IOT based monitoring and prediction model for parkinson s disease

Abstract

iv newlineABSTRACT newlineApproximately 10 million people throughout the world are suffering with Parkinson s disease. newline3.2 million people are living life with disability. There are three major causes of PD: - genetic, newlineidiic and drug induced PD. Most of PD patients are suffering with idiopathic Parkinson s newlinedisease in which cause of occurrence is unknown. Normally used treatment in early stages is newlinemedication which can improve the condition before going in severe conditions. As in severe newlinestage, test to detect PD severity are costly. So, its early detection becomes necessary. To solve newlinethis issue, a PD monitoring and progression kit is designed, which is non-invasive, cost and newlinetime effective kit for elderly. This kit recorded the gait signal of PD patient and on behalf of newlinevariation in gait, PD is detected. Various instrumented shoes have been designed by number newlineof researchers but internet of things (IOT) module in it still needs more research as it can newlineestablish a connection between patient and doctor by reducing the distance problem. Ultimately newlineit reduces the time of people living in remote areas. FTP wing server is used to store the data newlineof patient from where it is accessible by doctor, diagnosis center, caretakers and patients newlinethemselves also using required credentials. The kit can be easily run by semi-skilled health newlineworkers, so by implanting in remote areas early detection is possible. The early detection and newlinerehabilitation of PD can improve the treatment consistency of patients which may allow a fast newlinedecision of diagnosis. newlineThe whole work is divided into two parts: - one is prototype design and second is software newlinesection where online publically available gait signal datasets from website Physionet is taken newlineand analyzed. For it, different machine learning, deep learning techniques are applied. newlinePhysics-based algorithms contain limited applications for feature selection (FS) problems due newlineto dimensionality and high computational cost. Due to the lack of required information about newlineParkinson s disease (PDs), feature selection beco