Critically ill patients in intensive care often require artificial lung ventilation. This ventilation is performed by a special apparatus that provides gas exchange (oxygen supply and carbon dioxide removal). Adequate management of the process is one of the key factors of the whole treatment success. Different patients require different regimes of the artificial lung ventilation. It is very important to perform classification of patients into groups requiring similar regimes of ventilation.
PolyAnalyst processed the data from the computerized system “RICA”, developed at Ulm University, that contained data from monitoring devices, blood analyses, results of doctor examinations, XR, tomography, etc. The total number of training examples was about 200. The objective was to develop rules allowing classification of patients into separate groups with respect to certain parameters important for the artificial lung ventilation process settings. For example, patients with excessive, normal, and poor (hypoxia) oxygenation had to be identified.
PolyAnalyst has developed a set of nontrivial rules for each of classification tasks. These rules allow correct diagnostics in the 94% of cases. Here is just one example of derived rules. The fully integrated system for the artificial lung ventilation management, including RICA and PolyAnalyst, is planned for development. Results are described in more detail in the literature.
PolyAnalyst was very helpful for us in analyzing a large amount of data. The first results returned by the system are excellent. We hope to be able to make further use of assistance of the Megaputer’s team in building a fully automatic system to assess the respiration of an intensive care patient.
Bernd Classen, MD
University Clinic for Anesthesiology