Methodological Research On Early Screening And Detection Of Neurological Diseases

At this stage,the problem of aging population is increasing.The number of people suffering from neurological diseases is increasing year by year,which brings pressure to both patients and doctors.Neurological diseases generally have many symptoms such as disturbance of consciousness,sensory disturbances,and movement disorders.Current methods for detecting neurological diseases generally perform invasive examinations on patients.Patients also need to go to the hospital to check face to face which increase the burden of inconvenient moving patients.With the development of information technology,Internet health care has gradually emerged,and the medical ecology is being changed little by little.More and more people are exploring new technologies,adopting more convenient and more efficient methods to reaching disease inquiry,disease tracking and rehabilitation training.This article uses big data technology and uses Parkinson's disease as an example to explore the diagnosis and severity judgment of Parkinson's disease,aiming to provide a more convenient and efficient detection method.The study of Parkinson's diagnosis was conducted through two methods: speech recognition and keypoint recognition.Firstly,the preprocessing and feature extraction of speech and face video data are converted into data information that can be used for model construction;the speech part uses the linear and nonlinear features currently used to detect Parkinson's disease,and the key points of the face analysis uses two types of features that embody the facial expression amplitude and tremor.The diagnosis problem is abstracted as a classification problem,the severity of the disease is abstracted as a regression problem,and than the model is established and diagnosed separately.When constructing models of the classification through voice,the paper builds a model for each category after clustering the patients and conducts personalized disease diagnosis.In the case of a large number of data features,a regular method is used for feature selection;the key points of the face time series data were constructed and a Long Short-term memory model was constructed for diagnosis.The data in this article,the speech part comes from the public data set of the network,and the key points of the face are collected from the hospital.In addition,the above methods and models are also used in our system for users to use.Through experimental verification,it was found that Parkinson's patients can be distinguished from healthy people through voice and face video data.Afterwards,the system functions and architecture are briefly introduced,and some of the above mentioned algorithms are integrated into it to achieve the application value.