基于深度学习评价低管电压个性化对比剂方案腹部增强CT的效果  

Evaluation of low radiation dose and personalized contrast dose abdominal CT:using deep learning model for automated segmentation on CT images

作　　者：王可[1] 魏守奕 刘建新[1] 张耀峰 王祥鹏 林子楹 郭小超[1] 张晓东[1] 王霄英[1] WANG Ke;WEI Shouyi;LIU Jiancin;ZHANG Yaofeng;WANG Xiangpeng;LIN Ziying;GUO Xiaochao;ZHANG Xiaodong;WANG Xiaoying(Department of Medical Imaging,Peking University First Hospital,Peking 100034,China;Beijing Smart Tree Medical Technology Co.,Ltd.,Beijing 100032,China)

机构地区：[1]北京大学第一医院医学影像科,北京100034 [2]北京赛迈特锐医疗科技有限公司,北京100032

出　　处：《实用放射学杂志》2022年第11期1864-1868,共5页Journal of Practical Radiology

基　　金：北京大学医学青年科技创新培育基金项目(BMU2021PYB021).

摘　　要：目的应用基于深度学习的肝脏、血管分割模型探究低管电压(70kVp)个性化对比剂注射方案在腹部CT增强扫描中的应用效果.方法回顾性选取行70kVp个性化对比剂注射方案(70kVp组)患者48例,体质量指数(BMI)14.85~23.05kg/m^(2);120kVp固定对比剂量注射方案(120kVp组)患者93例,BMI16.67~35.16kg/m^(2).使用基于U-net的深度学习模型对上述CT图像进行分割,获得各期肝脏、动脉期主动脉、门脉期门静脉分割结果及相应区域平均CT值.计算患者门脉期CT体型特异性剂量估计值(SSDE).应用独立样本t检验比较70kVp组与120kVp组不同期相、不同脏器/血管CT值的差异、SSDE的差异及对比剂用量、速度的差异,P<0.05为差异有统计学意义.结果无论全部患者,还是BMI<24kg/m^(2)的患者,与120kVp组相比,70kVp组各期肝脏CT值、门脉期肝脏强化程度、动脉期主动脉CT值、门脉期门静脉系统CT值均更高,且差异有统计学意义.70kVp组门脉期SSDE、碘对比剂剂量、对比剂注射速度均显著低于120kVp组(P<0.05).结论与传统120kVp固定对比剂剂量注射方案相比,70kVp个性化对比剂注射方案辐射剂量更低、对比剂用量及速度更低,可获得更高的肝脏CT值、血管CT值及门脉期肝脏强化程度.基于深度学习的肝脏、血管分割模型可应用于临床科研工作中,提升影像医生工作效率.Objective To investigate the performance of low radiation dose(70kVp)and personalized contrast dose abdominal enhanced CT by using deep learning model for liver,blood vessel segmentation.Methods This study was a retrospective study.Patients who underwent 70 kVp personalized contrast dose CT[70 kVp group,48 patients,body mass index(BMI)14.85-23.05 kg/m]and 120 kVp fixed contrast dose CT(120 kVp group,93 patients,BMI 16.67-35.16 kg/m^(2))were enrolled.A U-net segmentation model was used to segment and analysis the average CT value of liver(on all phases),aorta(on arterial phase),portal vein(on portal venous phase).Size-specific dose estimate(SSDE)in portal venous phase was calculated.Independent-sample t test was used to compare the difference of CT value,SSDE,contrast medium volume and speed between 70 kVp and 120 kVp groups.Results For all patients and for patients whose BM<24 kg/m^(2),CT value of liver,aorta,portal vein of 70 kVp group was higher than that of 120 kVp group(P<0.05).SSDE,contrast medium dose and speed of 70 kVp group was lower than 120 kVp group(P<0.05).Conclusion Low radiation dose(70 kVp)and personalized contrast dose abdominal CT seek higher liver and vessel CT value,with lower radiation dose and contrast agent dose compared to 120 kVp fixed contrast dose method.The U-net model enables accurate automated segmentation of liver,aorta,portal vein on contrast enhanced CT images,with promising implications for quantitative imaging studies.

关 键 词：深度学习 分割 辐射剂量 计算机体层成像 

分 类 号：R814.42[医药卫生—影像医学与核医学] R816.5[医药卫生—放射医学]