基于人工智能的3D自动定位技术在大脑MRI扫描中的应用研究  

A Comparative Study of AI-based 3D Automatic Localization Technology in Brain MRI

作　　者：王忠良 沈雄杰 钟俊杰 梁斯俊 朱高杰 孙治国 陈志刚 宋子龙 卢珊珊 梁杨 王宇轩 李牡兰 WANG Zhong-liang;SHEN Xiong-jie;ZHONG Jun-jie;LIANG Si-jun;ZHU Gao-jie;SUN Zhi-guo;CHEN Zhi-gang;SONG Zi-long;LU Shan-shan;Liang Yang;WANG Yu-xuan;LI Mu-lan(Department of Radiology,Nanao People's Hospital of Dapeng New District,Shenzhen 518000,Guangdong Province,China;Shenzhen Anke High-Tech co.,Ltd,Shenzhen 518000,Guangdong Province,China)

机构地区：[1]深圳市大鹏新区南澳人民医院医学影像科 [2]深圳安科高技术股份有限公司,广东深圳518000

出　　处：《中国CT和MRI杂志》2025年第4期53-56,共4页Chinese Journal of CT and MRI

摘　　要：目的探讨基于AI的3D自动定位技术在大脑MRI不同体位、多次扫描的对比研究价值。方法使用深圳安科SuperMark 1.5T磁共振成像仪扫描49名志愿者,在头部正位、抬头、低头、左斜、右斜5种摆位下采集T1 FLAIR轴位,T2 FLAIR轴位,T1 FLAIR冠状位和T2WI失状位头部MRI图像。根据不同体位下采集的图像与正位的差异度打分,并统计得分情况。结果头部冠状位和轴位扫描图像,在四种不同摆位下有82.14~96.43%的图像满足与正位的同一层面的对比观察。失状位图像只有51.85~85.6%的图像满足与正位的同一层面的对比观察。四种不同摆位下T1 FLAIR冠状位,T1 FLAIR轴位和T2 FLAIR轴位与正位的图像对比得分没有明显差异,尤其是在抬头和低头体位图像与正位图像呈高度相似。结论基于AI的3D自动定位方法相比传统的手动定位,能保证多次扫描的层面保持一致,有利于临床对大脑病变的追踪评估和研究。Objective Investigate the usefulness of AI-based 3D automatic localization system for comparing brain MRI data obtained from various body positions,Methods Anke SuperMark 1.5T MRI was used to scan 49 subjects.The T1 FLAIR axial,T2 FLAIR aixal,T1 FLAIR coronal,and T2WI sagittal brain MRI images were scanned using five different head positions:standard,upward,lower,left oblique,and right oblique.The degree of deviation from the orthostatic position was used to grade the images taken in various positions,and the scores were tallied,Results 82.14~96.43 percent of the head's coronal and axial scanning images in each of the four positions met the same standard of comparison observation with the orthostatic posture.The dislocation images only met the same degree of comparison observation with the orthostatic position in 51.85~85.6 percent of the cases.The T1 FLAIR coronal,T1 FLAIR axial,and T2 FLAIR axial positions and the orthostatic position in the four different poses did not significantly differ in their image comparison scores.This was especially true for the images of the superior elevation and inferior hypostatic positions,which exhibited a high degree of similarity with the orthostatic images,Conclusion Compared to traditional manual localization,the AI-based 3D automated localization method guarantees that the dimensions of numerous scans can be kept consistently.This is advantageous for clinical follow-up assessment and research on brain lesions.

关 键 词：人工智能 核磁共振 3D自动定位 头部MRI 

分 类 号：R445.2[医药卫生—影像医学与核医学]