Our research
Leveraging multi-source datasets, including in-depth health and genetic data, at AI-PROGNOSIS, we will conduct research to develop and clinically validate novel, trustworthy AI models for personalised Parkinson's disease (PD) risk assessment and prognosis and combine them with digital biomarkers of risk and progression determinants from everyday devices.
Step 1 / Data
In AI-PROGNOSIS, we will amass and harmonise multi-source sets of in-depth health and genetic data labelled with incident PD diagnoses, disease progression history, or patients' response to medication.
Demographics,
clinical scales,
lifestyle data
Electrophysiological measurements and imaging data
Data from smartphones / wearables
SNPs and polygenetic risk score
Step 2 / Predictive modelling
We will then analyse the harmonised datasets to develop robust, explainable, and overall trustworthy AI models predicting PD risk, progression, and medication response.
PD risk prediction
Based on a person’s relevant characteristics and early signs, measured also with everyday digital devices.
PD progression prediction
In terms of time to higher disability transition, with emphasis on motor impairment and known disease milestones.
Medication response prediction
In terms of reduction of symptoms and of side effects, for a shortlist of common dopaminergic treatments.
Step 3 / Digital biomarkers
Digital biomarkers are health-related data collected by means of digital devices, such as smartphones and wearables. In AI-PROGNOSIS, we will develop new digital biomarkers of PD symptoms and along with existing ones, we will use them to inform the predictions of our predictive models.
Novel
(developed in project)
REM Behaviour Disorder (RBD)
Daytime somnolence
Existing
(validated or with PoC)
Rest tremor
Dyskinesia
Bradykinesia
Rigidity
Balance / posture
Cognitive deficits
Physical activity
Data comes first
Recent groundbreaking research efforts in the area of PD and beyond have created several relevant databases and biobanks with in-depth health and genetic data. We aim to leverage these data sources towards the development of our predictive models of PD risk, progression, and medication response.