Comprehensive classification models based on amygdala radiomic features for Alzheimer's disease and mild cognitive impairment
Comprehensive classification models based on amygdala radiomic features for Alzheimer's disease and mild cognitive impairment
复制标题
DOI：
10.1007/s11682-020-00434-z
复制
发表时间：
2021-02-04
影响因子：
3.2
通讯作者：
Ding, Zhongxiang
中科院分区：
文献类型：
--
作者：
Feng, Qi;Niu, Jialing;Ding, Zhongxiang
杏仁核是内侧颞叶的重要组成部分，在情绪和认知功能中发挥着关键作用。本研究的目的是建立并验证基于杏仁核放射组学特征的阿尔茨海默病（AD）和遗忘性轻度认知障碍（aMCI）的综合分类模型。对于杏仁核，从 97 名 AD 患者、53 名 aMCI 患者和 45 名正常对照 (NC) 的三维 T1 加权磁化准备快速梯度回波 (MPRAGE) 图像中提取了 3360 个放射组学特征。我们使用最大相关性和最小冗余（mRMR）以及最小绝对收缩和选择算子（LASSO）来选择特征。进行多变量logistic回归分析，建立三个分类模型（AD-NC组、AD-aMCI组和aMCI-NC组）。最后，评估了内部验证。经过两步特征选择后，AD-NC组保留了5个放射组学特征，AD-aMCI组和aMCI-NC组分别保留了16个特征。所提出的基于杏仁核放射学特征的逻辑分类分析，对于AD与NC分类，实现了0.90的准确度和0.93的ROC曲线下面积（AUC），对于AD与aMCI分类，实现了0.81的准确度和0.84的AUC，对于aMCI与NC分类，实现了0.75的准确度和0.80的AUC。杏仁核放射组学特征可能是检测 AD 和 aMCI 过程中脑组织微观结构变化的早期生物标志物。 Logistic 分类分析证明了 AD、aMCI 和 NC 组之间临床应用的有前景的分类性能。
The amygdala is an important part of the medial temporal lobe and plays a pivotal role in the emotional and cognitive function. The aim of this study was to build and validate comprehensive classification models based on amygdala radiomic features for Alzheimer's disease (AD) and amnestic mild cognitive impairment (aMCI). For the amygdala, 3360 radiomic features were extracted from 97 AD patients, 53 aMCI patients and 45 normal controls (NCs) on the three-dimensional T1-weighted magnetization-prepared rapid gradient echo (MPRAGE) images. We used maximum relevance and minimum redundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) to select the features. Multivariable logistic regression analysis was performed to build three classification models (AD-NC group, AD-aMCI group, and aMCI-NC group). Finally, internal validation was assessed. After two steps of feature selection, there were 5 radiomic features remained in the AD-NC group, 16 features remained in the AD-aMCI group and the aMCI-NC group, respectively. The proposed logistic classification analysis based on amygdala radiomic features achieves an accuracy of 0.90 and an area under the ROC curve (AUC) of 0.93 for AD vs. NC classification, an accuracy of 0.81 and an AUC of 0.84 for AD vs. aMCI classification, and an accuracy of 0.75 and an AUC of 0.80 for aMCI vs. NC classification. Amygdala radiomic features might be early biomarkers for detecting microstructural brain tissue changes during the AD and aMCI course. Logistic classification analysis demonstrated the promising classification performances for clinical applications among AD, aMCI and NC groups.