Using neural networks for prediction of central venous pressure during open-heart surgery

  • Authors:

  • Miljenko Križmaric;Peter Kokol;Mirt Kamenik;Dušanka Mičetic-Turk

  • Affiliations:

  • University Nursing College School, University of Maribor, Žitna ulica, Slovenia;FERI, University of Maribor, Smetanova, Slovenia;General Hospital Maribor, University of Maribor, Ljubljanska, Slovenia;University Nursing College School, University of Maribor, Žitna ulica, Slovenia

  • Venue:
  • AIC'04 Proceedings of the 4th WSEAS International Conference on Applied Informatics and Communications
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we describe a use of Neural Networks (NN) of formal neurons for predictions of Central Venous Pressure (CVP) during open-heart surgery. As input variables in the system of neural network we use primary non-invasive parameters and in the second simulation we added invasive physiological parameters of patient. CVP is a pressure in the right atrium of the heart, which is measured in very invasive way with catheter, which is placed directly into the heart. We assume that value of CVP can be predicted with only non-invasive parameters. At first we teach NN only with non-invasive input parameters captured from anesthesia machine, ECG monitor and pulse oximeter. In the second part of simulation we added invasive blood pressure (IBP) to non-invasive input parameters, which is measured in radial artery. Neural network has showed 75% correlation between central venous pressure (CVP) and non-invasive parameters. When we consider additional invasive parameter as neural network input, then success rate increase to 80%.