人工智能压缩感知技术在单髋关节三维质子密度加权成像中的应用研究  

Research on the application of artificial intelligence compressed sensing technology in three-dimensional proton density weighted imaging of the unilateral hip joint

作　　者：张道恩 徐旭 李函宇 胡斯娴 袁野 张高峰 张晓咏 夏春潮[1] 李真林[1] Zhang Daoen;Xu Xu;Li Hanyu;Hu Sixian;Yuan Ye;Zhang Gaofeng;Zhang Xiaoyong;Xia Chunchao;Li Zhenlin(Department of Radiology,West China Hospital,Sichuan University,Chengdu 610041,China;Equipment and Materials Department,West China Hospital,Sichuan University,Chengdu 610041,China;Department of Imaging Center,First Affiliated Hospital of Anhui University of Chinese Medicine,Hefei 230031,China;Department of Medical Imaging,the Second Affiliated Hospital of Zunyi Medical University,Zunyi 563000,China;Clinical Science,Philips Healthcare,Chengdu 610041,China)

机构地区：[1]四川大学华西医院放射科,成都610041 [2]四川大学华西医院设备物资部,成都610041 [3]安徽中医药大学第一附属医院影像中心,合肥230031 [4]遵义医科大学第二附属医院医学影像科,遵义563000 [5]飞利浦医疗临床科研部,成都610041

出　　处：《中华放射学杂志》2024年第12期1431-1436,共6页Chinese Journal of Radiology

摘　　要：目的探讨人工智能压缩感知技术(CS-AI)在单髋关节三维质子密度加权成像(3D PDWI)中对图像质量的影响。方法2023年1至7月纳入于四川大学华西医院行单髋关节高分辨MR检查的健康志愿者67名,分别行加速因子(AF)为4、6、8、10的CS-AI 3D PDWI扫描,按上述AF分为CS-AI 4、CS-AI 6、CS-AI 8、CS-AI 104组,并以CS-AI 4作为标准。记录扫描时间,由1名高年资医师对股骨头、肌肉及皮下脂肪的信号强度及噪声进行测量,计算图像信噪比(SNR)、对比噪声比(CNR),同时由2名观测者对4组图像的整体图像质量及伪影进行主观评分,并采用Friedman秩和检验进行统计学分析。结果CS-AI 4、CS-AI 6、CS-AI 8、CS-AI 10的采集时间分别为5 min 49 s、3 min 54 s、2 min 56 s、2 min 22 s。相较于CS-AI 4,CS-AI 6、CS-AI 8、CS-AI 10的扫描时间分别缩短了32.95%、50.14%、59.31%。在客观评价方面,CS-AI 6、CS-AI 8、CS-AI 103组的股骨头、肌肉的SNR、CNR稍低于CS-AI 4(P<0.05),差异有统计学意义,但CS-AI 6、CS-AI 8、CS-AI 103组间差异无统计学意义(P>0.05)。在主观评价方面,CS-AI 8的图像整体质量评分3(3,4),与CS-AI 4和CS-AI 6差异无统计学意义(P>0.05);CS-AI 4和CS-AI 6的评分均为4(4,4)分;CS-AI 10评分为3(3,3)分,与其他组间差异有统计学意义(P均<0.05)。CS-AI 4、CS-AI 6、CS-AI 8、CS-AI 10的伪影评分分别为4(4,4)、4(4,4)、3(3,4)、2(2,3)分,AF=10时伪影较大(P<0.05),其余组间差异无统计学意义(P>0.05)。结论AF为8时,CS-AI技术的3D PDWI序列可获得满足临床诊断要求的单髋关节高分辨率图像,同时缩短扫描时间。Objective To explore the impact of artificial intelligence compressed sensing technology(CS-AI)on image quality in three-dimensional proton density weighted imaging(3D PDWI)of the unilateral hip joint.Methods High-resolution unilateral hip imaging was conducted on 67 healthy volunteers at West China Hospital of Sichuan University from January to July 2023.Imaging was performed by using CS-AI 3D PDWI sequence with acceleration factors(AF)of 4,6,8,and 10,respectively.According to the AF,all subjects were divided into 4 groups:CS-AI 4,CS-AI 6,CS-AI 8 and CS-AI 10,with CS-AI 4 serving as a reference.Recording the scan time,the signal and noise intensity of the femoral head,muscle,and subcutaneous fat were measured by a senior radiologist and the signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)were then calculated.Additionally,two observers provided ratings for overall image quality and artifacts in the 4 groups,and statistical analysis was performed using the Friedman rank-sum test.Results The acquisition times for CS-AI 4,CS-AI 6,CS-AI 8,and CS-AI 10 were 5 min 49 s,3 min 54 s,2 min 56 s and 2 min 22 s,respectively.Compared to CS-AI 4,the scanning time for CS-AI 6,CS-AI 8,CS-AI 10 were reduced by 32.95%,50.14%,59.31%,respectively.The objective evaluation revealed that the SNR and CNR of the femoral head and muscle in groups CS-AI 6,CS-AI 8,and CS-AI 10 were slightly lower than those in group CS-AI 4(P<0.05),and the differences were statistically significant.However,no statistically significant differences were found among the 3 groups(P>0.05).The subjective evaluation indicated that the overall image quality scores of group CS-AI 8[3(3,4)]did not significantly differ from those of group CS-AI 4 and CS-AI 6(P>0.05);The mean scores of group CS-AI 4 and CS-AI 6 were 4(4,4);Scores of group CS-AI 10 was 3(3,3),which statistically significant differ from those of the other groups(P<0.05).The artifacts rating for groups CS-AI 4,CS-AI 6,CS-AI 8 and CS-AI 10 were 4(4,4),4(4,4),3(3,4),and 2(2,3)respectively.When AF

关 键 词：磁共振成像 髋关节 人工智能 压缩感知 加速因子 

分 类 号：R445.2[医药卫生—影像医学与核医学] R684[医药卫生—诊断学]