We are training the 15 innovation leaders in the vision of a “Personalised In-Silico Cardiology” through a unique 3 year PhD programme coordinated among 10 academic, industrial and clinical institutions, and 9 more associated partners.
Follow us on Twitter: https://twitter.com/PICnetEU
This project has received funding from the European Union’s Horizon 2020 Marie Skłodowska-Curie ITN Project under grant agreement No 764738.
Background and rationale
Cardiovascular diseases (CVD) have a huge impact on society in terms of mortality, morbidity and healthcare costs, being responsible for 1.9 million deaths in the EU annually (42% of all deaths) with a total cost of €169 billion. Improving healthcare systems in Europe in a period of ageing population and tightening financial constraints mandates a shift towards personalised and preventive management of disease. We need tailored and earlier treatments to increase the efficacy and efficiency of the healthcare system, as well as the quality of life of patients.
Healthcare provision can conceptually be simplified into three main processes: acquisition of clinical data, diagnosis & therapy planning, and delivery of treatment & intervention. Current technology allows a rich data acquisition, the use of sophisticated devices to monitor patients and deliver care. However clinical practice is guided by the use of averaged (population-based) metrics to define therapy strategies, missing many of the opportunities for disease prevention and tailoring of care for the individual patient.
In this context, recent scientific progress has created an exceptional capacity to simulate in-silico (i.e. on a computer) the heart and its interaction with the circulatory system. Patient-specific in-silico models provide a structured, reproducible and predictive framework for interpreting and integrating clinical data. This provides the pathway for developing personalised and preventive management strategies for cardiovascular diseases. In addition, recent advances in data science (i.e. machine learning, data mining) enable the extraction of novel insights and knowledge from the large repositories of clinical data of our health information systems.
PIC is the European ITN that will train a cohort of 15 future innovation leaders able to articulate and materialise the vision of Personalised In-silico Cardiology (see Fig. 1) where healthcare is guided by in-silico models. These models become virtual reconstructions of an individual, or avatars, to evaluate current health status and therapy options. PIC fellows will build both mechanistic and statistical models from clinical data (WP1), enabling the extraction of biomarkers for better diagnosis and prognosis of the individual patient. PIC fellows will apply models to maximise the value of clinical data (WP2) to inform diagnosis, and to optimise clinical devices & drug choices (WP3) to deliver a personalised therapy.
The 15 PIC fellows
The titles and host institutions of the 15 projects of the PIC fellows are:
- F1: Computational cardiac anatomy: novel shape biomarkers – King’s College of London
- F2: In-silico workbench for sensor acquisition and theraphy optimization – Maastricht University.
- F3: Cellular electrophysiology from electrical body recordings – Universidad de Zaragoza.
- F4: Machine learning of ultrasound data: robust motion and new risk scores – General Electric.
- F5: Cardiac accelerometers – Oslo Universitetssykehus.
- F6: Non-invasive estimation of central blood pressure and flow inefficiencies – King’s College of London.
- F7: Preclinical validation and assessment – Maastricht University.
- F8: Optimization of the choice and configuration of valve prosthesis – FEOPS.
- F9: Personalised HF theraphy with motion sensors and simulations – Medtronic.
- F10: Optimisation of activation patterns during next generation crt pacing – King’s College of London.
- F11: Personalised analysis of drug safety and efficacy – University of Oxford.
- F12: Improved control of cardiac pacemakers in heart failure – Oslo Universitetssykehus.
- F13: Cardiomyopathy under the lens of computer models – University of Oxford.
- F14: Use of non-invasive mapping to treat atrial arrhythmias – Université de Bordeaux.
- F15: Obstructive dilated cardiomyopathy and valve diseases – Hospital Clínic Barcelona.
