The current healthcare system does not take full advantage of the huge potential of new technologies, which could revolutionise patient care and treatment. YouForAll (Your digital twin for allowing a healthy society) is a project from the Alta Scuola Politecnica XVII cycle that, in cooperation with Dedalus Italia, a company active in the world of clinical software, undertakes the challenge to move a first step towards this revolution. Indeed, the goal of the project is to realise an efficient system to support clinical decisions based on Artificial Intelligence (AI) techniques. The main innovations of this system with respect to existing ones are the high efficacy of the support offered and the clear explanations behind the provided suggestions, which can be easily understood by doctors.
As a matter of fact, the current approach to patient care still relies for the most part on clinical guidelines, sets of rules that define questions to be asked to patients based on their symptoms and actions to be undertaken based on their answers (e.g. what is the final diagnosis or which exams should be performed given the patient’s health status). However, these guidelines fail to personalise the therapies and procedures to the specificities of each patient, leading to suboptimal treatments. Moreover, they need to be manually consulted by clinicians, resulting in poor healthcare efficiency.
In this context, YouForAll proposes NEAR (Neural imputed Explainable and Adaptive Risk score), an AI-based score to predict the risk of a clinical event. As proof of concept, NEAR is implemented to predict the risk of death and bleeding events for patients who have already suffered from cardiac disease, but the approach can be effortlessly extended to any other clinical condition for which a sufficient quantity of data from patients is available. Beyond its capability to model, in advance, the risk for a patient to incur in a given clinical condition, NEAR also provides easy-to-interpret explanations about its own predictions, thanks to which a clinician can immediately understand the clinical variables that contribute more to the score. Moreover, NEAR suggests actions to mitigate the risk or to have a more accurate prediction of the likelihood of the clinical event. Therefore, NEAR acts as a clinical decision support system for practitioners, who can integrate the suggestions with their professional experience to improve their diagnoses.
Principal Academic Tutor
Emanuele Della Valle, DEIB, PoliMi
Academic Tutor
Marco Agostino Deriu, DIMEAS, PoliTo
Michela Sperti, DIMEAS, PoliTo
External Institutions
Dedalus Italia S.p.A.
External Tutors
Davide Guerri, Dedalus Italia S.p.A.
Simone Paolucci, Dedalus Italia S.p.A.
Riccardo Banali, Dedalus Italia S.p.A.
Team members
Andrea Cavallo, Computer Engineering, PoliTo
Davide Fassino, Mathematical Engineering, PoliTo
Karim Kassem, Biomedical Engineering, PoliMi
Andrea Mario Vergani, Computer Science and Engineering, PoliMi