Repository files navigation

Abstract
Parkinson’s Disease (PD) is triggered due to the loss of dopaminergic neurons in thesubstantia nigra, disrupting the neural communication of the central nervous systemtowards reception and response of motor and cognitive senses of the patient. PD is aprogressive neural disorder which worsens with ageing. With no clearly outlined patternposed in symptoms, it is challenging for medical practitioners to identify the disease in itsprodromal stage. Inspired from the cause, this research’s objective is to predict the rate ofprogression based on the baseline assessment of a patient so that an appropriate treatmentplan can be designed for that individual patient. Biomarkers responsible for baselineassessment are extracted from multiple pre-clinical assessments designed to capture andscale the motor and cognitive impairments experienced by PD patients during theprodromal stage. The study performs clustering of PD patients into 3 clusters marking therate of progression based on the captured clinical feature of the patients and performancecomparison of 7 different ensembled and neural network-based classification model isconducted in this study. The study aims to assist medical practitioners in early diagnosisof risk PD among patients and adopt an appropriate measure to improve patient’s qualityof life.

Research Question
How well can combined analysis of clinical biomarkers help identify risk of Parkinson’s Disease progression at prodromal stage?

Source of Data
Parkinson's Progression Makers Initiative (PPMI) (http://www.ppmi-info.org), Michael J. Fox Foundation for Parkinson’s Research.Data is fetched from PPMI data repository via dedicated pypmi API package (https://pypmi.readthedocs.io/en/latest/index.html).

Undrestanding of data
Dataset offers a complete evaluation of patient clinical condition from baseline to5 years follow up visits. Captured data holds the clinical test score of various clinicalassessment trials conducted on the patient to identify different cognitive and motor developedimpairments. Datastore also maintains patient’s genetic data and brain MRI-Scan image datafor purpose of research. However, in this research clinical assessments, data will be used as anearly diagnostic biomarker of PD. Data is collected following standard data acquisitionprotocols with consent from the patient regarding the use of data for the purpose of researchwork. PPMI holds data for various clinical assessments for over 1800+ patients, belonging totwo prominent categories, suffering from PD and Healthy control (HC).15 clinical assessments capturethe premature motor and non-motor symptoms experienced by PD patients to scale the severity of the disease.In this research study important covariates are extracted from these set of assessments.

PPMI monitors symptomatic progression in PD patients at regular intervals. Participation ofthe patient at each interval is marked by a unique visit id. During every follow-up visit, patientis re-accessed for all baseline clinical assessments to capture the change in posed symptomsand understand the scenario of progression for that individual patient. These scheduled visitsare marked as BL (Basel Line) (Prashanth and Dutta Roy, 2018), which is first-time evaluationof the patient, later to which each incremental visit is marked between V01 – V12.Participationof patients declines gradually moving towards the last visit. Also, not each assessment ismandatorily conducted on every scheduled visit. For the purpose of study, subgroup of 476patients is created who showed active participation till last visit (V12), eliminating rest of thepatients who quitted the programme halfway. For the purpose of this study, we focus only on Healthy Control(HC) and Parkinson’s Disease (PD) patients. Also, data available for rest of the category is notsufficient for analytical study.