Read your latest personalised notifications
No account yet? Start here
Don't miss out
Ok, got it
Dr. Nico Bruining
The Digital Health (DH) symposia at the annual ESC congresses have without a doubt proven to be increasingly attractive to attendees both on-site and online. Despite the threefold expansion of available floor space for the DH symposia compared to the 2018 congress in Munich, practically all seats were long taken prior to even commencing the diverse sessions (Figure 1). This enthusiasm in digital health and specifically within the realm of cardiology, is also mirrored in the exponential growth of related publications. As eluded to in the inaugural editorial of the Digital Health Virtual Journal, keeping apace in this fast moving domain is a raison d'être of this novel Digital Health Section. While a significant number of ESC sessions were dedicated to digital health comprising a multitude of presentations all deserving commentary, four presentations were particularly noteworthy.
The first of which, Machine Learning in Cardiovascular Imaging, was presented by Jim Min from the Weill Cornell Medical College, New York during the Digital Revolution in Cardiology symposium. The nexus of this enlightening presentation was how machine learning will affect imaging ? His answer was exquisitely delineated into six distinct domains of advancement namely segmentation - using object recognition, integration - in terms of classification of data to perform big data analysis, prediction - using risk stratification , sensing - using continuous monitoring, decision making - in terms of changes in care and finally actuation – in other words machine guided treatment. His group had earlier published a first rate review to supplement his distinct presentation1.
In the same session, the current president of the American Heart Association (AHA), Robert Harrington, from Stanford University, gave an outstanding presentation on big data analysis and interpretation. He highlighted the importance of data interoperability in terms of source data collection from electronic health records (EHR) systems but also the challenges inherent with the monstrous volumes of data collected. For instance, he cited Project Baseline, from Verily, Alphabet’s science division, in which 6TB of data is collected per participant per year. Other facets pertaining to big data were accentuated such as wearables, which are enhancing patient engagement and in consequence positively disrupting the fundaments of future clinical trials. Besides accessing this presentation on the ESC 365 website, a recent paper from Robert Harrington’s group also reinforces the messages2. One final remark on this presentation. Twitter followers have noticed and applauded the magnificent gesture of his team to open up data for public use, a fantastic example of modern day leadership in academia (Figure 2).
Where artificial intelligence (AI) will be used in cardiology ? was the title of the presentation of Callum A. MacRae , Brigham and Women’s Hospital, Boston. He postulated that there will be a shift from the current flow of discovery, first in large populations where artificial intelligence mechanisms currently habituate and then on to therapies with multiple shots “on goal” towards a personalized approach with mechanistic based therapeutic trials. One could consider as an example how AI could predict drug effects which in turn would accelerate innovation or how one could build a “learning “system , a so called biological ecosystem. His team’s work encompasses also the field of genomics embracing advanced computational techniques3. However he noted in his presentation a number of deficits in information content in terms of genomics, drug development and care redesign. His presentation can also be viewed at ESC 365.
The fourth and final highlighted presentation was given by Nicholas Duchateau in the session Machine Learning State-of-the-art entitled Machine Learning what it is and what it is not. All fundamentals were touched upon, such as the need for annotated data, annotation with few landmarks (for imaging) or more landmarks. The need for robust training, validation and test datasets was accented. Furthermore the ever expanding lexicon was elegantly explained ( supervised vs. unsupervised, simple and deep learning neural networks, etc.) An admirable illustration of the potential of ML based cardiovascular image processing in a cohort of heart failure patients investigated to identify responders to cardiac resynchronization therapy has been recently published 4.
From the city of light, Paris, we traverse across Europe to Tallinn, Estonia, home of Skype, Spotify, Linux and e-citizens, for the first ESC Digital Health Summit (5-6 October 2019). This inaugural summit will cover amongst others topics ranging from the digital future, the role of mobile health, apps, wearables, man vs. machine, virtual and augmented reality but also ethics and practice, cybersecurity and in-silico trials. Also innovative presentation formats will be here launched including TED talk like presentations and key opinion leader roundtables.
With a bright but busy future in the various cardiovascular digital environs ahead of us, it’s worth recalling the jovial words of the famed author John Steinbeck – “Ideas are like rabbits, you get a couple and learn how to handle them and pretty soon you have a dozen”.
Declaration of Interests: The author has declared no conflict of interest in relation to this publication.
Our mission: To reduce the burden of cardiovascular disease.
© 2020 European Society of Cardiology. All rights reserved.