三维人工智能定位技术在CT胸部扫描中的应用

Application of three dimensional artificial intelligence localization technology in CT chest scanning

目的:探讨三维人工智能(AI)定位技术在胸部CT平扫中的应用价值。方法:回顾性分析2020年9月至10月在南京大学医学院附属鼓楼医院因新型冠状病毒肺炎筛查接受胸部平扫CT检查的患者100例,采用区组随机分组法分为人工定位组和三维AI定位组,每组50例。2组患者使用相同的胸部扫描协议。测量2组患者的定位偏离距离、CT剂量指数(CTDI)和剂量长度乘积(DLP),并记录患者的检查用时。同时由2名放射医师对患者纵隔窗图像质量进行客观评价,并对肺窗图像进行主观评分。采用独立样本 t检验对2组的定位偏离距离、CTDI、DLP、图像质量客观指标和检查用时进行比较。采用χ2检验对图像质量主观评分的患者分布进行比较。 结果:三维AI定位组的总体定位偏移距离较人工定位组减少了42.86%[分别为(8.8±7.2)、(15.4±9.7)mm, t=3.65, P<0.01],CTDI降低了10.67%[分别为(6.7±2.6)、(7.5±2.5)mGy, t=0.59, P=0.04],DLP降低了13.33%[分别为(234±86)、(270±95)mGy·cm, t=1.98, P=0.02]。三维AI定位组的检查用时较人工定位组减少了29.91%[分别为(150±14)、(214±26)s, t=15.79, P<0.01]。2组图像的背景噪声以及降主动脉、竖脊肌的信号噪声比及图像主观评分患者分布差异无统计学意义( P>0.05)。 结论:使用三维AI定位技术进行胸部CT扫描,可以在保证图像质量的前提下,提高定位准确度,降低患者的辐射剂量,大幅提高工作效率。

Objective To evaluate the clinical application of three dimensional artificial intelligence(3D-AI)localization technology in chest CT scan.Methods A total of 100 patients who underwent chest CT for COVID-19 screening in Nanjing University Medical School Affiliated Drum Tower Hospital were collected from September 2020 to October 2020 were analyzed retrospectively.The patients were divided into manual positioning group(n=50)and 3D-AI automatic positioning group(n=50)with block randomization method.All patients were scanned with the same CT scanning protocol.The off-center distance,CT dose index(CTDI),dose length product(DLP)and CT examination time were measured and recorded.Quantitative image evaluation of mediastinal window images and qualitative image evaluation of chest window images were assessed by two radiologists.The off-center distance,CTDI,DLP,CT examination time and objective indexes of image quality of two groups were compared by independent sample t test.The quantitative image quality scores were compared withχ2 test.Results Compared with manual positioning group,the overall off-center distance of 3D-AI automatic positioning group was reduced by 42.86%[(15.4±9.7)vs.(8.8±7.2)mm,t=3.65,P<0.01],CTDI was reduced by 10.67%[(7.5±2.5)vs.(6.7±2.6)mGy,t=0.59,P=0.04],DLP was reduced by 13.33%[(270±95)vs.(234±86)mGy·cm,t=1.98,P=0.02],the average examination time was reduced by 29.91%[(214±26)vs.(150±14)s,t=15.79,P<0.01].There were no significant differences in the background noise,signal to noise ratio of descending aorta and erecting spinal muscle,and subjective score between two groups(P>0.05).Conclusion The 3D-AI automatic positioning technology can greatly improve the accuracy of patient positioning and reduce the radiation dose for chest CT imaging,and improve work efficiency with qualified chest CT image quality.