Department of Biomedical Engineering, University of Arizona, Tucson, AZ, USA; Department of Kinesiology, University of Maryland, College Park, MD, USA.
Philips Research North America, Cambridge, MA, USA.
Comput Biol Med. 2020 May;120:103705. doi: 10.1016/j.compbiomed.2020.103705. Epub 2020 Mar 19.
In this study, we examined the uncertainty and local instability of motor function for cognitive impairment screening using a previously validated upper-extremity function (UEF). This approach was established based upon the fact that elders with an impaired executive function have trouble in the simultaneous execution of a motor and a cognitive task (dual-tasking). Older adults aged 65 years and older were recruited and stratified into 1) cognitive normal (CN), 2) amnestic MCI of the Alzheimer's type (aMCI), and 3) early-stage Alzheimer's Disease (AD). Participants performed normal-paced repetitive elbow flexion without counting and while counting backward by ones and threes. The influence of cognitive task on motor function was measured using uncertainty (measured by Shannon entropy), and local instability (measured by the largest Lyapunov exponent) of elbow flexion and compared between cognitive groups using ANOVAs, while adjusting for age, sex, and BMI. We developed logistic ordinal regression models for predicting cognitive groups based on these nonlinear measures. A total of 81 participants were recruited, including 35 CN (age = 83.8 ± 6.9), 30 aMCI (age = 83.9 ± 6.9), and 16 early AD (age = 83.2 ± 6.6). Uncertainty of motor function demonstrated the strongest associations with cognitive impairment, with an effect size of 0.52, 0.88, and 0.51 for CN vs. aMCI, CN vs. AD, and aMCI vs. AD comparisons, respectively. Ordinal logistic models predicted cognitive impairment (aMCI and AD combined) with a sensitivity and specificity of 0.82. The findings accentuate the potential of employing nonlinear dynamical features of motor functions during dual-tasking, especially uncertainty, in detecting cognitive impairment.
在这项研究中，我们使用先前验证过的上肢功能 (UEF) 检查了认知障碍筛查中运动功能的不确定性和局部不稳定性。这种方法的建立基于这样一个事实，即执行功能受损的老年人在同时执行运动和认知任务（双重任务）时会遇到困难。招募了年龄在 65 岁及以上的老年人，并将其分为 1）认知正常 (CN)，2）遗忘型轻度认知障碍（aMCI），和 3）早期阿尔茨海默病 (AD)。参与者在不计数的情况下以正常的节奏进行反复的肘部弯曲，同时倒计数一和三。使用肘部弯曲的不确定性（通过香农熵测量）和局部不稳定性（通过最大李雅普诺夫指数测量）来测量认知任务对运动功能的影响，并通过方差分析比较认知组之间的差异，同时调整年龄、性别和 BMI。我们根据这些非线性测量值开发了基于逻辑有序回归模型的预测认知组的模型。共招募了 81 名参与者，包括 35 名 CN（年龄=83.8±6.9），30 名 aMCI（年龄=83.9±6.9）和 16 名早期 AD（年龄=83.2±6.6）。运动功能的不确定性与认知障碍的相关性最强，CN 与 aMCI、CN 与 AD 和 aMCI 与 AD 相比的效应大小分别为 0.52、0.88 和 0.51。有序逻辑模型预测认知障碍（aMCI 和 AD 合并）的敏感性和特异性为 0.82。这些发现强调了在双重任务期间（特别是在不确定性方面），采用运动功能的非线性动力学特征来检测认知障碍的潜力。
