基于人工智能重建对颅内3D TOF MRA图像质量的改善  

Improvement of Intracranial 3D TOF MRA Image Quality Based on Artificial Intelligence Reconstruction

作　　者：陈楚玥 于雨洁 王茂雪[1] 王坤[1] 李茗[1] 马一鸣[1] 赵献策 韩鹏[1] Chen Chu-yue;Yu Yu-jie;Wang Mao-xue;Wang Kun;Li Ming;Ma Yi-ming;Zhao Xian-ce;Han Peng(Department of Radiology,Nanjing Drum Tower Hospital,Affiliated Hospital of Medical School,Nanjing University,Nanjing 210008,Jiangsu Province,China;Philips Health Care Division,Shanghai 200072,China)

机构地区：[1]南京大学医学院附属鼓楼医院医学影像科,江苏南京210008 [2]飞利浦医疗保健事业部,上海200072

出　　处：《中国CT和MRI杂志》2023年第10期60-63,共4页Chinese Journal of CT and MRI

摘　　要：目的探索人工智能重建技术对颅内3D TOF MRA图像质量的改善。方法回顾性收集2021年8月至2022年6月在南京大学医学院附属鼓楼医院进行头颅磁共振检查的43例受试者,平均年龄为56.95±35岁。采集不同压缩感知因子及重建技术的3D TOF MRM序列:CS4、CSAI4、CS8、CSAI8。定性分析四组图像血管显示能力、伪影、图像质量和诊断价值。采用F ri ed man检验比较四个序列间定性评分结果的差异。采用单因素重复测量方差分析定量比较四组图像CR值差异。结果图像质量定性分析中,在相同CS因子下,人工智能重建图像质量明显优于的常规CS重建图像(CSAI8:23.22±1.33 vs CS8;22.17+1.51,P<0.001;CSAI4:23.6±1.43 vs CS4:22.82±1.43,P<0.001)。图像定量分析中,人工智能重建的M RA图像CR值显著高于常规重建图像(大血管:CRCSAI4:11.10±0.75,CRCS4:9.10±0.48,CRCSAI8:10.75±0.75,CRCS8:7.86±0.38,P<0.001)。压缩感知加速因子不同时,CRCSAI8仍显著高于CRCS4(P<0.001)。结论人工智能重建技术可明显提高MRA图像质量。Objective To explore the artificial intelligence reconstruction technology to improve the quality of intracranial 3D TOF MRA image.Methods A total of 43 subjects with an average age of 56.95±35 years old underwent head magnetic resonance examination in Nanjing Drum Tower Hospital from August 2021 to June 2022 were retrospectively collected.3D TOF MRA sequences with different compressed sensing factors and reconstruction techniques were collected:CS4,CS_AI4,CS8,CS_AI8.Qualitative analysis of vascular display ability,artifact,image quality and diagnostic value of four groups of images.The Friedman test was used to compare the differences in qualitative scores between the four sequences.Single factor repeated measurement analysis of variance was used to quantitatively compare the CR values of the four groups.Results In the qualitative analysis of image quality,the quality of artificial intelligence reconstructed image was significantly better than that of CS reconstructed image with the a same CS factor(CS_AI8:23.22±1.33vs CS8:22.17±1.51,P<0.001;CS_AI4:23.60±1.43 vs CS4:22.82±1.43,P<0.001).In quantitative image analysis,the CR value of MRA reconstructed by artificial intelligence was significantly higher than that of conventional reconstructed images(great vessels:CRCS_AI4:11.10±0.75,CRCS4:9.10±0.48,CRCS_AI8:10.75±0.75,CRCS88:7.86±0.38,P<0.001).CRCS_AI8 was significantly higher than CRCS4(P<0.001),when compressive sensing acceleration factor was different.Conclusion Artificial intelligence reconstruction technology can obviously improve the quality of MRA image.

关 键 词：人工智能重建技术 压缩感知 3D时间飞跃法血管成像 

分 类 号：R322.12[医药卫生—人体解剖和组织胚胎学]