Alzheimer's disease (AD) is a common and serious form of elderly dementia, but early detection and treatment of mild cognitive impairment can help slow down the progression of dementia. Recent studies have shown that there is a relationship between overall cognitive function and motor function and gait abnormalities. We recruited 302 cases from the Rehabilitation Hospital Affiliated to National Rehabilitation Aids Research Center and included 193 of them according to the screening criteria, including 137 patients with MCI and 56 healthy controls (HC). The gait parameters of the participants were collected during performing single-task (free walking) and dual-task (counting backwards from 100) using a wearable device. By taking gait parameters such as gait cycle, kinematics parameters, time-space parameters as the focus of the study, using recursive feature elimination (RFE) to select important features, and taking the subject's MoCA score as the response variable, a machine learning model based on quantitative evaluation of cognitive level of gait features was established. The results showed that temporal and spatial parameters of toe-off and heel strike had important clinical significance as markers to evaluate cognitive level, indicating important clinical application value in preventing or delaying the occurrence of AD in the future.
阿尔兹海默症（AD）是一种常见且危害严重的老年痴呆病，但对其早期轻度认知障碍的检测与治疗有助于减缓痴呆症的进展。近年来有研究表明认知功能与运动功能和步态异常之间存在关系。本研究从国家康复辅具研究中心附属康复医院招募了302例受试者，按照纳入与排除标准最终纳入193例受试者，其中137例为轻度认知障碍患者（MCI），56例为健康对照者（HC）。使用可穿戴设备采集参与者在单任务（自由行走）、双任务（倒数100）时的步态参数。将步态周期、运动学参数、时间-空间参数等步态参数作为研究重点，使用递归特征消除法（RFE）选择重要特征，将受试者的MoCA分数作为响应变量，建立了一种基于步态特征量化评估认知水平的机器学习模型。研究结果显示，足趾离地角度和足跟着地角度这两种时间-空间参数作为评估认知水平的标志物具有重要临床意义，未来或对预防或延缓AD的发生具有重要的临床应用价值。.
Keywords: Cognitive dysfunction; Gait analysis; Machine learning; Mild cognitive impairment; Random forest.