Background
Cardiovascular Disease (CVD) accounts for 45% of all deaths in Europe and 37% of all deaths in the EU. 80% of premature heart disease and stroke are preventable, but this hinges on early recognition of risks and prognostics: accurate diagnostics helps to allocate the right intervention to the right people. General modelling approaches for risk factors exist. However, 1) they have been developed with datasets that do not reflect true population properties, especially for gender, this leads to a bias making results less valid for females, 2) they do not take well enough into account differences in risks for populations with different genetic backgrounds, and 3) they do not use all information that is available in raw and combined data, including daily-life data. The current approach in CVD diagnostics via ECG causes a high number of false-positive results, especially among females, causing unnecessary reference to costly examinations and anxiety. With advances in wearable technology and data analytics, acceptable medical quality ECG in real-life is rapidly becoming an affordable and accessible rich source of information. Data is available, and major hurdles have been overcome in the fields of AI and ML over the past few years.
Goal
PerCard explores the value of combining integrated heterogeneous data sources with AI/ML to increase the validity of risk assessment for cardiovascular disease in different populations. The consortium develops specialised models for populations as well as explicitly mitigates the issue of gender-bias in existing risk assessment methods.
PerCard develops and uses new data analysis methods (AI, ML, signal processing, including ensemble classifiers and Bayesian neural networks) to develop more powerful risk assessment methods that are accurate, robust, and explainable.
Impact
Methods for risk assessment, and diagnostics have been developed, however, there is a need to improve accuracy as well as applicability. Tools should take better into account actual population properties, especially regarding gender and different genetic backgrounds. Also, they should use all information that is available – this information is increasing and evolving thanks to advances in wearable technologies, and therefore AI/ML, and risk models should capitalise on this with new analysis approaches.
PerCard has the following impacts. Cost containment for the European health and social care systems. For patients and their nearest: improved quality-of-life. For medical device and healthcare service industry: better competitiveness by increased quality offering. For clinical professionals: more confidence in decision making. For science: new knowledge, open-science, standards contributions. For policy makers: increased knowledge about risk factors in different populations providing new tools for targeted preventive actions.
Funding
PerCard project is funded by ERA PerMed EU program (ERA PerMed (isciii.es))
Partners and co-operators
The PerCard consortium builds on decades of R&D experience in data-driven decision support for real-life healthcare challenges. It brings a unique opportunity to innovate; fast-track development with large retrospective datasets and thorough validation in a transnational setting.
The PerCard consortium is composed by Tampere University (lead), Politecnico di Milano, Centro Cardiologico Monzino and Protestant University of Ludwigsburg.