Trustroke, an European artificial intelligence project aimed to optimise stroke treatment
1.5M Strokes by 2025
460K Deaths/year
25% ends up disabled
Only 40% Adherence
Trustroke, an European artificial intelligence project aimed to optimise stroke treatment
1.5M Strokes by 2025
460K Deaths/year
25% ends up disabled
Only 40% Adherence
Assists in managing stroke patients and assessing disease progression using clinical data, focusing on:
Clinical worsening leading to unplanned hospital readmissions.
Poor mobility, incomplete recovery and unfavorable clinical long-term outcomes.
Stroke recurrence.
Assists in managing stroke patients
and assessing disease progression using
clinical data, focusing on:
Clinical worsening leading to unplanned hospital readmissions.
Poor mobility, incomplete recovery and unfavorable clinical long-term outcomes.
Stroke recurrence.
Main Objectives
Federated Learning network to enable secure AI training
Trustworthy and validated AI solution for stroke risk assessment
Framework for patient empowerment and communication
A flexible and scalable platform clinically validated
Proof of concept study to improve stroke patient pathway
Federated Learning network to enable secure AI training
Trustworthy and validated AI solution for stroke risk assessment
Framework for patient empowerment and communication
A flexible and scalable platform clinically validated
Proof of concept study to improve stroke patient pathway
What is a stroke?
A stroke happens when the blood supply to a part of the brain is cut off. This can be caused by a blockage in a blood vessel in the brain (an ischaemic stroke) or a bleed from a blood vessel in the brain (a haemorrhagic stroke).
The faster blood flow is restored to the brain, the better chance someone has of making a good recovery.
State of the art
NIHSS* score is determined on admission and at discharge. Has a predictive value related to functional outcome at 90 days.
Early supported discharge is not common clinical practice and there is a lack of clear
criteria to select suitable patients.
Modified Rankin Score (mRS) is a disability scale are mainly based on clinical reported
outcomes.
Existance of some acceptable models for short-term risk prediction but not widely used
due to the amount of inputs required.
Unmet needs
Lack of high-quality, harmonised
and transparent datasets.
Lack of trust in predictive tools: Current predictive models are not validated or clinically implemeted due to lack of robustness.
Lack of effective human-centred solutions.
The solution
Federated artificial intelligence model network trained by european hospitals to be used in the platform.
Stroke management platform deployed in the stroke units from hospitals at Spain, Italy, Belgium and Slovenia and validated over 2100 outpatients and their caregivers.
Evidence-based recommentadations for the use of trustworthy risk predictors to gain adherence to treatment.
Robust AI-based scoring system for five crucial end points in the management of stroke, with associated
The impact
Improvements in prevention
and case management
Reduce by 10-20% the risk of stroke recurrence, mortality and hospital readmissions
Novel quantitative metrics in clinical
practise, after regulatory approval
Revised international clinical guidelines
Treatment adherence increasement up to 80%
In 2030, over 1 milion acute ischemic stroke patients and 9 milions stroke survivors could benefit from the Trustroke tool.
An ischemic stroke happens when the blood supply to a part of the brain is cut off. This is caused by the blockage of blood flow in a vessel that supplies brain.
The faster blood flow is restored to the brain, the better chance someone has of making a good recovery.
NIHSS* score is determined on admission and at discharge. It has a predictive value related to functional outcome at 90 days.
Early supported discharge is not common clinical practice and there is a lack of clear
criteria to select suitable patients.
Modified Rankin Score (mRS) is a disability scale mainly based on clinical reported
outcomes.
Existance of some acceptable models for short-term risk prediction but not widely used
due to the amount of inputs required.
Lack of high-quality, harmonised
and transparent datasets.
Lack of trustworthiness of predictive models: Current predictive models are not validated or clinically implemeted due to lack of robustness.
Lack of effective human-centred solutions.
Federated artificial intelligence model network trained by European hospitals to be used in the platform.
Stroke management platform implemented in the stroke units from hospitals at Spain, Italy, Belgium and Slovenia and validated over 2100 outpatients and their caregivers.
Evidence-based recommentadations for the use of trustworthy risk predictors to gain adherence to treatment.
Robust AI-based scoring system for five crucial end points in the management of stroke, with associated
Improvement in prevention
and case management
Reducing by 10-20% the risk of stroke recurrence, mortality and hospital readmissions
Novel quantitative metrics in clinical
practice, after regulatory approval
Revised international clinical guidelines
Treatment adherence increase up to 80%
In 2030, over 1 milion acute ischemic stroke patients and 9 milions stroke survivors could benefit from the Trustroke tool.
Publications
Publications
Project Resources and Publications
Accelerating stroke prevention
Partners
Partners
Project Coordination Team
Carlos Molina
Clinical Lead Head of Stroke Unit
Hospital Universitari Vall d'Hebron / VHIR
carlosav.molina@vallhebron.cat
Estela Sanjuan
Horizon Europe Project Coordinator
Hospital Universitari Vall d'Hebron / VHIR
estela.sanjuan@vhir.org
Project Coordination Team
Carlos Molina
Clinical Lead
Head of Stroke Unit
Hospital Universitari Vall d'Hebron / VHIR
carlosav.molina@vallhebron.cat
