Work packages – Artillery

Project introduction

Women with breast cancer are at increased risk of cardiovascular disease (CVD), partly due to anthracycline-, trastuzumab- and radiation-induced cardiotoxicity, and partly due to shared risk factors like overweight and physical inactivity.

Most breast cancer patients (~65%) are treated with radiotherapy. Planning computed tomography (CT)-images are obtained for delineation and computation of radiation dose distribution fields. These CT-images contain information on risk factors for CVD, e.g. calcifications in the epicardial coronary arteries and aorta. This potentially valuable but unrequested information is currently not systematically assessed nor reported by professionals, largely due to time constraints and unfamiliarity with its potential importance.

In the ARTILLERY project funded by Horizon Europe, we aim to develop, validate, and prospectively evaluate AI systems for automated early detection of CVD risk (factors) in women with breast cancer by using routine CT-images.

WP 1

ARTILLERY Coordination and management

Objectives: To secure a smooth and timely execution of the Work Plan in fulfilment of the Grant- and Consortium Agreement, adapt to necessities and ensure that ARTILLERY’s objectives and outputs are achieved on schedule, within budgetary limits and within the resource constraints; To perform quality control, monitor and report on project implementation, progress and resources and timely handle risks and contingencies; To manage the overall day-to-day administrative, legal and financial aspects of ARTILLERY project and facilitate optimal internal and external communication.

WP 2

AI development and validation

Objectives: In this WP, we will develop AI-based image analysis tools to quantify markers of i) cardiovascular disease (CVD), ii) body composition, iii) osteoporosis and iv) lung and airway abnormalities. Subsequently, we will use these markers to design AI-based models predicting hospitalization and all-cause mortality. Finally, a framework for integration of the designed AI-based image analysis into clinical workflow will be developed.

WP 3

Training and validation of AI systems in breast cancer patients in a Real-World Data (RWD) repository

Objectives: Create a Real-World Data (RWD) repository; Generate governance for access to, and use of, RWD repository; Use RWD repository to evaluate robustness, reproducibility and clinical added value of AI systems developed in ARTILLERY project.

WP 4

Clinical implementation and evaluation of AI systems

Objectives: In this work package, we will work towards clinical implementation and evaluation of an AI system- based workflow. Specific goals include: Prioritization of AI systems that are relevant to breast cancer patients and their doctors; Development of manuals / recommendations for healthcare professionals on management of patients identified by AI systems to be at increased risk of chronic diseases; Conduct a decision impact study (ARTILLERY-DI) to evaluate the uptake and acceptability of AI systems in routine care and to measure to what extent the use of AI systems affects clinical decision making.

WP 5

Ethical and legal aspects of implementation of AI-systems

The success of any new clinical application greatly depends on the acceptance and trust in the application amongst its end-users, which are in our case patients and healthcare professionals, as well as society (as a whole). The aim of WP5 is to ensure trust and acceptance of imaged-based AI systems in clinical practice amongst patients and healthcare professionals.

WP 6

Dissemination, Exploitation and Communication

Objectives: To ensure efficient and effective dissemination of ARTILLERY’s approaches, goals and results to stakeholders to maximize the potential societal and scientific impact of the project; To enable engagement and input from all stakeholders (incl. clinicians, patients, patient organizations, and the general public) so that the needs of all groups can be considered to facilitate the uptake of the results.

Methods

A large data repository including routine CT scans of 26.000 women with breast cancer along with relevant patient variables and CVD outcomes will be set-up from multiple hospitals in Europe (i.e., University Medical Center (UMC) Utrecht, Amsterdam UMC, Region Hovedstaden, Champalimaud Foundation).

This repository will be used by computational scientists to develop and train AI systems for early detection of CVD risk. Using the same repository, methodologists will evaluate the validity of these AI systems.

In parallel, clinical practice guidelines for management of patients automatically identified by AI systems to be at increased risk of CVD will be created by a multidisciplinary team of (breast cancer) professionals for patients with increased risk of CVD and identified by AI systems.

In a multicenter prospective decision impact trial, the impact of implementation of the AI systems on patients’ risk profiles, health status and wellbeing will be evaluated. Finally, we will work towards the valorization of trustworthy AI-based software and product development with the aim to guarantee that ethical principles of trustworthy AI-systems in regular breast cancer patients care will be met.