Data Analysis Of Wearable Devices Of Alzheimer’s Patients Based On Functional Data Clustering

Alzheimer’s disease(AD)is an incurable neurodegenerative disease and is the most common type of dementia.In clinical study,the AD patients suffer from progressive cognitive impairment and movement impairment,which have great effects on people’s daily life and health,and brings a heavy economic burden to society and families.On the other hand,with the rapid development of data acquisition technology and data storage capabilities,large-scale medical data has gradually emerged.In clinical medicine,the analysis of medical big data with complex structure can aid diagnosis of diseases and treatment plans,which is a very important and meaningful research topic.This paper mainly focuses on the functional data clustering analysis of wearable device data from patients with AD,where the data are collected using AX6 devices attached on the waists of the patients with waist bands.The AX6 consists of a single three-axis accelerometer and three-axis gyroscope sensor.In this paper,the AX6 is mainly used to collect acceleration signals from the horizontal/sideway movement,upward/downward movement and forward/backward movement during the walking process,and the acceleration is recorded at a frequency of 100 Hz.Due to the high dimensionality,complex structure and noise pollution of the collected wearable device data,it is a great challenge to use the wearable device data to evaluate the gait activity of AD patients.In this paper,signal vector magnitude(SVM)and sliding window technique are used to preprocess wearable device data.First of all,the sliding window technique is used to divide the data into sliding windows and then the data from 55 AD patients and 18 patients with Parkinson’s disease are fitted using B-spline basis expansion.Then,the fun HDDC algorithm is used to project the fitted functional data into a low-dimensional subspace for clustering.This method simultaneously completes dimensionality reduction based on functional principal component analysis(FPCA)and clustering based on FPCA scores,and the model parameters are estimated by the EM algorithm,and the clustering results of the wearable device data of patients with AD are obtained,and compared with the clustering results of patients with Parkinson’s disease.Finally,a new feature is constructed for the clustering results of 73 patients,and a logistic regression model is established by combining the new features and the personal information of the patients including age,height and weight.This paper mainly studies the diagnosis of AD patients,so the main concern is the accuracy of the prediction model.In this paper,the accuracy rate of the prediction model obtained by leave-one-out crossvalidation is 94.5%,which shows that the results of this study will provide a theoretical basis for the diagnosis of AD patients in clinical medicine.It also provides theoretical support for the early diagnosis of AD,which is of great significance to the diagnosis and treatment of dementia in our country.