基于多模态神经影像的脑年龄估值差的综合研究

Comprehensive exploration of brain age gap estimation based on multimodal neuroimaging

目的:基于多模态神经影像探究脑年龄估值差(brain age gap estimation,BrainAGE)与非影像衍生标记物(noni-maging derived phenotypes,Non-IDPs)的关系。方法:以英国生物银行27 842例受试者的6种影像模态[T_(1)WI、弥散加权成像(diffusion-weighted imaging,DWI)、磁敏感加权成像(susceptibility-weighted imaging,SWI)、T_(2)WI、静息态功能成像(resting-state fMRI,rsfMRI)和任务态功能成像(task fMRI,tfMRI)]共7种特征集(FSL、Freesurfer、DWI、SWI、T_(2)WI、tfMRI、rsfMRI)为基础,采用相关向量回归模型对大脑年龄进行预测,并采用平均绝对误差(mean absolute error,MAE)评估模型的性能;将经偏差校正后的BrainAGE与223个Non-IDPs进行回归分析,以探究BrainAGE与Non-IDPs的关系。结果:相关向量回归模型预测脑年龄的MAE为2.767年。通过多元线性回归分析发现服用治疗药物的数量、全谷物摄入量、糖尿病诊断、收缩压、心室率以及吸烟状况6个Non-IDPs与BrainAGE之间存在显著相关。结论:BrainAGE是一项综合性脑健康评估指标,需要考虑多种健康信息和生活方式来进行综合分析。

Objective To explore the relationship between brain age gap estimation(BrainAGE) and non-imaging derived phenotypes(Non-IDPs) based on multimodal neuroimaging.Methods Brain age was predicted using a correlation vector regression model and the performance of the model was assessed using mean absolute error(MAE) based on the data of 27 842subjects from UK Biobank involving in 6 imaging modalities(T_(1)WI,diffusion-weighted imaging(DWI),susceptibility-weighted imaging(SWI),T_(2)WI,resting-state fMRI(rsfMRI) and task fMRI(tfMRI)) and 7 feature sets(FSL,Freesurfer,DWI,SWI,T_(2)WI,tfMRI and rsfMRI);bias-corrected BrainAGE was regressed against 223 Non-IDPs to explore the relationship between BrainAGE and Non-IDPs.Results The correlation vector regression model showed an MAE of 2.767 years for BrainAGE.Significant correlations were found by multiple linear regression between 6 Non-IDPs and BrainAGE including medicine intake,whole grain intake,diagnosed diabetes,systolic blood pressure,ventricular rate and smoking status.Conclusion BrainAGE is a comprehensive brain health assessment metric involving in complicated health and lifestyle information.