Keywords

CLinical decision support, oncology, melnoma, prostate cancer.

Abstract
In current oncological practice, disease management is commonly organized through multidisciplinary oncology consultations (MOCs). With the emergence of individualized medicine, and the increasing amount and complexity of available medical data, a growing need exists for the development of computer-aided clinical decision-support systems (CDSS) for the clinic, based on prediction models of treatment outcome. In oncology, such models should combine all available biological and clinical data to achieve the highest accuracy to predict tumour response and follow-up events.
Patient empowerment is important aspect of modern oncological care, with demonstrated benefit for patient outcomes, and shared decision-making one if its main drivers. A necessary step towards the shared decision- process is to adequately inform the patient on his pathology and treatment options. As cancer treatment is becoming more and more complex, latter task is becoming increasingly complicated.
The continuously increasing complexity of the multi-modal biological data to be considered, has led to a need for computer-based decision support systems. Machine learning has the power to extract knowledge from large sets of routinely acquired clinical data, and is of huge potential for the area of personalized and precision medicine. A downside of the approach concerns the limited interpretability of the resulting models, which is in direct contradiction with the need for verifiability of decision support for the health care professional, and the transparency towards the patient.
The MOC-UP project aims to have positive impact on cancer disease management, by targeting one of the principal instruments employed in the decision-making process of current oncological care: the multidisciplinary oncology consultation. The rationale underlying our approach is that for clinical decision- support to be relevant in a MOC-setting, it should be multi-factorial (based on data coming from multiple disciplines), easy-to-use (allowing it to be integrated in the clinical workflow), and understandable (allowing interpretation for both the caregiver and the patient).
The MOC-UP consortium is composed of a highly interdisciplinary team of reseachers with high degree of expertise in their domain. It includes health care professionals of multiple medical disciplines, psychologists with a extensive experience in studying cancer patient psychology, a leading non-profit organisation in oncological research, industrial players in the area of medical software and patient empowerment communities, experts on business development in life sciences, and researchers in digital health and biomedical engineering.
We will study two pathologies in detail: prostacte cancer and melanoma. The particularity of our approach stems from the fact that, contrary to the current trend of proposing increasingly complex prediction models, we are targetting the interpretability of multi-factorial decision support from the point of view of the researcher, phycisian and patient thereby putting the human back in the loop.
The principal outcomes of the MOC-UP project include clinical decision-support models for treatment response and outcome based on multi-modal data, that can assist in ing the optimal treatment for a certain patient interactive visualizations for high-dimensional dimensional data, allowing to enhance the interpretability of decision support systems for expert clinical researchers integrated dashboards allowing to visualize the multi-modal patient data during MOCs, monitor the evolution of the patient during follow- up, and compare it to the relevant population infographic representations of the clinical knowledge and decision formation, adjustable to the patient needs in terms of health literacy, aimed at improving patient- empowerment and promoting shared-decision making.

Runtime: 2018 - 2021