摘要
目的 探讨人工智能压缩感知(ACS)技术在踝关节MRI中应用的可行性。资料与方法 前瞻性收集2023年9—10月在首都医科大学附属北京友谊医院行踝关节扫描的32例健康志愿者,应用3.0T MR进行基于ACS及并行成像(PI)技术的踝关节MRI,采集矢状位质子密度加权成像(PDWI)、冠状位PDWI、横轴位PDWI及矢状位T1WI,并将其分为ACS组和PI组,ACS组ACS综合加速倍数为3.05(ACS 5.0),PI组PI加速倍数为2(PI 2.0)。测量距骨、跟腱及软骨的信号强度以及拇长屈肌的信号强度及标准差,以拇长屈肌为背景噪声计算信噪比(SNR)和对比噪声比(CNR);对两组成像客观评估及主观评估进行统计学分析,各序列图像质量以PI 2.0为标准参考。结果 ACS组SNR和CNR较PI组更高,两组在矢状位PDWI序列各解剖结构比较,差异均有统计学意义(t=-2.937、-1.981、-4.058、-3.879,P<0.05);在冠状位PDWI序列上软骨SNR、距骨CNR差异有统计学意义(t=-3.310、-3.567;P=0.002,P<0.001);在轴位PDWI序列上ACS组与PI组距骨CNR、软骨CNR差异均有统计学意义(t=-4.270、-4.382,P<0.05)。主观评价2位影像诊断医师对两组图像质量评分观察者间一致性强(Kappa=0.977,P=0.009);两组图像质量评分差异均无统计学意义(Z=-0.248、-0.747,<0.001、-0.071,P>0.05)。ACS组与PI组采集总时间分别为337 s及610 s,与PI组相比,ACS组扫描总时间缩短44.8%。结论 基于ACS技术的踝关节MRI不仅能有效缩短扫描时间,还进一步改善图像质量,具有可行性。
Purpose To explore the feasibility of artificial intelligence compressed sensing(ACS)technique in ankle joint MRI.Materials and Methods From September to October 2023,32 healthy volunteers who underwent ankle joint scanning in Beijing Friendship Hospital,Capital Medical University were prospectively collected.MRI of the ankle joint based on ACS and parallel imaging(PI)technology was performed on 3.0T MR.The sagittal proton density weighted imaging(PDWI),coronary PDWI,transverse PDWI and sagittal T1WI were acquired,and all data were divided into test group and control group,with ACS to accelerate the multiples of 5(ACS 5.0)in test group,whereas PI speed ratio of 2(PI 2.0)in control group,respectively.The signal intensity of talus,achilles tendon and cartilage were measured,the signal intensity and standard deviation of the long hallux flexor were obtained,and the signal noise ratio(SNR)and contrast to noise ratio(CNR)were calculated via long hallux flexor as background noise.The data of objective and subjective evaluation of the two sequences were statistically analyzed,and the image quality of each sequence was evaluated via the standard reference of PI 2.0.Results SNR and CNR in ACS group were higher than those in PI group,and the anatomical structure of sagittal PDWI sequence between the two groups had statistical significance(t=-2.937,-1.981,-4.058,-3.879,P<0.05).There were significant differences in cartilage SNR and talus CNR in coronal PDWI sequence(t=-3.310,-3.567;P=0.002,P<0.001).In terms of axial PDWI sequence,there were statistically significant differences in talus CNR and cartilage CNR between ACS and PI groups(t=-4.270,-4.382,P<0.05).The subjective evaluation of the image quality scores of the two groups by the two diagnostic imaging doctors showed a strong observer consistency(Kappa=0.977,P=0.009).There was no significant difference in image quality scores between the two groups(Z=-0.248,-0.747,<0.001,-0.071,P>0.05).The total collection time of ACS group and PI group was 337 s and 610 s,respectively
作者
蒋学涛
程天馨
李菲菲
袁颖
江林
韦捷
徐辉
JIANG Xuetao;CHENG Tianxin;LI Feifei;YUAN Ying;JIANG Lin;WEI Jie;XU Hui(Department of Radiology,Beijing Friendship Hospital,Capital Medical University,Beijing 100050,China;不详)
出处
《中国医学影像学杂志》
CSCD
北大核心
2024年第11期1164-1169,共6页
Chinese Journal of Medical Imaging
基金
科技部重点研发计划(2022YFC2409403)
国家自然科学基金项目(82371904)。
关键词
踝关节
人工智能
压缩感知
磁共振成像
图像质量
Ankle joint
Artificial intelligence
Compressive sensing
Magnetic resonance imaging
Image quality