F1: Computational cardiac anatomy: novel shape biomarkers
Host: King’s College of London (KCL)
Objectives: to propose novel clinical diagnostic metrics based on the concept of statistical models of anatomy. This project will benefit from the huge amount of imaging data stored in health information systems, and build statistical atlases of the cardiac ventricles and main vessels. The description of the range of variability in healthy population will then offer the opportunity to detect earlier deviations caused by disease (F12, F13, F14 & F15). The patient-specific and anatomically detailed geometrical models will also be used to run simulations to predict the impact of therapy in heart failure, arrhythmia or valve conditions (F8, F10 and F14).
Planned secondment(s):
- Industrial exposure to GEVU. For 2M in Y2, in order to investigate the full automation in the generation of atlases using machine learning technologies.
- Clinical exposure to HCB/AQuAS (co-located in Barcelona), for 2M in Y1, in order to get clinical data and to identify clinical problems that could benefit from the atlas technology developed.
F2: In-silico workbench for sensor acquisition and theraphy optimization
Host: Maastricht University (UM)
Objectives: To simulate phonocardiographic and motion sensor signals during synchronous and dyssynchronous heart failure in the well-known CircAdapt model of the heart and complete (pulmonary and systemic) circulation. The CircAdapt model allows to derive signals that mimic either sounds (from the closure/opening of valves) or that mimic motion sensors from known cardiac strains. Diseased situations, as well as interventions by means of CRT, will be simulated. Results will guide optimization of sensor configuration and therapy delivery (F9), and will thus serve as basis for biosensor development (F5) and for an efficient pre-clinical testing (F7) and clinical evaluation (F12).
Planned secondment(s):
- Academic/clinical: OUS and KCL. For 1M in Y1 at each institution, to relate measurements of deformation and motion sensors with clinical data and to exchange modelling technologies and insights.
- Industrial: MDT. For 2M in Y2, to develop the models of the specific characteristics of the products of this company.
F3: Cellular electrophysiology from electrical body recordings
Host: Universidad de Zaragoza (UZ)
Objectives: To personalise cell parameters of electro-physiological (EP) models from electrical activity observed from clinical data, including signals measured on the body surface (ECG). Inter- and intra-individual variability in cardiac electrical activity will be represented through identification of cell parameter values across the myocardium. The developed methods will rely on the access to a wide range of clinical / experimental data and our expertise on the formulation of stochastic EP models (UOXF, UB, UM, KCL). The cell parameters will have diagnostic and prognostic value, and will be used in simulations to predict individualised risks to drugs (F11) and the optimal treatment of heart failure (F10) or arrhythmias (F14).
Planned secondment(s):
- Clinical/academic: LYRIC, for 2M in Y1, in order to get anatomical data and work on non-invasive reconstruction of heart electrical activity from body surface electrocardiographic data.
- Academic: UOX, for 2M in Y2 to work on development and validation of EP models (with F11 & F13).
- Industrial: IBM, for 2M in Y2 to work on methods for parameter estimation / inference to personalise EP models.
F4: Machine learning of ultrasound data: robust motion and new risk scores
Host: General Electric (GEVU)
Objectives: to develop machine learning solutions for robust extraction of anatomical, pathological and pathophysiological features from cardiac ultrasound images. Risk scores will be constructed by the identification of hidden patterns in the already available information in the images and clinical data currently held in hospital information systems. Statistical models of motion will be integrated with the anatomical models from F1, flow insufficiency parameters from F6, ECG data from F3 and markers used in the clinical guidelines for the management of CVD (F12–F15). Gaining statistical power will require a large number of patients, with a range of outcomes, which is a strength of PIC clinical partners (OUS, HCB, JRH, KCL) and AQUAS.
Planned secondment(s):
- Clinical: OUS (Hospital). For 2M in Y2, to test the novel methods, and get feedback from clinical doctors.
- Academic & clinical: KCL. For 2M in Y1 to build 3D computational models of shape, motion, and flow, and to get access to the repository of electronic data held in St Thomas’ Hospital.
- Academic: UZ. For 1M in Y2 to learn about ECG analysis, and to integrate ECG data in risk scores.
