This project aims to develop a semi-supervised hybrid patient data model. The team has currently demonstrated a fundamental machine learning model for dissolved oxygen estimation and only introductory papers on the mechanistic model have been published so far. The goal of this project, meanwhile, is to build a new model-based patient assessment tool (improving or deteriorating condition) that provides a hybrid solution between a known or implicit mechanistic model and expert rules that are known from past experience. This study is relevant in intensive care units where patients are connected to artificial lung ventilation and timely knowledge based on a cross-cutting assessment is a key tool for the clinician to consider a decision to change the program or its settings. When combined redundantly with existing specific known signal trends, this tool in the form of a model would be for the clinician more than just another criterion to assess the patient’s condition. Clinical trial data will be provided to the University by JSC Cumulatis for the development of the model, subject to the approval of the Bioethics Committee. Our team hopes that the model will be of particular importance for future classification problems in predicting inflammation types in short-term diagnostics.
Project funding:
Research Council of Lithuania, Designated Programme “Information technologies for the development of science and knowledge society”
Period of project implementation: 2025-09-01 - 2027-08-31
Project coordinator: Kaunas University of Technology