#	Pagina
attuale pagina	/open-h2020/projects/217377/results.html

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - AI4EMS (Artificial Intelligence for Emergency Medical Services: a smart digital assistant for faster and more accurate cardiac arrest recognition during emergency calls)

Teaser

Summary

What is the problem/issue being addressed?
In the EU-28, Cardiovascular Diseases (CVD) account for 37% of all deaths, over 1.8 million deaths annually, and it costs the EU economy an estimated â‚¬210bn per year mainly due to health care costs and productivity losses. As CVD prevalence and costs are projected to increase substantially, Out-of-Hospital Cardiac Arrest (OHCA) is projected to rise in the same proportion during next years. To date, OHCA is one of the leading causes of death in Europe, and worldwide. Currently, Emergency Medical Services (EMS) dispatchers use set medical protocols as a support to recognize medical conditions, like OHCA, during emergency calls prior to activating the emergency response system. However, EMSs are struggling as calls have increased in Europe from 100 million calls in 2003 to 300 million in 2018, stretching already thinned resources to the limit. Assistant decision tools will be necessary to help EMSs to faster identify time-critical conditions like a OHCA situations, and in a more sustainable way.

Our solution, AI4EMS, is a Software-as-a-Service (SaaS) platform that integrates state-of-the-art speech recognition and machine learning for augmenting in real-time the performance of human EMS dispatchers. AI4EMS is the first and only smart digital assistant for EMS dispatchers that supports triage decision-making process by (1) real-time processing and analysis of emergency calls; (2) real-time recognition of cardiac arrest in an evidence based process for Danish, English, French and Italian (and more languages to come); and, (3) presenting the most important insights to dispatchers in a user-friendly way. With AI4EMS, OHCA recognition is faster (reducing the EU average of 3â€™39â€™â€™ to 50 seconds) and more accurate (increasing the EU average of 73.9% human accuracy to 95%).

Why is it important for society?
Healthcare is like every other market driven by the relationship between supply and demand, and patients demand medical expertise since nobody wants mediocre medical treatment. This means that every job function is usually highly specialized, and it takes a lot of training and retraining to keep every employee up to date with best practices. This drives up the cost of treating each patient, which in turn makes resource utilization and allocation more important than ever before. Resources that are already scarce considering the growing aging population and the widely documented imbalances and shortages of health workforce in the European region, and globally. In this scenario, Emergency Medical Services (EMSs) are no exception. Effective decision assistant tools will have a large impact in healthcare, bridging the widening gap between medical professional resources and patients\'needs.

What are the overall objectives?
AI4EMS will impact the economy and society at large by improving access and quality of healthcare. Our goal is to disrupt the Artificial Intelligence (AI) market for healthcare becoming world leaders in EMS artificial intelligence, saving lives and unnecessary costs. Sales will render revenues of â‚¬86.7 million in the first five years of commercialization and a total of 127 new jobs will be created by 2024, generating an estimated total economic benefit for all our end users of â‚¬108m.

Work performed

During the first period of the AI4EMS project, we have been primarily focusing on the technology maturation and final prototyping of our product, that allows us to reach a higher technology readiness level. This entailed a major allocation of resources and time towards the optimization of several aspects of our core product (speech recognition model, triage classifiers and cardiac arrest detection model, user interface and technical configuration/ integration work).
In line with the planned timeline of the project the first milestone \'Finalization of piloting requirements\' has been reached and it will set the foundation for the \'Large-scale piloting and validation\' during the second period. The European piloting sites selected through our collaboration with the European Emergency Number Association (EENA) are the EMSs of AREU (Milan, IT) and SAMU74 (Annecy, FR). This allows AI4EMS to double the number EU languages that it is able to recognize (Danish, English, Italian and French) and it enables the deployment and configuration of the AI4EMS technology on each piloting activity site for large-scale testing and training purposes. The early pilot sites of Copenhagen (DK) and Seattle (USA) have already implemented the AI4EMS technology at their premises and training of their end-users.

Clearly, AI4EMS is moving towards its commercial readiness expected for the end of the project, Q1-2 2020. To that regard, the foundation for effective Sales and Customer support unit have been established, and during the second period they will be expanded as we approach market launch.
A wide range of communication activities has been carried out during the first period in order to raise awareness on the project, reaching out to the general public as well as selected target groups (PSAPs, Policy makers, Investors, scientific community, etc.). We attended key industry and scientific conferences and have been working towards publishing research articles, some of which have been already published: one scientific article and four conference papers. During the second period of the project we will focus on maximizing the dissemination of project results, ensuring accessible dissemination materials via open access.

Final results

AI4EMS hinges on the basic proposition of proving the proficiency of a machine learning model functioning in a realm of multi-modal data, usually only operated by humans. If the technology lives up to itâ€™s promise, as it is proving to do here at the first reporting period, we are on the verge of taking an extremely promising technology out of the realm of untested and highly localized, all the way to the mass market as it will be working across location, dialect and language.

This will impact the global debate on what tasks and roles are uniquely human, and how we define our jobs in the future healthcare system, which is an extremely important debate that Corti as a company will need to play a role in shaping and facilitating, as the technology matures and expands to new markets and use-cases, changing our perception of clinical care and good emergency management.