Database Reference
In-Depth Information
OHA has been extended with patient similarity concepts to help physi-
cians take decisions while leveraging past experiences gathered from sim-
ilar patients that have been monitored in the past [30]. In [31], the MI-
TRA system, introduced as an extension of OHA, allows physicians to
query for similar patients and use records from these similar patients to
make predictions on the health evolution of a patient of interest. An
in-silico study using physiological sensor data streams from 1500 ICU
patients obtained from physionet [32] shows how MITRA may be used
to forecast the trajectory of blood pressure streams and help predicting
acute hypotensive episodes in ICUs. In [33], similar approaches to time-
series forecasting with applications to intensive care are also reported.
Patient similarity techniques are described in this thesis as a way to
extract robust features for forecasting purposes. Sequential learning
techniques with Linear Dynamical Systems and Hidden Markov Models
are proposed for the modeling stages.
State-of-the-art Analytics for Intensive Care Sensor Data Mining:
State of the art analytics and mining approaches for in hospital sen-
sor data monitoring tend to generate innovations on data pre-processing
and transformation. Modeling is typically done with well known families
of machine learning techniques such as classification, clustering and dy-
namic system modeling with sequential learning. These analytical tech-
niques often attempt to derive features from physiological time series to
model the inflammatory response of the body, as it is known to be highly
correlated with early sign of complications in general. The inflammatory
response is a reaction from the body to different harmful stimuli such
as pathogens, various irritants or even damaged cells. Hence, accurate
modeling of it enables a wide range of early detection applications in in-
tensive care. In particular, devastating complications such as sepsis are
known to produce an inflammatory response well before the appearance
of clinical symptoms [51].
The inflammatory response is controlled by the autonomic nervous
system, consisting of the sympathetic and parasympathetic nervous sys-
tems [40]. These systems regulate several involuntary actions such heart
beats, respiration, salivation, transpiration etc. Inflammation results in
poor regulation of these systems, and is often correlated with the Sys-
temic Inflammatory Response Syndrome (SIRS) [41], [42]. The poor
regulation manifests itself in loss of signal variability associated with
physiological sensor streams. As a result, several researchers have at-
tempted to model the inflammatory response using various measures
Search WWH ::




Custom Search