不同扫描方式对CCTA人工智能重建与分析的影响

Effects of Different Scanning Methods on Artificial Intelligence Reconstruction and Analysis in CCTA

目的分析与评价两种不同扫描模式对于冠状动脉CT血管成像(Coronary Computed Tomography Angiography,CCTA)人工智能(Artificial Intelligence,AI)后处理结果评估的影响。方法选取本院2021年4月至12月收治的100例行冠状动脉CT血管造影(CT Angiography,CTA)复查的患者,将患者随机分为A组和B组,每组50例。A组和B组再按照心率将每组分为两个亚组:心率≤75次/min为A1组(33例)、B1组(35例),心率>75次/min为A2组(17例)、B2组(15例)。A组使用容积扫描模式,B组使用螺旋扫描模式。将所有图像传送至数坤AI后处理站进行重建与分析,按照不同亚组分别对AI自动重建的容积再现(Volume Rendering,VR)图像和曲面重建(Curved projection Reformation,CPR)图像进行主观质量评分;通过对冠状动脉兴趣区进行测量,包括CT值、标准差(Standard Deviation,SD)并计算对比噪声比(Contrast to Noise Ratio,CNR)、信噪比(Signal to Noise Ratio,SNR),以测量客观图像质量;比较A组和B组的辐射剂量。结果在VR主观图像质量评分中,A1组与B1组差异无统计学意义(P>0.05),A2组图像质量评分高于B2组,差异有统计学意义(P<0.05);在CPR主观图像质量评价中,A1与B1相比无统计学差异(P>0.05),A2组图像质量评分低于B2组,差异有统计学意义(P<0.05);客观图像质量对比中,A组与B组图像在低心率亚组前降支远段的SNR值比较差异无统计学意义(P=0.8091),其余所有节段的SNR值B1组均显著高于A1组(P<0.05);而高心率亚组中,远段节段的SNR值B2组均小于A2组,近中段节段的SNR值则表现为B2组均大于A2组(P<0.05)。结论采用容积扫描AI后处理的图像在心率≤75次/min时,与螺旋扫描的图像质量相近;心率>75次/min时,螺旋扫描的图像质量更好,但由于容积扫描辐射剂量大幅度减低,因此在心率较低(≤75次/min)的情况下推荐使用容积扫描,心率较高时(>75次/min)使用螺旋扫描。

Objective To analyze and evaluate the effect of two different scanning modes on the evaluation of post-processing results of coronary computed tomography angiography(CCTA)and artificial intelligence(AI).Methods A total of 100 patients who underwent coronary CT angiography review in our hospital from April to December 2021 were selected,and the patients were randomly divided into group A and group B,with 50 cases in each group.Group A and group B were then divided into two subgroups according to heart rate:the heart rate was≤75 times/min for the A1(33 cases)and B1(35 cases)subgroups,and the heart rate was>75 times/min for the A2(17 cases)and B2(15 cases)subgroups.Group A used volume scanning mode,and group B used helical scanning mode.All images were sent to the Shukun AI post-processing station for reconstruction and analysis,and subjective quality scores were performed on the volume rendering(VR)images and curved projection reformation(CPR)images automatically reconstructed by AI according to different subgroups.By measuring the coronary area of interest,including CT value,SD value,and calculating the contrast signal to noise ratio(CNR)and signal to noise ratio(SNR)values,objective image quality measurement was performed,and the radiation doses between the two group A and group B were compared.Results In the VR subjective image quality score,there was no statistically significant difference between group A1 and group B1(P>0.05),and the image quality of group A2 was slightly higher than that of group B2(P<0.05).In CPR subjective image quality evaluation,there was no statistically significant difference between group A1 and group B1(P>0.05),the score of image quality of group A2 was lower than that of group B2,and the difference was statistically significant(P<0.05).In the objective image quality comparison,there was no statistically significant difference in the SNR value of the distal segment of the anterior descending artery in the low heart rate subgroup between group A and group B(P=0.8091),and the SNR values of all other segments in group B1 were significantly higher than group A1(P<0.05).While in the high heart rate subgroup,the SNR value of the distal segment in group B2 was lower than that in group A2,and the SNR value of the proximal and middle segments in group B2 was greater than that in group A2(P<0.05).Conclusion When the heart rate is≤75 times/min,the image quality of the post-processed images of the volume scanning AI was similar to that of the helical scanning.When the heart rate is>75 times/min,the image quality of the helical scanning is better.But because the radiation dose of volumetric scanning is greatly reduced,it is recommended to use volume scanning when the heart rate is low(≤75 times/min),and use helical scanning when the heart rate is high(>75 times/min).