Using machine learning to ensure safety of patients who cannot remain still during magnetic resonance imaging

超高场磁共振成像（UHF-MRI，>= 7 T）提供了以前无法达到的图像质量和组织细节水平。然而，一些患者群体在磁共振成像期间可能不会保持静止，包括儿科患者和患有痴呆症、帕金森氏症、图雷特氏症和亨廷顿氏症的人。我们最近证明，扫描内患者运动可能会使患者发热增加至安全限值的3.1倍以上（Kopanoglu，MRM，2020）。尽管镇静可以减轻运动，但它可能会导致不良副作用，甚至导致住院治疗。我们证明，如果我们知道患者运动对UHF-MRI扫描动力学的影响，我们可以在扫描过程中纠正它们（Kopanoglu，ISMRM，2018）。不幸的是，这些影响无法测量，但我们使用机器学习方法（方法1）来估计这些变化的可行性研究产生了令人兴奋的初步结果。有了这个缺失的环节，我们可以确保那些在获取高质量图像时无法保持静止的患者的安全。

Ultrahigh field magnetic resonance imaging (UHF-MRI, >=7T) offers previously unattainable levels of image quality and tissue detail. However, several patient populations may not stay still during magnetic resonance imaging, including paediatric patients and people with dementia, Parkinson's, Tourette's and Huntington's. We have recently demonstrated that within-scan patient motion may increase patient heating by up to 3.1-fold above the safety limits (Kopanoglu,MRM,2020) Although sedation can mitigate movement, it may cause adverse side effects and even, with an adverse effect, hospitalization. We demonstrated that, if we know the effect of patient motion on UHF-MRI scan dynamics, we could instead correct them during the scan (Kopanoglu,ISMRM,2018). Unfortunately, these effects cannot be measured, but our feasibility study on using a Machine Learning method (Method-1) to estimate these changes yielded exciting initial results. With this missing link complete, we can ensure safety of patients who cannot remain still while acquiring high quality images.