AI-POD structure and Work Packages

Welcome to AI-POD, a cutting-edge project aimed at improving cardiovascular health in obese individuals through advanced AI research. Our mission is divided into eight key work packages, each bringing together cutting-edge technology, advanced healthcare research, and ethical guidelines to shape the future of AI in healthcare.

Work Package 1

Our journey begins with WP1, where we’re building a high-powered, cloud-based data platform for managing and processing large volumes of healthcare data. The platform will harmonize and streamline data across diverse clinical partners and integrate into existing clinical information systems.

WP lead: Hilde Andersen, CMRAD

Building a Cloud-Based Data Platform for Advanced Healthcare Research

Work Package 1 (WP1) focuses on the development of a specialized cloud-based data platform to support the project’s data-intensive healthcare research. This high-powered platform is dedicated to managing, curating, and processing large volumes of multi-modal healthcare data.

Key objectives of WP1 include creating the data platform and aligning it with consortium requirements, ensuring its compatibility with different data formats (such as DICOM, laboratory data, and clinical data), and equipping it with robust data privacy protection and information safety supervision functions.

The cloud-based data platform will be operationalized in three main stages: platform development and deployment, clinical partner preparation and deployment, and data ingestion and structuring.

The platform development and deployment stage involves the initial setup, agile development of specific features requested by consortium partners, and quality assurance. The subsequent stage prepares the clinical partners for data upload, ensuring a harmonized workflow across different partners. The final stage is focused on setting up data routing paths, extracting data, and integrating the platform into clinical information systems.

Ultimately, WP1 aims to create a versatile and secure data platform that facilitates effective data management and accessibility for advanced AI-driven healthcare research.

Work Package 2

In WP2, we collect and process critical data, particularly cardiac CT examinations. Our innovative photon-counting CT protocol will help us capture high-resolution images with minimal radiation exposure. With the help of advanced software, we extract vital features and meticulously curate and annotate the data.

WP lead: Hatem Alkhadi, UZH

Data acquisition, processing, curation and annotation

Work Package 2 (WP2) focuses on the comprehensive management of data required to create an effective AI model for analyzing heart health in obese patients. This entails the collection of a significant volume of cardiac CT examinations and related health data. Once gathered, these data undergo thorough curation to ensure they’re optimized for AI analysis.

A substantial aspect of this work package involves the development of an innovative photon-counting CT protocol. This technology will provide high-resolution images at the lowest possible radiation dose, making it a safer and more efficient tool for imaging obese patients.

Furthermore, this work package will employ advanced prototype software to extract vital features from the collected imaging data. This software aims to reveal details about the cardiovascular system that aren’t typically part of standard clinical examinations.

Finally, WP2 ensures all data are meticulously curated and annotated. This rigorous process ensures high-quality data are available for AI model development and validation, providing a robust foundation for subsequent project stages. By harmonizing and refining data from a diverse range of sources, WP2 helps create a comprehensive resource to aid in the analysis and prediction of cardiovascular health in obese patients.

Work Package 3

As we move onto WP3, our focus is on creating explainable AI models for predicting individual risk of adverse cardiac events. The AI-POD risk score is born from rigorous research and feature extraction from both imaging and non-imaging data. This scoring system forms a cornerstone of the project, mapping to the development of the Clinical Decision Support System (CDSS) and the Citizen App.

WP lead: Georg Langs, MUW

Explainable AI models for prediction

Work Package 3 (WP3) aims to develop an AI-POD risk score to predict individual risk of adverse cardiac events in obese people. This work involves creating two versions of the risk score; the first (AI-POD-1) employs features identified from previous research, while the second (AI-POD-2) includes additional variables identified algorithmically.

A significant portion of WP3 involves the extraction of features from imaging data and development of machine learning methods to quantify aspects of heart health, such as plaque composition, pericardial fat, and myocardial mapping. In addition to imaging data, non-imaging data, including laboratory results and lifestyle parameters from the Citizen App, will be harnessed and harmonized across different sites to support risk calculation.

Furthermore, WP3 focuses on creating integrative scoring models that map both imaging and non-imaging features to an individual risk score. It also looks into continually adapting these models to align with changes in real-world clinical environments, ensuring that the AI-POD scoring algorithm stays robust amidst advancements in imaging technology or changes in patient populations and clinical pathways. Ultimately, the risk scoring engine produced by WP3 will serve as a cornerstone for the development of the CDSS and the Citizen App.

Work Package 4

WP4 is the hub where all this data is put to work. Here, we develop an AI-assisted CDSS and a Citizen App to provide accurate, personalized risk predictions. Users will gain insight into their risk factors, while healthcare professionals will use the dashboard to make well-informed decisions.

WP lead: Maik Bido, MV

Integrative Clinical Decision Support System and Citizen App

Work Package 4 (WP4) aims to develop and validate an AI-assisted Clinical Decision Support System (CDSS) and a complementary Citizen App for assessing and predicting the risk of cardiovascular diseases in obese individuals. The CDSS incorporates a variety of data, including medical histories, laboratory parameters, lifestyle data, and demographic information, to create accurate, personalized risk predictions. The system will utilize a combination of machine learning and rule-based approaches, forming a white-box model, to derive diagnostic suggestions.

A significant component of WP4 is the creation of a Citizen App, which will connect patients to their health data and the AI-POD risk score. The app will be designed under Design Thinking methodologies, making it user-friendly and adaptable. The app will allow patients to better understand their risk factors, as well as help physicians make informed clinical decisions.