F5: Cardiac accelerometers
Host: Oslo Universitetssykehus (OUS)
Objectives: To extract clinical information from kinematic measurements (accelerations, velocities, displacements) of the heart using novel miniaturized accelerometers (technology patented by CAR). The study will focus on model interpretation of data, extracting indices to quantify cardiac function, providing alarms when the heart becomes dysfunctional, and providing information of the functional state such as response to medications and fluid loading. It will work in close industrial collaboration (CAR, MDT) in order to design an end-product fulfilling clinical regulations, and for the analysis and interpretation of data using machine learning (F4). The project will combine data recorded in previous studies as well as simulated data (F2) when predicting cardiac kinematics during normal and pathological (F12) conditions.
Planned secondment(s):
- Industrial: CAR (SME, P5). Continuous meetings in order to further develop and miniaturize the sensor. And MDT for 2M in Y2, in order to learn constrains in implementing accelerometers in devices and how to embed algorithms, with specific computational constrictions on top of signal processing techniques.
- Academic: UZ, for 2M in Y1, in order to learn about signal processing techniques to extract the valid accelerometer data.
F6: Non-invasive estimation of central blood pressure and flow inefficiencies
Host: King’s College of London (KCL)
Objectives: To investigate the feasibility and performance of methods to estimate pressure differences from imaging data (MRI or ultrasound) removing the need of invasive catheters. Clinical diagnostic guidelines of conditions such as valve stenosis use risk markers based on the pressure drop through the flow obstruction. This project involves interpreting clinical images within the context of physical constraints encoded in mathematical models. Using a novel mathematical model patented at KCL, this project will investigate how to better characterise flow inefficiencies based on viscous dissipation or the sudden change of linear momentum. In order to achieve an easy clinical adoption, the project will also investigate the use of planar wave imaging or contrast agents in ultrasonography.
Planned secondment(s):
- Industrial: FEOPS, for 1M in Y2 to work on the industrial adoption of the methods in the problem of TAVI planning.
- Academic: UM, for 1M in Y2 to run experiments and verify the flow inefficiency characterization (with F7).
- Clinical: IDIBAPS, for 2M in Y1, to get clinical data and generate initial evidence comparing to in-silico workbenches.
F7: Preclinical validation and assessment
Host: Maastricht University.(UM)
Objectives: To propose and test independent hypotheses of the electro-mechanical cardiac physiology in animal models. This project will build and use models of synchronous and dyssynchronous heart failure and related therapies (CRT), and models of valve stenosis that introduce an additional burden to the heart. It will then specifically investigate on the mechanistic links between an external insult and the development of heart failure, with measurements including electrophysiological mapping, echocardiography, MRI and hemodynamics. Complementarily, this project will provide the access to the experimental models and data to perform the pre-clinical evaluation of the results from other fellows (F6, F10 and F12).
Planned secondment(s):
- Industrial: MDT. For 2M in Y2, in order to be exposed to the methodologies for the design and testing of pacemaker leads and clinical devices, and for the identification and characterization of dyssynchrony.
- Academic/clinical exposure at OUS. For 2M in Y3 in order to connect measurements of deformation and motion sensors from the animal experiments and exchange technologies and insights.
F8: Optimization of the choice and configuration of valve prosthesis
Host: FEOPS
Objectives: To evaluate the prediction of the optimal choice and configuration of the valve prosthesis used in transcatheter valve implantation (eg Transcatheter Aortic Valve Implantation – TAVI, Transcatheter Mitral Valve Repair/Replacement – TMVR). Advanced finite element analyses and computation fluid simulations will be used to test the outcome different implantation strategies (a library of models of the different prosthesis will be built for this purpose), guided by anatomical models (F1) and data-driven boundary conditions (F6). The proposed optimal configuration will be compared to the actual clinical outcomes in coordination with F15.
