The decision to stop ventilatory support (ventilator weaning) is a complex task. Application of current population-based guidelines often leads to repeated failing weaning attempts or complications due to a prolonged assisted mechanical ventilation. Patient-specific methods allowing for a personalized follow-up in this context are thus of major interest. The possibility of acquiring and processing the increasing amounts available monitoring data opens the possibility to improve the prediction of successful weaning.
- Acquisition of intensive care monitoring data is a complex problem. Specific tools are needed
- The observed monitoring data is noisy, multivariate, heterogeneous and time-dependent, thus difficult to process.
- Explainable ML methods are needed in this context
To evaluate the feasibility of the application of a set of ICT tools and methods proposed by the project partners for patient-specific prediction of successful weaning. If this exploratory project is successful, we would extend the obtained tools and methods to propose a larger project, focused on the optimal, patient-specific management of ventilated patients.
This project will involve 3 complementary partners from CominLabs, that have started initial collaborations in this field in the past and that would benefit from a synergic support to kickstart collaborative work in this subject:
- LTSI-INSERM UMR 1099 (Rennes). Resp. A. HERNANDEZ
- LaTIM - INSERM UMR 1101 (Brest). Resp. E. L'HER
- Lab-STICC - CNRS UMR 6285 (Brest). Resp. J. PUENTES
- Also, CHRU de Brest, through the way of LaTIM, will be involved in this project.