直肠癌MRI:人工智能辅助压缩感知技术与并行成像技术的对照研究  

MRI for rectal cancer:a controlled study of AI-assisted compression sensing technique and parallel imaging technique

作　　者：许桂晓[1] 刘海彬 吴尚远 黄颖怡 谢传淼[1] 何浩强[1] XU Gui-xiao;LIU Hai-bin;WU Shang-yuan(Department of Radiology,Sun Yat-sen University Cancer Center,State Key Laboratory of Oncology in South China,Collaborative Innovation Center for Cancer Medicine,Guangzhou 510060,China;不详)

机构地区：[1]中山大学肿瘤防治中心影像科、暨华南肿瘤学国家重点实验室/肿瘤医学协同创新中心,广东广州510060 [2]中山市人民医院磁共振科,广东中山528499

出　　处：《放射学实践》2023年第4期446-451,共6页Radiologic Practice

摘　　要：目的:通过与传统并行成像(PI)技术对比,探讨人工智能辅助压缩感知(ACS)技术在直肠癌MRI中的应用价值。方法:回顾性将2021年1月1日-2022年3月1日在本院经病理检查首诊为直肠癌的74例患者纳入研究。所有患者在确诊前行直肠3.0T MRI扫描,扫描序列主要包括横轴面、矢状面和冠状面FSE-T1WI和FSE-T_(2)WI,其中矢状面FSE-T_(2)WI分别采用ACS技术和传统PI技术。比较采用ACS和PI技术的扫描时间及两组图像上病灶和正常肠壁的信噪比(SNR)和对比噪声比(CNR)。由两位医师采用李克特量表分别对两组图像上病灶显示情况、病灶边缘清晰度、图像伪影和整体图像质量4项指标进行评分,并对两位医师评分的一致性进行评估。两位医师分别基于ACS和PI技术的矢状面FSE-T_(2)WI(结合横轴面和冠状面高分辨率T_(2)WI)进行直肠癌T分期,比较两种技术的诊断准确性。结果:采用ACS技术的检查时间明显短于PI技术[(92.03±1.16)vs.(144.36±0.97)s,P<0.001]。两组图像上SNR病灶和SNR正常肠壁的差异均无统计学意义(P>0.05),ACS组病灶和正常肠壁的CNR值较PI组高(P<0.05);图像质量主观评价结果显示,在病灶显示情况、病灶边缘清晰度、图像伪影和图像整体质量方面,ACS组的评分均高于PI组(P<0.001);两位观察者对图像质量4项评估指标的评分结果的一致性为良好或极好(Kappa值:0.603~0.929,P均<0.001)。两位医师基于ACS和PI技术诊断直肠癌T分期准确率分别为85.14%(63/74)和81.08%(60/74),诊断符合率无统计学差异(P>0.05)。结论:与传统并行采集技术相比,人工智能结合压缩感知(ACS)技术不仅可以缩短扫描时间,还可以改善图像质量,且不影响对直肠癌术前T分期的诊断准确性,具有较好的临床应用价值。Objective:By comparing with traditional parallel imaging(PI)technique,to explore the application values of artificial intelligence(AI)assisted compressed sensing(ACS)technique in rectal cancer MRI examination.Methods:Retrospectively,a total of 74 patients who were first diagnosed as rectal cancer by pathological examination from January 12021 to March 12022 in our hospital were included in this study.All patients underwent 3.0T MR scan before surgery,and the main sequences included transverse,sagittal and coronal FSE-T_(2)WI,of which sagittal FSE-T_(2)WI was scanned with ACS and conventional PI techniques,respectively.The signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)of the lesions(Les)and normal intestinal walls(NIW),and scan time of both sets of images obtained by ACS and PI techniques were compared.Four indexes,including lesion display,margin sharpness of lesions,artifacts,and overall image quality on two sets of images,were scored by two radiologists with the Likert scale respectively,and the consistency between the two ra-diologists was assessed.Preoperative T-staging of rectal cancer based on sagittal FSE-T_(2)WI with ACS and PI techniques(combined with high resolution transverse and coronal T_(2)WI)was performed by 2 radiologists,and its accuracy was analyzed.Results:The scan time of FSE-T_(2)WI with ACS technique was significantly shorter than that with PI technique[(92.03±1.16)vs.(144.36±0.97)s,P<0.001].Quantitative image analysis showed that there was no statistical difference between SNRLes and SNRNIW(P>0.05).The CNRLes to CNRNIW measured by ACS technique was significantly higher than that by PI technique(P<0.05).In the image quality subjective evaluation,the scores of lesion display,margin sharpness of lesions,artifacts,and overall image quality obtained by ACS technique were higher than those obtained by PI technique(all P<0.05).The consistency of the scoring results of the four evaluation indexes of image quality between the two radiologists was good-to-excellent(P<0.001).The T-staging

关 键 词：直肠肿瘤 人工智能 压缩感知 并行成像 磁共振成像 

分 类 号：R445.2[医药卫生—影像医学与核医学] R735.3[医药卫生—诊断学]