In addition, WP4 aims to design a dashboard to summarize all relevant data for integrated diagnostics in clinics, aiding healthcare professionals in making well-informed decisions. Strict compliance with GDPR will be ensured during the entire process. Moreover, a patient monitoring feature will be developed to alert physicians of changes in a patient’s health condition that exceeds defined thresholds.

Work Package 5

In WP5, we address the ethical aspects and challenges of using AI in clinical tools. We’re actively engaging with stakeholders, ensuring transparency, inclusivity, and ethical alignment. We’re particularly interested in the patient experience and will be continually refining our tools to meet their needs.

WP lead: Pascal Borry, KUL

Ethical aspects and challenges of AI based clinical tools

Work Package 5 (WP5) in this project focuses on the ethical aspects and challenges of AI-based clinical tools. With an emphasis on Responsible Research and Innovation (RRI), WP5 seeks to ensure transparency and inclusivity in the development and deployment of AI innovations in healthcare, specifically relating to the prediction of obesity-related vascular diseases. The package will assess and analyse current ethical frameworks and governance models, taking into account issues like informed consent, data protection, and equity of access.

WP5 will also involve engagement with various stakeholders such as physicians, AI developers, patients and public health officials to gauge their views on using AI for predicting obesity-related diseases. This input will inform the ethical alignment of AI technology with societal needs and expectations.

Furthermore, WP5 aims to evaluate the patient experience of obese individuals using the mobile Citizen App, investigating factors such as acceptability, motivation, and user satisfaction. These assessments will help refine the app to ensure it effectively meets the needs of the user population in a respectful and ethical manner.

Work Package 6

Next, in WP6, we put our AI-POD tools to the test in a real-world setting. We’ll be conducting an observational study with 1200 obese patients suspected of having cardiovascular disease, comparing our tools to standard risk assessment methods.

WP lead: Ulrike Attenberger, UKB

Proof of concept study

Work Package 6 (WP6) is aimed at validating the AI-POD tools which include a risk score, a Clinical Decision Support System (CDSS), and a Citizen App. This validation will be conducted in a real-world, multi-disciplinary setting involving multiple clinical stakeholders.

To achieve this, WP6 plans to conduct a prospective observational study involving 1200 obese patients who are suspected to have Cardiovascular Disease (CVD). The participants will undergo two cardiac CT scans, one at the beginning and one after two years, along with regular tracking of various clinical and laboratory parameters.

Half of the participants will be randomly selected to use the Citizen App and receive a separate fitness tracker device. The patients’ progress and responses to these AI-POD tools will be monitored and compared with the standard risk assessment methods for CVD.

The aim is to evaluate the efficacy of these AI-based tools in predicting and managing CVD among obese patients. The results will guide refinements to the AI-POD tools and inform future development in AI-aided healthcare solutions.

Work Package 7

WP7 focuses on spreading the word about AI-POD, involving end users, and planning for the project’s long-term viability. We’re excited about sharing our research outputs and ensuring that the AI-POD system continues to develop and make an impact.

WP lead: Katharina Krischak, EIBIR

Dissemination, end user involvement and exploitation

Work Package 7 (WP7) of the AI-POD project, led by EIBIR, encompasses three primary goals: dissemination, end user involvement, and exploitation. This package aims to make the research outputs of AI-POD widely known and used among relevant target groups and ensure the long-term viability of the product.

Dissemination efforts will involve various outreach activities, the creation of a communication kit, and a strategy for communicating project information effectively. AI-POD’s identity will be developed and conveyed through a new website, social media, print materials, and two promotional videos.

In terms of end-user involvement, a Stakeholder Board comprising clinicians, patients, AI developers, and decision makers will provide feedback throughout the project. There will also be workshops, tutorials, and a survey to gather user feedback and adapt the software accordingly.

The exploitation and sustainability part of the package is focused on ensuring the continued use and development of the AI-POD research, system, and app after the project concludes. A strategic plan will be set in place to leverage project results for future research activities, policy-making, and potential commercialization. Information about IP developments, patenting, licensing, and more will be collected and evaluated for future applications.

Work Package 8

Finally, in WP8, we oversee the entire project’s management, ensuring its smooth operation, maintaining communication among consortium members, and adhering to principles of open science and data management.

WP lead: Ulrike Attenberger, UKB

Project Coordination and management

Project coordination and management
Work Package 8 (WP8) of the AI-POD project, led by UKB, is dedicated to project coordination and management. It aims to handle all administrative, scientific, and financial aspects of the project over the span of 48 months.

The project coordination and management aspects involve overseeing the progress and achievements of the project, managing resources, and monitoring activities. A consortium of bodies will facilitate effective communication, decision making, and coordination of scientific and technological activities.

An essential part of WP8 is data management, ensuring data from the project aligns with the principles of open science and FAIR data. A Data Management Plan will be drafted and updated continuously throughout the project, detailing what data will be shared and what will remain confidential due to GDPR or IPR concerns.

Another key aspect is the management of the External Advisory Board (EAB), which consists of international experts who offer independent assessment and advice on the project’s scientific, technical, or ethical aspects. The EAB will also act as multipliers for the project’s dissemination activities and provide progress reports, recommending any necessary corrective actions.

We believe that the integration of AI in healthcare holds immense promise, and we're thrilled to share this journey with you as we make strides towards a healthier future.