Antonio Mosca
Modeling Hemodynamic Stability in LVAD Patients: A Machine Learning Framework Combining Clinical and Pump-Derived Data in the ICU.
Rel. Marco Knaflitz, Thomas Schlöglhofer, Lukas Ruoff. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Biomedica, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (16MB) | Preview |
Abstract
Despite increased utilization of left ventricular assist devices (LVADs), postoperative management in the intensive care unit (ICU) remains challenging. Real-time interpretation of clinical and pump-derived parameters is essential for informed therapeutic decision-making. Therefore, this exploratory study aimed to characterize phases of hemodynamic stability in patients implanted with the Abbott HeartMate 3 (HM3) LVAD, a device currently in clinical use, using a machine learning-based, data-driven approach, proposing a new method to gain insights into patient recovery in the ICU. The analysis was based on data collected from patients treated in the ICU at the Medical University of Vienna, Vienna General Hospital. In this prospective, single-center cohort study, parameters from the ICU digital health records (DHR), including hemodynamics, laboratory values, and pharmacological support summarized by the Vasoactive-Inotropic Score (VIS), were utilized to delineate phases of cardiovascular stabilization following LVAD implantation.
Postoperative days (PODs) were clustered per patient using hierarchical clustering on daily clinical metrics, allowing data-driven identification of three stabilization phases based on the structure of individual dendrograms
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
