value is significantly reduced,and the automatic segmentation accuracy performance is slightly improved,which can be popularized and used in daily clinical diagnosis.