Planned secondment(s):
- Clinical: IDIBAPS (HCB). Continuous short visits during the whole project in order to learn about the clinical workflow, and the decision making process, and work together with F1, F6 (also in secondment at IDIBAPS) and F15.
- Academic/clinical: KCL. During year 2, a series of short visits to coordinate data acquisition and the validation study, and to identify synergies with the modelling work at KCL.
F9: Personalised HF theraphy with motion sensors and simulations
Host: Medtronic (MDT)
Objectives: (1) To design an implanted device able to pace and simultaneously provide motion tracking information. This project will investigate the use of electro-magnetic tracking systems, and will work together with F5 (working with accelerometers) towards an optimal design. (2) To optimize the location and configuration of the new sensors in the heart using in-silico models, in close collaboration with F2 (simplified cardiac and systemic models to define optimal sensing) and F10 (anatomically detailed models to define optimal pacing). The technology will be evaluated in coordination with F12.
Planned secondment(s):
- Clinical: OUS, for 2M in Y1, to run experiments to evaluate the two alternative tracking technologies.
- Academic: KCL, for 2M in Y2, to learn the in-silico technologies to optimise device settings.
F10: Optimisation of activation patterns during next generation CRT pacing
Host: King’s College of London (KCL)
Objectives: To build in-silico detailed personalised biophysical finite element models in order to predict the optimal configuration of pacing leads (location and timing) in the application of Cardiac Resynchronization Therapy (CRT) to heart failure. The goal is to deliver the best possible activation pattern to a failing heart. This project will combine mathematics, numerical methods, image processing and software development to use inverse ECG measurements (F3), tomographic imagery (F1) and computer simulations to predict patient activation patterns. It will then investigate (1) the impact of the rate of activation or of an early access to Purkinje network, and (2) the benefit of the adoption of multi lead, multi-pole and endocardial pacing devices. Evaluation of results will be coordinated with F12.
Planned secondment(s):
- Industrial: MDT, for 1M in Y1 to understand the process and practicalities of the Medtronic CRT system; and GEVU, for 1m in Y1 to automate the model generation through the adoption of their image segmentation expertise.
- Academic: UM, for 1M in Y2 to validate the model against patient and pre-clinical data.
- Academic & Clinical: UB, for 3M in Y2 to work on the introduction of models of Purkinje network into models, and to investigate their role if any in CRT activation patterns.
F11: Personalised analysis of drug safety and efficacy
Host: University of Oxford (UOXF)
Objectives: To investigate the major differences in the pro-arrhythmic and anti-arrhythmic mechanisms underlying human whole-ventricular EP response to drugs in non-diseased versus diseased conditions. This project will build personalised models of human electro-physiology (in collaboration with F3 and F1), and simulate the activation of the heart to predict the safety and efficacy of pharmacological action in the individual subject. Special emphasis will be given to avoid arrhythmogenic effects, which currently have a huge interest for regulatory bodies (see letter of commitment by FDA). Results will be evaluated in collaboration with F14.
Planned secondment(s):
- Industrial: JAN (P9, pharma industry). For 2M in Y2 to test the ideas in the company’s database of compounds.
- Regulatory experience provided by interactions with the FDA (P4).
- Clinical: co-supervision and interactions within cardiologists at the Oxford JRH (P7).
F12: Improved control of cardiac pacemakers in heart failure
Host: Oslo Universitetssykehus (OUS)
Objectives: To improve the control of implanted pacemakers for cardiac resynchronization therapy (CRT) in heart failure. CRT still has significant challenges when determining the best placement of the leads and in the optimization of the pacemaker parameters to maximize its benefit. This project will tackle this challenge by using miniaturized motion sensors incorporated in pacemaker leads (F9), and will measure the impact of the pacemaker on heart wall motion. The capacity of the method for evaluating cardiac function will be tested using a combination of in-silico (F2, F10), physical and animal (F7) models.
