Over the past years, multiple AI-based algorithms for automatic quantification of cardiovascular risk, body composition, osteoporosis, and chronic obstructive pulmonary disease (COPD) from CT exams have been proposed. Many of those algorithms, including the ones developed by the ARTILLERY consortium, have demonstrated excellent performance. However, very few methods were developed for quantification in radiotherapy planning CT scans and hence, available methods are typically not directly applicable to this data. Moreover, the development of those algorithms has been mostly focussed on achieving accurate performance, while issues regarding trustworthy employment have not gained much attention. Consequently, clinicians lack tools for their daily practice to perform a simple and reliable assessment of actionable chronic disease risks in their patients with breast cancer. Hence, we aim to develop trustworthy AI-based software tools for analysis of medical images addressing four distinct but related, conditions in breast cancer patients: 1) cardiovascular disease, 2) unfavourable body composition, 3) low bone density / osteoporosis, and 4) COPD.
In the current project, we will establish a Real-World Data (RWD) infrastructure for algorithm development at an unprecedented scale: a large repository of routine (annotated) CT scans of >26.000 breast cancer patients, which can be combined with other patient variables and outcomes, and which can be used by partners and other parties for derivation, validation and clinical evaluation of new algorithms. Further, we will set up a collaboration, unique in the field of breast cancer, where AI system developers (AI scientists, software developers) and AI system users (health care professionals and patients) will actively interact along the different stages of AI system development. This way, the ARTILLERY project will facilitate the development of AI systems that are beneficial for and trusted by healthcare professionals and patients, are evidence-based, and ready to be implemented in clinical settings.
Overall, the ultimate ambition of the ARTILLERY consortium is to enhance long term health of breast cancer survivors by making optimal use of routine imaging to identify patients with risk factors or pre-symptomatic chronic conditions. This will empower patients to make targeted lifestyle changes based on their individual risk profile. At a teachable moment during the patient journey (ie. shortly after a life changing breast cancer diagnosis), physicians are facilitated to make use of objective and quantitative parameters of the patients’ general health status, which will help facilitate discussions about the need for risk reducing strategies.
