Post on 17-Aug-2020
Bioinformatics to Achieve Multiple Modality Monitoring
Peter JD Andrews University of Edinburgh
Lothian University Hospitals Division
PAndrews@ed.ac.uk
Hellenic Society of Intensive Care Medicine
November 9-11, 2012
Introduction
• Multimodal monitoring
• Neurocritical care bioinformatics
• New monitors
KEY POINTS
• Monitoring for secondary brain injury is a fundamental aspect of neurocritical care
• Advances in neuromonitoring technologies now include the ability to directly monitor brain oxygenation, cerebral blood flow, and cerebral metabolism in real time
• BUT data from bedside monitors in neurocritical care are evaluated by clinicians in much the same way as forty years ago!
KEY POINTS
• Informatics has fundamentally changed many fields in medicine - epidemiology, genetics, and pharmacology.
• New data acquisition, storage, and analytic tools applied to neurocritical care data to harness the large volume of data now available to clinicians.
• Neurocritical care bioinformatics will require collaboration between clinicians, computer scientists, engineers, and informatics experts to bring user-friendly real-time advances to the patient’s bedside.
Data Acquisition, Integration, and Time Synchronization
• Paper charts are still the most commonly used data record in the ICU
• Electronic medical records have the potential for: – reducing medical errors, increasing ease of record-
keeping, and aid regulatory oversight • Physiological signals can have frequencies
exceeding 200 Hz, information is lost from data sampled below 0.5–1 kHz (1–2 ms sampling)
“Collecting and archiving data is the crucial first step to information management”
Data Integration • It is impossible to find a comprehensive set of all
physiologic data, patient records, lab work, imaging findings, etc… for a patient in one place
• The inability of different information technology systems and software applications to exchange data accurately and consistently - interoperability – main reason
• “Integrated Clinical Environment” - the American Society for Testing and Materials
• Kiosk or Distributed Systems
Data Integration • It is impossible to find a comprehensive set of all
physiologic data, patient records, lab work, imaging findings, etc… for a patient in one place
• The inability of different information technology systems and software applications to exchange data accurately and consistently - interoperability – main reason
• “Integrated Clinical Environment” - the American Society for Testing and Materials
• Kiosk or Distributed Systems
Data Time Synchronisation
• High-resolution time synchronisation • Automated data cleaning algorithms are needed
to avoid interpreting artifactual data • Once a comprehensive database of integrated,
precisely time-stamped physiologic signals is created (without artifact), – complemented by relevant clinical observations,
laboratory results, and imaging data, clinicians can then begin to formulate and test
hypotheses about the underlying dynamic physiological processes in patients.
Translating Data into Information
• AUC…temperature, cerebral perfusion pressure, PBRO2
• Autoregulation. PRX , ICP/ MAP • Neither uses sophisticated analytic tools to
tackle complex multivariable modeling – but even these simple informatics applications,
digital data acquisition and real-time data analysis are required.
Advanced NCCU Bioinformatics: Making Better Sense of Complexity
Advanced NCCU Bioinformatics: Making Better Sense of Complexity
Self-organizing heat map of physiological variables for neurocritical care. In this neurocritical care heat map physiological variables that cluster on the basis of association within and across patients such that they become hierarchically clustered into three groups of patients. As expected, MAP and ABP cluster together. In this specific case, ICP and FiO2 were unexpectedly clustered, leading to the identification of previously unrecognized ICP elevations during bedside tracheal suctioning in this set of patients with TBI who were mechanically ventilated
Advanced Bioinformatics
• Data-driven methods use existing data to learn to predict an outcome based on newly supplied data – neural networks, regression/ decision trees
• Supervised (know outcomes) or unsupervised learning (data mining) – clustering analysis
• Analysis of nonlinear systems (complex systems analysis)
McQuatt A, Sleeman D, Andrews PJ D et al., Discussing Anomalous Situations using Decision Trees: A Head Injury Case Study. Methods of Infromation in Medicine. 2001: 40 (5); 373-379.
Detrended Fluctuation Analysis “Fractal” scaling
• “Fractal” scaling - similar patterns of variation across multiple time scales
• Conceptualising patients as existing in pathophysiologic “states”
• These states and the transitions are invisible to clinicians using current paper or spreadsheet-based ICU patient records
Avert-IT • Consortium “AVERT-IT” (www.avert-it.org) hospital
intensive care centres, a Scottish software company (C3-Global) and the University of Glasgow National e-Science Centre are developing and assessing software technology for predicting arterial hypotension adverse events in patients with TBI.
• The first year of the study involved training a Bayesian advanced arterial neural network on existing time-series data collected from patients with head injury to predict the occurrence of hypotensive events. The following three years will conduct a two stage clinical trial assessing the AVERT-IT prediction technology.
Hospital Database (s)
PDA/PC Data Collection
Tool used by ITU Nurse
ITU Monitoring
Data Server
Internet PC
Used by ITU Staff
GRID – “Role Based Secure Access Technology”
Project Technology
GRID – “Data Conversion/Mapping Technology”
Hospital or Centre n of N Network Connecting N Centres…
Data Data Data
Request Data or
Grid Model
n+1 n+3
Trial Monitor
ICU
ClinicalAPI
ClinicalClient
Data Push
System Architecture
Hypotension Predicted?
YES
No
Neural Network
Middle layer is a set of modelling Equations that relate the inputs to The output…
Primary Question
• Does use of the AVERT-IT technology lead to a reduced burden of arterial hypotension adverse events during the acute intensive care management of patients with traumatic brain injury?
• Burden will be quantified using: • Duration of arterial hypotension per day.
Secondary Question • Does use of the AVERT-IT technology lead to a reduced
burden of arterial hypotension adverse events during the acute intensive care management of patients with traumatic brain injury?
• Burden will be quantified using • Pressure Time Index (PTI) per day.
Pressure time-index (PTI) is a two dimensional index quantifying in a physiological parameter (such as arterial blood pressure) the duration the signal is beyond a specific threshold and the depth to which it is impaired.
The BrainScopeTM Inc. device connects an 8 electrode, 5- channel EEG, sound for audio evoked potentials and a single channel electrocardiogram (ECG) lead to a palm-top computer. The patient interface is a disposable strip of electrodes placed from ear to ear. Data acquisition requires about 2-3 minutes of noise-free signal.
BrainScope