Planned secondment(s):
- Academic: KCL and UM. For 1M in Y1 at each location in order to perform testing in in-silico models, cardiac 3D finite element models at KCL, and simpler cardiac but comprehensive systemic models at UM (CircAdapt).
- Industrial: CAR (SME, P5), through regular contact with the providers of the accelerometer technology in order to propose and evaluate novel features in the sensor (in coordination with F5 & F9).
- Industrial: MDT. For 2M in Y2, to learn about Medtronic’s manufacturing processes & hardware constraints in the design of pacemakers.
F13: Cardiomyopathy under the lens of computer models
Host: University of Oxford (UOXF)
Objectives: To better understand and characterise one of the underlying causes of heart failure and arrhythmias, the disease of the cardiac cells (cardiomyopathy). This project will analyse how much structural and electrical remodelling affects the performance of the heart using anatomically detailed simulations. Specifically, it will use a combination of personalised models and machine learning techniques to study the correlation of this performance with disease geno- and phenotypes for hypertrophic and dilated cardiomyopathy. Access to a detailed phenotype will be enabled through the collaboration with F1 (anatomical) F4 (mechanical) and F3 (electrophysiological).
Planned secondment(s): (fellow hosted in an academic institution)
- Industrial: IBM (P6). For 2M in Y2, to learn about complementary modelling strategies using HPC.
- Clinical: co-supervision and interactions within cardiologists at the John Radcliffe Hospital (JRH, P11).
F14: Use of non-invasive mapping to treat atrial arrhythmias
Host: Université de Bordeaux (UB).
Objectives: To investigate how in-silico models can help to better plan ablation to reduce the procedure time, as well as minimize and predict recurrence. Atrial arrhythmia is a complex and multifactorial problem. The access to in-silico models is essential to isolate factors, and then analyse their isolated and complementary effects. The plan is to construct personalized atrial models based on clinical data (MRI, late gadolinium enhanced MRI, electroanatomic mapping, non-invasive mapping) and initiate arrhythmias. Model performance will be compared with clinical data to improve parameterization techniques. . Ablations sites will be planned in silico and tested first retrospectively on clinical data and then prospectively, sharing tools and experiences with F11. This project will also contribute to validate inverse mapping solutions (F3) by use of forward and inverse calculations to determine the information that can be characterised in a clinical context.
Planned secondment(s): (hosted in an academic & clinical instit.)
- Industrial: MDT, 2M in Y2 + 2M in Y3, to study (with in-silico models including anatomy and clinical data) how inter-patient variability affects the efficacy of AF therapy by ablation or electrical interventions.
- Clinical & academic: KCL, for 2M in Y2, to learn about shape analysis technology (with F1), to integrate investigation of atrial fibrillation, to collect further EP data and to prepare a multi-centre study.
F15: Obstructive dilated cardiomyopathy and valve diseases
Host: Hospital Clínic Barcelona (IDIBAPS)
Objectives: to investigate the clinical added value of the adoption of in-silico models in the management of two obstructive heart conditions: obstructive hypertrophic cardiomyopathy and aortic stenosis. By a deep understanding of physiopathology, individual haemodynamics, force generation and cardiac remodelling, we aim to obtain a personalized approach to the management of obstructive cardiac diseases. This will permit the selection of the optimal time of intervention, improve treatment selection and increase the patients’ prognosis and quality of life. Specifically, the burden to the heart caused by the obstruction will be measured by the novel biomarkers based on pressure gradients (F6), anatomical remodelling markers (F1) and in their optimal integration together with traditional metrics through machine learning techniques (F4). Besides, the optimal planning of the surgical procedures (F8) will be evaluated in a cohort of severe aortic stenosis subjects.
Planned secondment(s): (fellow hosted in clinical institution)
- Academic/clinical: KCL. 2M in Y1 to learn the analysis tools, image acquisition and design the validation strategy.
- Industrial: FEOPS. For 1M in Y2, to learn the use of the planning system, and to design its adoption & validation.