HomeResearchHealth Data Management & SemanticsMining Large Complex Datasets

Mining Large Complex Datasets

In order to process/mine continuous physiological time series data, collected through various instruments in modern intensive care units (ICUs) and during surgery, we aim to develop novel mechanisms for data compression, trend analyses and data classification.

Time Series Data Compression

In processing large scale, high frequency time series data, a key requirement is the development of efficient and effective compressing algorithms in near real time. Our current work is to develop algorithms that support low time/space complexities. For example, our wavelet based tools support lossy online compression on one/multi dimensional data, with certain error bound guarantees, leading to improvements in processing time and lower storage requirements.

Clinical decision are often influenced by the analysis of trends in physiological signals data (e.g. blood pressure, heart rate, respiration rate, ECG, body temperature, SPO2, CO2(FI), CO2(ET), N2O(FI), N2O(ET), O2(ET), O2(FI)). Simple forms of trends, such as Up, Down, are frequently used as key descriptions of clinical symptoms. For instance, if a patient experiences hypotension (i.e. blood pressure goes down) and bradycardia (i.e. heart rate goes down) during an operation, clinicians may consider the treatment of anaphylaxis or cardiac arrest. In this project, we are developing algorithms and tools that detect these trends in physiological data. The technology is likely to be incorporated into patient monitors to provide warnings to clinicians for adverse trends.

Recent research suggests a strong association between certain characteristics of blood pressure (BP) and the early outcomes of Acute Ischemic Stroke. Extreme hypertension and hypotension on admission have been associated with adverse outcomes in acute stroke patients. In this project, we are investigating short period, beat-to-beat, no-invasive BP of acute ischemic stroke patients, within 72 hours of ictus. We are applying our algorithms and tools in order to predict stroke outcomes based on acute phase short-period BP monitoring

Anaesthetic Data Analyser

Our Anaesthetic Data Analyses (ADA) prototype is a clinical trend annotation tool which can analyse stored or real-time physiological data. ADA includes four basic components: data compression, data visualisation, data trend annotation and online vital trend warning.

 

Last Updated on Thursday, 29 September 2011 12:39

 
Go to top

Contact

Dr Michael Lawley

 +61 7 3253 3609
This email address is being protected from spambots. You need JavaScript enabled to view it.