cardioRisk Personalized Cardiovascular Risk Assessment through Fusion of Current Risk Tools

Description

This project addresses the coronary artery disease (CAD) and, specifically, the management of myocardial infarction (MI) patients. The main scientific goal is the development of personalized clinical models for cardiovascular (CV) risk assessment of acute events (e.g. death and new hospitalization), in order to stratify patients according to their care needs. The models significance will be assessed trough an in-hospital observational study, addressing patients admitted in the Intensive Cardiac Care Unit with a first episode of acute MI. The work will follow three major directions: fusion of risk tools, clustering of patients’ data and biosignals analysis techniques. This project will be partially supported by the results achieved by the team in the European project HeartCycle (FP7-216695) (pHealth applications to improve compliance and effectiveness in heart failure and CAD closed-loop management). Of particular importance are i) the development of fusion approaches to merge CV risk tools, ii) strategies for the prioritization of patients and iii) biosignal processing techniques to achieve personalized diagnosis. The team is composed by two research institutions, the University of Coimbra, Politecnico di Milano (annex 1) and the Centro Hospitalar Leiria Pombal (annex 2). It is also supported by a major company in the eHealth area, criticalHealth (annex 3). Motivation Although fundamental for the assessment of the patient’s CV risk, risk tools used in daily clinical practice present some limitations. First of all, they ignore the information provided by previously developed risk tools. Moreover, they are average models, derived for a general population and thus not ideal for a specific patient. Additionally, these tools are rigid and do not allow the incorporation of new risk factors or clinical expertise. The derivation of personalized models that cope with the referred drawbacks, would represent a significant improvement for currently available tools, with a positive impact on a patient cardiac risk evaluation. They would assist professionals in managing each specific patient, helping to identify those in need of urgent attention and/or review of care plans. Furthermore, improvement of diagnosis could have a large social and economic impact. In fact, each year CV disease causes over 1.9 million deaths in the EU (42% of all deaths), and contributes to health costs by €105 billion. Methods By implementing fusion methodologies, several available individual tools will be combined in order to mitigate aforementioned drawbacks. Bayesian probability theory assumes a key role in this context, given its capability to support a common representation for those tools (thus facilitating their combination) and, simultaneously, to provide a basis for clinical expertise description. A fusion scheme, using Bayesian inference, was derived in the HeartCycle project, combining in a global model several risk tools. The second direction of research will be the development of strategies to identify groups of patients with similar characteristics, to achieve personalization through the selection of the most adequate model for each patients’ group. To this aim, computational intelligence methodologies, including dimensional reduction and clustering approaches, will be explored. Furthermore, in order to improve accuracy as well as personalization, the incorporation of additional parameters will be investigated. Among the parameters recognized as clinically significant to improve risk stratification but that are not currently incorporated in risk assessment tools, are those derived from heart rate variability (HRV) measurements. HRV, an electrocardiogram (ECG) derived signal, is a strong and independent predictor of mortality in patients following acute MI. In these patients, HRV measures were found to have a significant association with all-cause mortality, cardiac death, and arrhythmic death. As a result, by combining HRV derived parameters with other parameters, risk stratification might be markedly improved, particularly for what concerns the identification of risk groups. Results Within the project an integrated clinical platform will be implemented, including the developed models and algorithms, will be assessed under real conditions in a hospital environment (CHLP, Leiria). These models will consider as inputs historic data, such as demographics, biomarkers and clinical exams that can be obtained from the hospital information system. To obtain HRV measurements, the ECG will be acquired using a Holter device, to be purchased by the project. The observational study will address patients admitted in the ICCU with a first episode of acute MI, managed according to the current European guidelines. Three exclusion criteria will be considered: artificial pacing, previous heart failure and valvular heart disease. The primary endpoint will be death, plus new hospitalization due to heart failure 30 days after the event.