中华放射医学与防护杂志

杨哲,张翊婷,鲍永华,等.人工智能照射野自动控制技术在胸部数字X射线摄影中的应用研究[J].中华放射医学与防护杂志,2026,46(7):707-711.Yang Zhe,Zhang Yiting,Bao Yonghua,et al.Application of AI-based automatic collimation in chest digital radiography[J].Chin J Radiol Med Prot,2026,46(7):707-711

人工智能照射野自动控制技术在胸部数字X射线摄影中的应用研究

Application of AI-based automatic collimation in chest digital radiography

投稿时间：2025-10-27

DOI：10.3760/cma.j.cn112271-20251027-00378

中文关键词: 数字X射线摄影人工智能照射野自动控制技术智能定位辐射剂量

英文关键词:Digital radiography Artificial intelligence Automatic collimation Intelligent positioning Radiation dose

基金项目:

作者	单位	E-mail
杨哲	浙江大学医学院附属第一医院放射科, 杭州 310003
张翊婷	浙江大学医学院附属第一医院放射科, 杭州 310003
鲍永华	浙江大学医学院附属第一医院放射科, 杭州 310003
童雪明	浙江大学医学院附属第一医院放射科, 杭州 310003
刘星雨	浙江大学医学院附属第一医院放射科, 杭州 310003
汪启东	浙江大学医学院附属第一医院放射科, 杭州 310003	wangqidong@zju.edu.cn

摘要点击次数: 1988

全文下载次数: 15

中文摘要:

目的探究人工智能(AI)照射野自动控制技术在胸部数字X射线摄影(DR)辐射剂量控制和肺野定位精度优化中的应用价值。方法采用简单随机抽样法回顾性纳入胸部DR患者3 048例,均固定采用125 kV、自动曝光控制(AEC)的摄影条件。其中AI控制组1 524例,采用AI智能定位及AI照射野自动控制技术;对照组1 524例,采用常规定位,手动照射野控制。比较两组间的剂量面积乘积(DAP)、摄影照射野面积、照射野中心与肺野中心偏差以及影像质量评分等指标。结果 AI控制组的DAP为[0.66(0.50, 0.78) dGy·cm²],低于对照组的[0.86(0.70, 1.05)dGy·cm²](Z=-24.98, P＜0.01);AI控制组摄影照射野面积为[1 578.10(1 419.85, 1 711.55) cm²],低于对照组的[1 600.00(1 496.36, 1 720.00) cm²] (Z=-5.89, P＜0.01);摄影照射野中心与患者肺野中心的偏差值同样表现为AI控制组[1.48(0.85, 2.24)cm]小于对照组[1.58(1.01, 2.43)cm] (Z=-4.19, P＜0.01);AI控制组影像质量评分均在4分以上,与对照组差异无统计学意义(P＞0.05)。结论胸部DR应用AI照射野自动控制技术可显著降低受检者辐射剂量,有效减少照射野面积,并提高肺野中心的定位精度。

英文摘要:

Objective To investigate the application value of artificial intelligence (AI)-based automatic collimation in radiation dose control and lung field localization accuracy optimization in chest digital radiography. Methods A total of 3 048 patients who underwent chest digital radiography were retrospectively enrolled via simple random sampling. All patients were examined with the X-ray exposure parameters fixed at 125 kV and automatic exposure control mode. The AI control group (n=1 524) underwent AI-based positioning and automatic collimation, while the control group (n=1 524) was examined with conventional positioning and manual collimation. The dose area product, collimation area, center deviation between collimation field and lung field, and image quality scores were compared between the two groups. Results The dose area product in the AI control group was [0.66 (0.50, 0.78) dGy·cm²], which was lower than that in the control group [0.86 (0.70, 1.05) dGy·cm² ] (Z=-24.98, P<0.01). The size of the collimation area in the AI control group was [1 578.10 (1 419.85, 1 711.55) cm²], which was significantly lower than that in the control group [1 600.00 (1 496.36, 1 720.00) cm² ] (Z=-5.89, P<0.01). The center deviation between the collimation field and lung field in the AI control group was [1.48 (0.85, 2.24) cm], which was also lower than that in the control group [1.58 (1.01, 2.43) cm ] (Z= -4.19, P<0.01). The image quality scores of the AI control group were all above 4 points, with no statistically significant difference compared to the control group (P>0.05). Conclusions The application of AI-based automatic collimation in chest digital radiography can significantly reduce the radiation dose to examinees, effectively decrease the collimation area, and improve the localization accuracy of the lung field center.

HTML 查看全文查看/发表评论下载PDF阅读器

关闭