Overview

The main aim of FAITH is to apply the latest Artificial Intelligence (AI) and Big Data analytics techniques to better model and predict disease/treatment trajectories of cancer patients, with the goal of improving their quality of life and aftercare. To protect privacy of the individual, but still gain insights that are beneficial to the broader population, FAITH will be applying the concept of federated machine learning, which makes it possible to build machine learning systems without direct access to personal treatment data that will be used for training in machine learning. Devices private to the patient will run their own personalised AI models, via the project’s ‘AI Angel’ application, while a global AI model aggregates the individual model learnings (rather than the traditional approach of a central repository of holding all private patient data). FAITH’s ‘AI Angel’ will remotely analyse depression markers, predicting negative trends in their disease trajectory, giving their healthcare providers advanced warnings to allow for timely intervention. These markers are treated under several distinct categories: Activity, Outlook, Sleep, and Appetite, in accordance with the 3M strategy for population health: Monitor–Measure–Manage. Central to the vision of the FAITH project is to measure population health deeply, it is necessary to monitor individuals on a continuous basis to cast a wide enough net over a user’s health data. A key strength of FAITH is the involvement of eminent cancer hospitals and specialists in the consortium to provide relevant applicable cancer care related use cases that can effectively leverage a big data framework using computational intelligence approaches and methodologies that can be used for long term cancer care health risk and symptom minimisation for patients. FAITH has trial sites in Madrid, Waterford, and Lisbon, with real end users to assess and validate the adoption and usage of the FAITH technologies and platform.

FAITH: a Federated Artificial Intelligence solution for moniToring mental Health status after cancer treatment. The main aim of FAITH is to apply the latest Artificial Intelligence (AI) and Big Data analytics techniques to better model and predict disease/treatment trajectories of cancer patients, with the goal of improving their quality of life and aftercare. The concept of the project is based on a federated machine learning approach, which builds localised machine learning systems without direct access to global personal treatment data. Individual patient devices will run their own personalised AI models, and a global AI model will aggregate the individual model learnings and then synchronise the local models with the updated global model. The AI model will analyse depression markers predicting negative trends to enable timely intervention. The project consortium has trial sites in several European countries to assess and validate the adoption and usage of this new analytical platform.

Implementation

In order to ensure that FAITH’s project objectives are achieved, a clearly defined methodology has been drawn up to show the interdependency and flow between the workpackages and is detailed in the diagram below. The project will start with gathering requirements (WP2) from our end user organisations (hospitals/doctors/patients) as well as relevant stakeholder and policy maker organisations. These requirements drive the architecture specifications, the data reference models as well as real-life use case scenarios (WP2), which are then acted upon for the building of our platform components (WP3, WP4 & WP5). Data from the various sources, including the FAITH App, any connected sensors and our federated AI models are fed into our framework, which through the in-built intelligence will be used to monitor specific mental state markers and further provided for analysis (WP4). This provides the hospital liaison person with a full overview of what is happening in relation to the person who has finished cancer treatment (WP3). These interactions will be trialled (WP6) in consortium end user hospitals, gathering feedback and being validated by doctors and patients with regards to their specific needs. This feedback from these trials will be fed back into a second, and third, round of requirements gathering. After the third iteration, and once we are sure that we have a framework that meets the market needs, we will explore market deployment activities (WP8), which is fully supported by our dissemination, policy & stakeholder engagement activities (WP7).

Objectives

The goal of the FAITH project is to provide an ‘AI Angel’ that remotely analyses depression markers, using federated learning, to predict negative trends in their disease trajectory, giving their healthcare providers advanced warnings to allow for timely intervention. In FAITH we have brought together a number of key partners that can realize the full vision of supporting patients with depression undergoing cancer treatment. The expertise includes defining the use cases and gathering of data sets, defining the ICT infrastructures to process the data (e.g., Cloud and network infrastructure), data and privacy protection, as well as Federated AI to enable data privacy preservation. The partners also include hospitals that will allow test trials to be conducted, which will help ensure that the technologies will be available soon to play a major role in contributing to society. Within FAITH, we will pursue its realisation through 5 concrete objectives:

  • Specify the scope of the of the FAITH framework through elicitation of use-cases; leading to the development of functional and non-functional requirements, user stories to cover these, and a reference architecture.
  • Successfully integrate the appropriate middleware, software components, tools and libraries in order to deliver a federated learning framework for secure experimentation, composition, exploration, and ultimately deployment in line with requirements.
  • Deploy Federated AI models across a broad population base to deliver distinct personalised models.
  • Implement a voice/NLP interface to gauge a user’s mental outlook.
  • Demonstrate the applicability, usability effectiveness and value of the FAITH concepts, models, mechanisms, and techniques in real-world scenario under pragmatic conditions against a pre-defined set of use-cases.

Project Achievements

Project Launched on 1st of January 2020

Contact

Gary McManus (Project Coordinator) – gmcmanus@tssg.org

Philip O’Brien (Technical Coordinator) – pobrien@tssg.org

Funding Body: European Commission Call Topic Identifier: H2020-SC1-DTH-2018-2020, Project Type: Research and Innovation action, Grant Agreement number: 875358