摘要
目的:探讨人工智能辅助下的压缩感知(Artificial intelligence compressed sensing,ACS)技术在肩关节MRI成像中对图像质量和缩短扫描时间的应用价值.方法:招募健康志愿者27例,应用联影3.0T MR行肩关节MRI扫描,对受试者均应用并行采集技术(PI)以及ACS技术扫描肩关节横轴位PDWI、矢状位PDWI、冠状位T2WI、冠状位T1WI序列,4个序列使用PI技术的综合加速因子分别为1.77、1.72、1.60、1.87,使用ACS技术的综合加速因子分别为3.23、3.01、2.81、2.86.由两名医师从关节图像结构显示情况、脂肪抑制情况、伪影及图像整体情况四方面进行评分.在两组图像相同层面位置测量冈上肌、肱骨头均匀组织的信号强度(SI)和噪声值(SD),计算冈上肌和肱骨头SNR和CNR.采用配对样本t检验和Mann-Whitney U检验对主观评分和客观评价指标进行统计学分析.结果:PI在横轴位PDWI、矢状位PDWI、冠状位T2WI、冠状位T1WI的主观评分别为4.89±0.33、4.85±0.46、4.85±0.46、4.85±0.46,ACS 为 4.92±0.27、4.92±0.27、4.88±0.33、4.92±0.27,差异无统计学意义(P＞0.05)o PI 在横轴位 PDWI、矢状位 PDWI、冠状位 T2WI、冠状位 T1WI 测量冈上肌 SNR 分别为 25.08±4.08、30.40±6.93、10.37±1.95、17.53±3.17,低于 ACS 的 31.77±6.83、38.95±7.05、13.22±2.47、20.09±5.95,差异有统计学意义(P＜0.05).ACS 组总扫描时间为 241 s,较 PI 组(438 s)比较,减少了 45％.结论:应用ACS技术扫描肩关节MRI,可以在保证图像的质量的前提下,大幅减少扫描时间,为临床提供高效的检查.
Abstract
Objective:To explore the application value of artificial intelligence compressed sensing(ACS)technology in shoulder joint MRI imaging for image quality and shortening scanning time.Methods:Twenty-seven healthy volunteers were recruited,and the shoulder joint MRI scanning was performed using 3.0T MR scanner.Parallel acquisition technology(PI)and ACS technology were used to scan shoulder joint transverse PDWI,sagittal PDWI,coronal T2WI,and coronal T1WI sequences in all subjects.The combined acceleration factors for the four sequences were 1.77,1.72,1.60,and 1.87 by using the PI technique;3.23,3.01,2.81,and 2.86 by using the ACS technique,respectively.Two doctors scored the joint image structure display,fat suppression,artifacts and the overall image.The signal intensity(SI)and noise value(SD)of the homogenous tissue of the supraspinatus muscle and humeral head were measured at the same level of the two images,and the SNR and CNR of the supraspinatus muscle and humeral head were calculated.Statistical analyses of subjective scores and objective evaluation indicators were performed using paired-sample t-tests and Mann-Whitney U-tests.Results:The subjective scores of PI on axial PDWI,sagittal PDWI,coronal T2WI and coronal T1WI were 4.89±0.33,4.85±0.46,4.85±0.46 and 4.85±0.46,while those of ACS were 4.92±0.27,4.92±0.27,4.88±0.33 and 4.92±0.27,with no significant statistical difference(P＞0.05).The SNR of the supraspinatus muscle measured by PI on axial PDWI,sagittal PDWI,coronal T2WI and coronal T1WI were 25.08±4.08,30.40±6.93,10.37±1.95 and 17.53±3.17,which were lower than 31.77±6.83,38.95±7.05,13.22±2.47 and 20.09±5.95 of ACS,and the difference was statistically significant(P＜0.05).The total scan time in the ACS group was 241 s,which was a 45％reduc-tion compared to the PI group(438 s).Conclusion:The application of ACS technology to scan shoulder joint MRI can signifi-cantly reduce the scanning time and provide efficient examination for the clinic on the premise of ensuring the image quality.
基金项目
科技部重点研发计划(2022YFC2409403)
国家自然科学基金项目(52227814)
维普
万方数据