Skip to main content

Towards a novel paradigm for cardiac function assessment from imaging

Periodic Reporting for period 2 - CardioFunXion (Towards a novel paradigm for cardiac function assessment from imaging)

Berichtszeitraum: 2017-09-01 bis 2019-08-31

CardioFunXion has contributed to a novel paradigm for integrated assessment, combining (image-based) measurements with (model- and machine learning-based) incorporation of physiological knowledge and clinical guidelines. Project outcomes include assessment tools for validation (improving reproducibility) and integration (simplifying interpretation) of cardiac function. These novel approaches took place by connecting the imaging industry with the academic and clinical worlds through a combination of research projects with dedicated training initiatives to develop an intersectorial and interdisciplinary culture.

CardioFunXion had 2 main aims: to develop a framework for the integrated Assessment in Cardiovascular Diseases, and to develop a culture for integrating higher-education, clinical users and technology providers.

The first aim has revolved around the implementation of 4 projects for the multimodal assessment of cardiac function: longitudinal assessment of cardiac function; integration of multimodal measurements into a consistent, physiologically plausible patient representation; designing realistic models for explicit validation tools of novel image analysis algorithms; automated construction of decision trees for heart failure assessment.

The second aim was promoted by the organisation of training events such as the Barcelona VPH Summer School (4 editions), the CardiofunXion winter schools (Paris, 2 editions); the establishment of collaborations at the level of the individual projects with a broad spectrum of actors; and the support to policy actions such as the HUMAINT project from the Joint Research Center of the EC.

The active, enthusiastic and useful involvement of a broad variety of organisations from the industrial, clinical, academic and policy world has confirmed the relevance of the inter-sectorial approach that CardiofunXion proposed with the support of the Marie Curie program.
All projects progressed as expected, with all ESRs defending their PhD thesis in the following months.

CardioFunXion consolidated an open network of academic, industrial and clinical researchers via its open training events. Specifically, 4 Summer and 2 Winter Schools have been organized, covering topics aligned with its objectives (cardiovascular physiopathology; flow phenomena in biomedicine; mechanistic and machine-learning based modelling). They were open to the overall community, to increase the impact of the training actions and of the Marie Skłodowska Curie support within the research and industrial community. They also created opportunities for ESRs to interact with junior and senior profiles, and give ESRs an experience in the organization of research and training events. It involved a wide variety of relevant actors (such as VPH Institute, European Society of Biomechanics, Simula, QUAES Foundation). The last edition was organised in collaboration with the PIC (Personalised in-silico Cardiology) Marie Curie Initial Training Network, in order to promote its continuity after CardioFunXion ends.

Although not explicitly described in the original proposal, the project realised that its impact could be higher, if the platform for industrial-clinical-academic collaboration could also include interaction with policymakers and international organisations, contributing with the expertise available in the project from the unique scientific, industrial and clinical setting that CardioFunXion represents, and providing new opportunities for career development for the early stage researchers in the program.

For this purpose, the CardioFunXion team has been collaborating with the HUMAINT project within the JRC's Centre for Advanced Studies of the EC, which aims to understand the impact of machine intelligence on human behaviour, with a focus on cognitive and socio-emotional capabilities and decision making. The postdoctoral researcher Sergio Sánchez has supported the transfer of knowledge generated in CardioFunXion to the JRC team, in order to allow the researchers make an optimal participation in this collaboration, while not getting distracted from their main goal during this last part of their research project. Both projects co-organized the “Workshop on Artificial Intelligence (AI) in healthcare” (Barcelona, 5/07/19), aimed at AI researchers, biomedical scientists and physicians interested in the applications of AI in healthcare, and its technical and ethical implications. Bart Bijnens, coordinator of the CardioFunXion project, provided the talk “AI for clinical decision-making: challenges and opportunities” (video available in the webpage of the workshop).

The team has also involved collaborations with other reputed international organisations such as the World Health Organisation, the Auckland Bioengineering Institute of the Cardiology Care for Children in USA, just to name a few, providing students valuable contacts for their research and their future professional development, as well as new channels for the translation of the results of the project into other academic, clinical and policy contexts.
All ESRs published their results in peer reviewed conferences or submitted them to international journals:

Regarding electromechanical modeling of the heart, we are innovating in the developments of meshless modeling techniques, alleviating an important source of manual editing and pre-processing.

For longitudinal data analysis, we developed a new statistical framework for integrating heterogeneous measurements collected over successive timepoints, allowing us to address the analysis of challenging datasets such as continuous multimodal stress-testing recordings during nonstandardized protocols, which hold a lot of potential for cardiovascular disease screening.

Regarding decision trees, we extended state of the art algorithms to account for the uncertainty present in clinical assessments. Moreover, we proposed a novel approach to learn a set of decision trees non-greedily that allows the adaptation to constraints of the decision problem.

Regarding integration of data from different modalities, we developed a methodology to relate 2D and 3D echocardiography, assessing the reliability and quantifying the noise level and information that is only available in one modality.

All theses results need to be further consolidated and integrated into software prototypes for being evaluated by clinical and industrial project partners. Enhanced methodological developments are being published in international journals upon completion of the 4 PhD theses initiated thanks to CardioFunXion (two already deposited and to be defended in the last term of 2019, the other two in the process of being deposited). Additional support has been received by the María de Maeztu DTIC-UPF program for the succesful end of the thesis. Thanks to the possibilities offered by the Marie Curie support in establishing and implementing contacts with a large number of entitities, including policy makers, via the individual projects and the open training events organised, the impact of the results go beyond the specific advances at the level of research and the training to the individual researchers involved, and manage to confirm the need for inter-sectorial approaches that involve all relevant actors in the cardiology domain to significantly improve the way scientific results address important clinical challenges for the overall benefit of our society.