人工智能自动定位技术在不同体型患者胸部CT成像中的应用

Application of artificial intelligence automatic positioning technology in chest CT imaging of patients with different body types

目的:探讨不同体型患者胸部CT成像中AI自动定位技术对图像质量及辐射剂量的影响。方法:收集武汉大学中南医院行胸部CT检查的患者290例,按照定位方式不同分为两组,其中A组(AI自动定位操作组)及B组(常规手动定位操作组)扫描各145例,根据体质量指数(BMI)将两组患者进一步分为A1组和B1组:BMI<24kg/m^(2);A2组和B2组:24 kg/m^(2)≤BMI<28 kg/m^(2);A3组和B3组:BMI≥28 kg/m^(2)。对各组的定位扫描精度、客观图像质量、主观图像质量及辐射剂量进行评价。结果:AI定位组在不同体型患者中的偏中心距离均低于常规定位组,A1组、A2组及A3组偏中心距离较B1组、B2组及B3组分别减小38.67%、61.04%及65.90%(均P<0.01);A组各分组较B组在主动脉、气管、竖脊肌和肺野的CT值均增高,图像噪声均减低,信噪比(SNR)均增高,且差异均有统计学意义(均P<0.05);A1组、A2组及A3组CT容积剂量指数(CTDIvol)较B1组、B2组及B3组分别减少7.57%、10.57%及15.27%(均P<0.01);A1组、A2组及A3组剂量长度乘积(DLP)较B1组、B2组及B3组分别减少15.22%、16.74%及19.04%(均P<0.01)。结论:AI自动定位技术适用于不同BMI患者的胸部CT检查,可以优化检查流程,提高定位准确性,改善图像质量,并有效减少患者辐射剂量。

Objective:To investigate the effect of artificial intelligence(AI)automatic positioning technology on image quality and radiation dose in chest CT imaging of patients with different body types.Methods:A total of 290 patients who underwent chest CT examination in our hospital was enrolled in this study.145 patients were examined with AI automatic positioning operation(group A)and 145 patients were examined with conventional manual positioning operation(group B).According to the body mass index(BMI),the patients within the two groups were further divided into subgroups as:A1and group B1,BMI<24 kg/m^(2);group A2 and group B2,24 kg/m^(2)≤BMI<28 kg/m^(2);and group A3and groupB3,BMI≥28 kg/m^(2).The off-center distance,positioning accuracy,objective image quality,subjective image quality and radiation dose of each group were compared.Results:The off-center distance of patients with different body types in the AI positioning group was lower than that in the conventional manual positioning group.Compared with that in group B1,group B2 and group B3,respectively,the off-center distance of group A1,group A2 and group A3 decreased by 38.67%,61.04%and 65.90%,respectively(all P<0.01).The CT value and SNR of aorta,trachea,erector spine and lung field of group A were higher than those of group B.The image noise was lower than that of group B(all P<0.05).The CTDIvol of groups A1,A2 and A3 decreased by 7.57%,10.57%and 15.27%,respectively(all P<0.01).The DLP of groups A1,A2 and A3 decreased by 15.22%,16.74%and 19.04%,respectively(all P<0.01).Conclusion:AI automatic positioning technology is appropriate for chest CT examination of patients with different body types,and helps to simplify the examination workflow,improve the accuracy and the image quality of chest CT examination,and reduce the radiation dose of patients.