Detection of mild cognitive Impairment from gait using Adaptive Neuro-Fuzzy Inference system

• Published in: Biomedical signal processing and control Vol. 71; p. 103195
• Main Authors: Seifallahi, Mahmoud, Hassani Mehraban, Afsoon, Khamseh, Fatemeh, Salehi, Masoumeh, Hajati, Gholamreza, Khansari, Mehrdad, Ahmadzadeh, Mina, Babaie, Mehrzad
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2022
• Subjects: Adaptive Neuro-Fuzzy Inference System (ANFIS); Gait; Kinect Camera; Mild Cognitive Impairment; Skeletal Data; Gait; Mild Cognitive Impairment; Kinect Camera; Adaptive Neuro-Fuzzy Inference System (ANFIS); Skeletal Data
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2021.103195

•A novel method is proposed for the detection of elders with MCI form gait using ANFIS classifier and descriptive statistical analyses.•Kinect-V.2 camera is used as a new tool instead of usual previous systems like GaitRite for recording subjects’ walking in various conditions.•Dual-task gait parameters (DTW) are the optimal collection of gait features for the detection of MCI using the proposed ANFIS.•The proposed methods classify the elderly subjects with an accuracy of 93% and an F-score of 91.5% using DTW into MCI and HC groups. Early detection of mild cognitive impairment (MCI) among elders is very important for prevention of fast progress to probable Alzheimer’s disease. Studies show that motor abilities e.g. walking can be impaired for elderly people with mild to severe dementia and elders with MCI as well as memory defects. Thus, in this study, a novel method is presented to detect MCI among elders from skeletal data of elders’ gait recorded with a Kinect.V2 camera and using a proposed Adaptive Neuro-Fuzzy Inference System(ANFIS). The study included 70 elders with MCI and 80 healthy control (HC) elderly subjects who were selected based on specific criteria. Three sets of gait parameters as single-task walking (STW), dual-task walking (DTW), and cost of dual-task walking (DTC) were extracted from recorded gait tests. After comparing the extracted gait parameters using descriptive statistical methods, classification of elderly subjects into MCI and HC was done via the proposed ANFIS (Adaptive Neuro-Fuzzy Inference System) classifier using various extracted gait features. The results showed that cognitive tasks can dramatically increase the capability of gait analysis for the detection of MCI. Also, the comparison of the proposed ANFIS with several non-fuzzy classifiers showed that the ANFIS classifier can more accurately find the ambiguous boundary between MCI and HC in comparison to non-crisp methods. Generally, the proposed ANFIS classifier had values ​​of 93%, 90%, and 91.5% for accuracy, sensitivity, and F-score, respectively when all extracted gait parameters in the DTW condition for the classification of elderly subjects to MCI and HC groups.