摘要
目的探讨老年卒中后认知障碍(CI)患者磁共振成像(MRI)纹理早期特征的价值。方法回顾性分析160例老年卒中患者72 h内T1加权MRI图像,比较海马和内嗅皮层纹理特征,并比较了有CI和无CI患者之间的差异。相关性研究确定了这些特征与蒙特利尔认知评估量表(MoCA)和简易精神状态检查量表(MMSE)临床评分的关联。随后与经典预后因素(年龄和性别)结合,构建了一个机器学习算法作为CI发生的预测模型。采用了大脑中动脉暂时性阻塞模型。结果在临床研究中,两个纹理特征、峰度和反向差异矩,显示出有CI与无CI患者之间的显著差异,并与MoCA和MMSE评分显著相关。该预测模型的准确率为(88±3)%。临床前模型显示出纹理特征与缺血区域对侧海马的神经密度之间存在显著相关性。结论MRI图像的纹理特征能够代表神经变化,可能成为老年卒中后CI早期预测筛查策略的一部分。
Objective To investigate the value of early magnetic resonance imaging(MRI)texture features in patients with post-stroke cognitive impairment(CI).Methods T1-weighted MRI from 160 stroke patients within 72 hours were retrospectively analyzed to compare the texture features of the hippocampus and entorhinal cortex,and to compare the differences between patients with and without CI.Correlation studies determined the association of these features with MoCA and MMSE clinical scores.The salient features were then combined with classical prognostic factors(age and gender)to construct a machine learning algorithm as a predictive model for CI occurrence.The middle cerebral artery temporary obstruction model was used.Results In clinical studies,two texture features,kurtosis and reverse difference moments,showed significant differences between patients with and without CI and were significantly associated with MoCA and MMSE scores.The accuracy of the prediction model is(88±3)%.Preclinical models showed a significant correlation between texture features and nerve density in the contralateral hippocampus of the ischemic region.Conclusion The texture features of MRI can represent neural changes and may be a part of the early prediction and screening strategy for CI after stroke.
作者
马玉宝
陈晶
刘华卿
Ma Yubao;Chen Jing;Liu Huaqing(Department of Neurology,the First Medical Center of the People's Liberation Army General Hospital,Beijing 100853,China)
出处
《脑与神经疾病杂志》
2025年第12期756-760,共5页
Journal of Brain and Nervous Diseases
基金
国家自然科学基金面上项目(82372440)
解放军总医院青年自主创新科学基金项目(22QNFC082)
青岛市医药卫生科研指导项目(2023-WJZD196)。
关键词
认知障碍
磁共振成像
纹理特征
卒中
Cognitive impairment
Magnetic resonance imaging
Texture feature
Stroke
