Atrial fibrillation (AF) is the most prevalent heart arrhythmia globally, resulting in severe complications, substantial financial costs, and significant resource use.[1 ] AF frequently goes unnoticed until the patient presents with AF-related complications (e.g., stroke, heart failure, dementia, and hospitalisations), particularly with brief episodes of AF that spontaneously revert to sinus rhythm.
In Europe, stroke (a major complication of AF) is a leading cause of death and the top cause of disability. The pathophysiology of AF-related stroke (AFRS) involves severe neurological deficits, which considerably worsens prognosis. While risk factors for poor stroke outcomes are known, current AF prediction models have limitations and fail to account for dynamic changes in risk profiles.[2 ]
[3 ] In addition, the importance of various stroke risk factors in AF may have changed over the years, for example, sex differences in AFRS risk.[4 ]
[5 ] This has had implications for stroke risk stratification, concerning the use of the well-validated CHA2 DS2 -VASc score or a nonsex version (CHA2 DS2 -VA).[6 ]
[7 ]
[8 ] Nevertheless, recognizing the residual cardiovascular risks associated with AF despite anticoagulation, the management of this condition has moved toward a more holistic or integrated care approach, which has been associated with better clinical outcomes.[9 ]
[10 ] This has led to such an approach recommended in contemporary guidelines.[11 ]
[12 ]
[13 ]
[14 ]
Stroke prevention is central to AF management.[15 ] Indeed, oral anticoagulant treatment in AFRS presents a dilemma: early initiation may increase haemorrhagic transformation risk, whereas delays can lead to recurrent ischaemic strokes. Poststroke rehabilitation, crucial for reducing risks and improving outcomes, lacks consensus on effective protocols, particularly personalised approaches based on functional outcomes in AFRS patients. This uncertainty can result in the exclusion of patients who would benefit from rehabilitation or the inefficient use of health care resources on those unlikely to help.
Given this background, the European Union, through the Horizon Europe Research Program, has funded the “Health Virtual Twins for the Personalised Management of Stroke Related to Atrial Fibrillation” (TARGET) Project (grant agreement no. 101136244). TARGET's consortium involves 19 partners including universities, hospitals, companies, and a charity. The project kicked started in January 2024 under the scientific leadership of Prof. Sandra Ortega-Martorell (Principal Investigator, Liverpool John Moores University, LJMU), the methodological leadership in artificial intelligence (AI) and virtual twins of Prof. Ivan Olier (LJMU), the clinical leadership in AF and stroke of Prof. Gregory Lip (University of Liverpool), and the coordination of Prof. Mattias Ohlsson (Lund University).
TARGET aims to address several clinical challenges within the AFRS disease pathway by focusing on a three-pillar approach: (Pillar I) Risk Prediction and dynamic, longitudinal monitoring of AF and the subsequent risk of developing AFRS; (Pillar II) Diagnosis and Management of AFRS, including early identification of stroke etiology, prediction of outcomes, and risk of stroke recurrence; and (Pillar III) Rehabilitation, focusing on identifying predictors of functional independence and quality of life in AFRS survivors and facilitating personalisation of rehabilitation. The project will be underpinned by the development of virtual twins of patients, which will be used to model novel, causal AI models embedded into decision-support tools for point-of-care applications. These novel models and tools will be evaluated via in silico simulated clinical trials and on newly collected data from clinical observational studies ([Fig. 1 ]).
Fig. 1 Summary of main activities of the EU project TARGET.
TARGET has a strong focus on the personalisation of health technologies for improved and more cost-efficient solutions in disease prevention, diagnosis, treatment and monitoring, better patient outcomes and well-being in the AFRS disease pathway, and reduced disease burden. TARGET will work closely with health care professionals (HCPs) and patients, who will be at the core of the research and the project, to codevelop the tools and ensure their acceptance and adoption.
For Pillar I, one of the tools (Tool 1) will embed TARGET models to provide HCPs with personalised risk prediction scores along with the causal factors and help patients understand how modifiable factors, e.g., lifestyle changes, could impact risk (e.g., increase or decrease) over time (dynamic, longitudinal monitoring of risk). TARGET will also build on Isansys' (partner) Patient Status Engine (PSE) to integrate and evaluate personalised risk prediction models when monitoring patients. The PSE is an end-to-end medical device (CE Class IIa) and a configurable platform that generates and analyses real-time physiological data. TARGET will adapt the PSE dashboard (Tool 2) embedding novel risk prediction and AF detection models, to dynamically estimate patients' clinical trajectories and monitor AF.
The tool for Pillar II (Tool 3) will provide HCPs with information about stroke etiology, personalised outcome prediction, and recurrence risk scores (including the dynamic changes in the risk) along with associated causal effects or factors and optimised recommendations for oral anticoagulation resumption in AFRS. For patients, it will be limited to personalised outcome prediction and recurrence risk scores, where considered that this information would benefit patients.
