I am trying to study the dynamic interactions in time(time series) between physiological variables (heart rate, respiratory rate and blood oxygen saturation) as they progress toward the development of hemodynamic instability in continuously monitored patients. More aptly, I'm looking for data mining resources to perform temporal pattern searches for specified motifs multivariately (eg: search for instances of increasing heart rate and decreasing respiratory rate etc).
Does the KNIME community have any specific thoughts on how I would proceed with data mining this type of a problem?
What you're describing sounds like the area of predictive analytics. A series of attributes (your physiological variables) to the prediction of some kind of label (your development of hemodynamic instability).
Have you looked into the use of neural networks for this kind of work. This kind of work has gained substantial popularity over the last few years and I would believe there is some work being done in this area that you can investigate and build upon. One suggestion is to perform a search for hemodynamic instability and neural networks on PubMed.gov.
Thank you for the advice. Yes, certainly, I am considering neural networks (multilayer perceptron) as one of my options.
This type of work is exciting and rewarding. I did a search and it produced 35,000 results all from a scholarly recognized site. Here is the search I used on Google to obtain those results specific to hemodynamic instability and neural networks. I continue to wish you the best in your research endeavors!
hemodynamic instability AND neural networks site:www.ncbi.nlm.nih.gov/pubmed