The study aims to provide initial proof-of-concept validation data of an artificial intelligence-based model to estimate individual Parkinson's disease risk using demographic, clinical, genetic information and digital biomarker data collected via a smartwatch and a mobile application.
Background: Everyday electronic devices may detect subtle motor and non-motor abnormalities years before the clinical diagnosis of Parkinson's disease (PD) providing opportunities for early detection. Study aim and impact: This study aims to validate an artificial intelligence based model that provides an individualised risk of PD based on demographic, clinical, genetic and digital biomarker data (smartwatch and a phone app). An early diagnosis will allow timely interventions to manage symptoms and risk stratification of participants for early clinical trials. Methods: 60 people at risk of PD (either with polysomnography confirmed REM sleep behaviour disorder; OR neurogenic orthostatic hypotension; OR objective hyposmia on smell test) will be recruited. Participants will complete study assessments to provide PD risk estimation using current research clinical criteria and the artificial intelligence model. Study assessments will include: * In-person visits (baseline and 6 months) to complete validated questionnaires and a neurological examination (including cognitive and motor assessments). * Brain dopamine (DAT) scan (baseline only). * blood tests for PD polygenic risk score (baseline only) and plasma urate (in males only at baseline and 6 months). * Smartwatch and phone app: a smartwatch linked to the participants' smartphone will provide digital biomarker and additional clinical information through questionnaires via study phone app. An artificial intelligence based model (AI-PROGNOSIS model) will use these digital data in combination with demographics, clinical and genetic information to provide an individualised PD risk estimation. Accuracy measures of the risk estimates from the current research diagnostic criteria and artificial intelligence model using the presence of abnormal dopamine DAT scan as the ground truth for PD diagnosis will be provided.
Study Type
OBSERVATIONAL
Enrollment
60
Wearing a smartwatch and using a mobile phone application for 6 months in order to provide digital biomarker data and additional self reported clinical information.
Centre Hospitalier Universitaire de Toulouse
Toulouse, France
Fundación Iniciativa para las Neurociencias. Hospital Ruber Internacional.
Madrid, Spain
Queen Mary University of London
London, United Kingdom
Classification performance of the PD risk artificial intelligence-based model
Classification performance of the model in predicting dopaminergic degeneration defined as a binary outcome: a participant will be considered to have dopaminergic degeneration if putamen specific binding ratio (SBR) on the most affected side is below 2 standard deviations of age-matched normative data or shows abnormal visual inspection by a qualified nuclear medicine specialist on dopamine transporter SPECT imaging.
Time frame: From enrolment to 6 months
Usability of study digital environment (mAI-Health phone app)
System Usability Scale (SUS) scores. The SUS includes 10 statement items regarding the usability of the study phone application that will be rated on a scale of 1 - 5 (strongly disagree - strongly agree). Range 10-50 with higher scores meaning a better outcome.
Time frame: At 6 month visit
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.