For Pillar III, one of the tools (Tool 4), will provide HCPs with personalised predictions of functional outcomes along with associated causal effects or factors, the individual rehabilitation needs of patients, and a dynamic and personalised assessment of independence level and health-related quality of life after rehabilitation. Patients will use this tool to learn their personalised prediction of functional outcomes, and how treatment adherence could impact their recovery trajectory over time (dynamic assessment). A second tool (Tool 5) will be a serious game, which will recommend personalised therapeutic sessions involving specific rehabilitation goals and motivation mechanisms; game genres suitable for stroke patients' rehabilitation, with visually appealing and intuitive environments; capturing patients' gestures and gait during gameplay to provide real-time feedback; and inclusion of different types of exercises/movements into the games that gradually increase difficulty and complexity as patients' rehabilitation progresses.
Before wider implementation into clinical practice, external validation of the novel virtual twins-based AI models and tools is required. To this aim, TARGET will perform the four prospective cohort studies ([Table 1 ]) in four countries ([Fig. 2 ]), where clinical and nonclinical information will be collected from the participants.
Table 1
Brief summary of TARGET's four clinical observational studies
CS1
Name
I
n
telligent m
o
nitoring
t
o pr
e
dict
a
trial
f
ibrillation (NOTE-AF)
Setting
Liverpool University Hospital NHS Foundation Trust (LUHFT), Liverpool, UK
Leader
Prof. Ingeborg Welters
Description
This study will validate Pillar I novel models and tools (Tools 1 and 2) for prediction of clinically relevant AF episodes in patients hospitalised for elective surgery or with acute illness. Continuous monitoring will be performed using the Isansys PSE
CS2
Name
Using vir
t
u
a
l tw
i
n-based AI mode
l
s to detect AF and impr
o
ve st
r
oke outcomes (TAILOR)
Setting
Stroke Unit at Hospital del Mar (IMIM). Barcelona, Spain.
Radboud (RUMC). Nijmegen, The Netherlands
Leaders
Dr Eva Giralt-Steinhauer (IMIM) and Dr Anil Tuladhar (RUMC)
Description
This multi-centric study will validate Pillar II novel models embedded into Tools 2 and 3, on patients with ischaemic stroke and no history of AF. The study aims to enhance AF detection using Tool 2, find novel image markers, identify AFRS etiology, and predict stroke outcomes (e.g., at 3 mo) using Tool 3
CS3
Name
Im
p
roving recov
e
ry:
a
new screening tool for selecting high-intensity and moderate-intensity AFRS
r
ehabi
l
itation programs (PEARL)
Setting
Rehabilitation Unit at IMIM. Barcelona, Spain
Leader
Dr Esther Duarte
Description
This study will determine differences in rehabilitation needs of acute poststroke patients and 1-, 3-, and 6-mo poststroke outcomes (i.e., body structure and function, activity, and participation outcomes). Outcomes will be assessed at baseline (stroke unit discharge) and will be compared with Tool 4 assessments
CS4
Name
Fo
llow-up
s
tudy of s
t
roke pati
e
nts using senso
r
s for the evaluation of functional ability and compensatory movements (FOSTER)
Setting
Revalidatieziekenhuis Inkendaal (RI). St-Pieters-Leeuw, Belgium
Leader
Prof. Degelaen
Description
This study will follow up stroke patients admitted to RI after hospital discharge, to assess the use of sensor measurements obtained by Tool 5 to quantify patient recovery, which will be compared with standardised clinical assessments. We will also assess whether sensors can help detect compensatory strategies during daily life-based activities
Fig. 2 TARGET will conduct four prospective observational clinical studies (CS) in four countries. Purple dots represent the nineteen partners of the TARGET Consortium.
In silico clinical trials will be employed to accelerate the translation of the virtual twin models developed. They will be developed based on TARGET's clinical observational studies to allow for further evidence generation on the effect of these models and their impact on translational outcomes covering the AFRS disease pathway (Pillars I–III). For this, virtual populations of patients will be derived from data collected in the observational clinical studies, as well as from heldout data used for the development of the virtual twins. The in silico clinical trial simulations will then test candidate virtual twin-based AI models in scenarios such as: (1) predicting onset of AF episodes (CS1), (2) predicting stroke outcomes (CS2), and (3) improving selection of therapy intensity (CS3). Whereas the real-world studies are observational, the in silico trials will be interventional to determine the impact of the tools on clinical decision-making relevant to the clinical studies and whether improved outcomes are observed.
TARGET will generate a high-societal, scientific, technological, and economic impact and foster translational biomedical research into practice by increasing and accelerating our understanding of the drivers of AF and AFRS, enhancing the knowledge of the disease onset and progression, and developing better tools for improved care management and treatment of AFRS patients.
