Information Technology Reference
In-Depth Information
2.2
Related Work
Some researches had been made in order to understand the relationship between the
Plateau pressure and the occurrence of barotrauma.
One study was designed to demonstrate how high pressure could increase the risk
of barotrauma in patients with Acute Respiratory Distress Syndrome (ARDS). The
respective study [6] verified that in ventilated patients with ARDS the occurrence of
barotrauma is unusual these days. The occurrence of barotrauma may vary with the
severity of lung disease, but the standard mechanical ventilation could explain the
occurrence of issues, especially when Plateau Pressure is above 35 cm O [6]. Also
the incidence of Mediastinal Emphysema (ME) and air Pneumothorax were analysed
to identify radiographic patterns and risk factors for the occurrence of Barotrauma.
Another study was conducted in a population of patients receiving mechanical
ventilation in an intensive care unit. The study found that the peak values of
inspiratory pressure, the pressure level at the end of expiration, the respiratory rate
and the tidal volume were significantly higher in patients who developed barotrauma
when compared with the other patients. High values reflect the high incidence of
barotrauma in patients with ARDS [7].
2.3
INTCare
This study is being developed under the research project INTCare. INTCare is an
Intelligent Decision Support System (IDSS) [8], which is in constantly developing
and testing. It is deployed in the ICU of the Hospital Santo António - Centro
Hospitalar do Porto (CHP).
The system monitors patient condition and uses data mining techniques for
predicting patient outcome, patient organ failure, readmissions, length of stay, suggest
procedures, treatments and therapies. The forecasting of Barotrauma is also one of its
objectives. This work is the first step to predict Barotrauma.
2.4
Data Mining
For technical aspects, DM is a process that uses artificial intelligence techniques,
statistics and mathematics to extract useful information and knowledge (or patterns)
from large volumes of data. The discovery of patterns in the data may be in the form
of business rules, affinities, correlations, or terms of prediction models [9]. For this
work, the application of DM was achieved by using a statistical environment R.
R is presented as a programming language and environment for statistical
development [10]. The e1071 library [11] was used to implement the techniques
Support Vector Machine (SVM), Decision Tree (DT) and Naive Bayes (NB). To
perform the evaluations of DM models it was used the rminer library [12].
Search WWH ::




Custom Search