人工智能辅助压缩感知技术在上腹部T2WI压脂序列中的应用  

Application of artificial intelligence-assisted compression sensing technology in upper abdominal fat-suppressed T2WI sequence

作　　者：雷漫诗 邓锶锶 汪昕蓉 黄锦彬 向青 熊安妮 孟占鳌 Manshi Lei;Sisi Deng;Xinrong Wang;Jinbin Huang;Qing Xiang;Anni Xiong;Zhan'ao Meng(Guangzhou Xinhua University,Guangzhou 510520,China;Department of Radiology,the Third Affiliated Hospital of Sun Yat-sen University,Guangzhou 510630,China;Bayer Healthcare Co.,Ltd.,Guangzhou 510730,China;Guangzhou Health Science College,Guangzhou 510450,China)

机构地区：[1]广州新华学院,510520 [2]中山大学附属第三医院放射科,广州510630 [3]拜耳医药保健有限公司,广州510730 [4]广州卫生职业技术学院,510450

出　　处：《中华肝脏外科手术学电子杂志》2023年第5期551-556,共6页Chinese Journal of Hepatic Surgery(Electronic Edition)

基　　金：广州市科技计划项目(202007030007)。

摘　　要：目的探讨人工智能辅助压缩感知(Acs)技术在上腹部T2WI压脂序列中的应用。方法本研究对象为2022年6月至2022年10月在中山大学附属第三医院行上腹部MRI检查的30例患者。患者均签署知情同意书,符合医学伦理学规定。其中男21例,女9例;年龄31~76岁,中位年龄55岁。所有患者均分别采用常规并行采集(PI)方案和Acs方案进行磁共振扫描,PI组采用轴位、门控触发、频率选择饱和技术(AX-T2WI-FS-RT),Acs组采用Acs技术(AX-T2WI-FS-BH-Acs)。客观评价指标包括图像信噪比(SNR)、对比噪声比(CNR)、扫描时间等。两组比较采用配对t检验。两组图像质量评分和病灶检出数量比较采用Wilcoxon检验。2名医师的图像质量评分一致性评估采用Kappa检验。结果Acs组图像SNR平均为24.3±8.2,明显高于PI组11.7±4.4(t=13.00,P<0.05)。Acs组图像CNR为4.2±2.3,亦明显高于PI组的2.2±1.3(t=9.20,P<0.05)。Acs组扫描时间为66 s,明显短于PI组的156 s。Acs组图像质量评分中位数为4.4(4.3,4.6)分,明显高于PI组的4.1(3.4,4.4)分(Z=3.98,P<0.05)。对于呼吸紊乱患者,Acs组图像质量评分为4.6(4.3,4.7)分,亦明显高于PI组的3.4(3.1,3.4)分(Z=3.80,P<0.05)。两组病灶检出数量分别为1(0,3)、1(0,2)个,差异无统计学意义(Z=0.50,P>0.05)。2名医师的图像质量评分一致性强(κ=0.96)。结论与上腹部PI组AX-T2WI-FS-RT序列相比,Acs技术的T2WI-FS-BH-Acs序列可在不降低病灶检出率的前提下,明显缩短扫描时间,提高图像质量,尤其可极大提高呼吸紊乱患者检查图像质量。Objective To investigate the application of artificial intelligence-assisted compressed sensing(Acs)technology in the upper abdominal fat-suppressed T2-weighted imaging(T2WI)sequence.Methods 30 patients underwent MRI examination of the upper abdomen in the Third Affiliated Hospital of Sun Yat-sen University from June 2022 to October 2022 were recruited in this study.The informed consents of all patients were obtained and the local ethical committee approval was received.Among them,21 patients were male and 9 female,aged 31-76 years with a median age of 55 years.All patients received MRI by conventional parallel imaging(PI)and Acs sequences,respectively.Axial position,gating and triggering,and frequency-selective saturation techniques(AX-T2WI-FS-RT)were adopted in the PI group,and Acs technique(AX-T2WI-FS-BH-Acs)was employed in the Acs group.Objective assessment indexes included image signal-to-noise ratio(SNR),contrast-to-noise ratio(CNR)and scanning time,etc.Comparison between two groups was performed by paired t test.The image quality score and the number of detected lesions between two groups were compared by Wilcoxon test.The consistency in the image quality scores between two physicians was evaluated by Kappa test.Results The average SNR in the Acs group was 24.3±8.2,significantly higher than 11.7±4.4 in the PI group(t=13.00,P<0.05).The CNR in the Acs group was 4.2±2.3,significantly higher than 2.2±1.3 in the PI group(t=9.20,P<0.05).The scanning time in the Acs group was66 s,significantly shorter than 156 s in the PI group.The median image quality score in the Acs group was 4.4(4.3,4.6),significantly higher than 4.1(3.4,4.4)in the PI group(Z=3.98,P<0.05).For patients with respiratory disorders in the Acs group,the image quality score was 4.6(4.3,4.7),significantly higher than 3.4(3.1,3.4)in the PI group(Z=3.80,P<0.05).The number of detected lesions in two groups was 1(0,3)and 1(0,2),and no significant difference was observed(Z=0.50,P>0.05).The consistency in the image quality score was high between two p

关 键 词：磁共振成像 人工智能辅助压缩感知(Acs) 深度学习图像重建 屏气序列 呼吸触发 人工智能(AI) 光梭成像 

分 类 号：R445.2[医药卫生—影像医学与核医学] TP18[医药卫生—诊断学]