Information Technology Reference
In-Depth Information
burden is put on how to take an optimal advantage of available resources, i.e. machines,
softwares, network resources, etc. For instance, there are many attempts for inventing
new generic methods for delegating the reconfiguration capabilities to the middleware.
In these approaches the knowledge of the high-level application goals is not fully ex-
ploited. We think that such knowledge can be used for processing and abstracting input
data as well as for guiding the selection and instantiation of high-level goals from input
data features. We advocate the use of a central pilot to achieve this task. Some authors
[9,19] have advocated distributed adaptation. We think that distributing control knowl-
edge is difficult since high-level and low-level aspects are strongly connected and must
often be treated synchronously.
In addition, we have proposed a supervised method to learn such temporal patterns
automatically from annotated temporal data. In [17], a self-adaptive software is intro-
duced to modify the processing chain according to some predefined events such as sen-
sor loss. But this approach does not deal with high-level goals as temporal reasoning
which is crucial for monitoring evolving systems. In
ASGAARD
[32], the data abstrac-
tion module focusses on specific input sources according to a contextual plan which
can be adapted when the processed inputs contradict expected values. However, data
abstraction subtasks are scheduled in rigid sequences and diagnosis information is not
used for adapting the current system.
3
Application Domain: Monitoring Cardiac Patients
Several areas in medicine require monitoring and management, e.g. temporary paralysis
of the respiratory center in the brain, renal damage, surgical anesthesia or myocardial
infarction. In this latter domain, our concern, intensive coronary care units (ICCU) were
initiated in the late sixties, for monitoring the vital functions of a patient after a seri-
ous cardiac attack. Such intensive care units are still active nowadays and have the
same main goal i.e. to prevent, detect and control lethal arrhythmias and cardiac sudden
events by therapeutic actions. To this end, ECG and arterial pressure signals as well as
cardiac rhythm are analyzed in real time and displayed to an operator: the trends of the
main parameters, such as the cardiac frequency, are computed and alarms are gener-
ated. All this information assists the clinician operator who is in charge of analyzing
the situation, validating the alarms and deciding which action to perform.
3.1 Analysis of Cardiac Signals
The electrocardiogram (ECG, EKG) is a non invasive observation technique that can
provide may details about the cardiac condition of a patient. As such, it is widely used
for cadiac examination as well as in ICCU where it can be analyzed online. The electro-
cardiogram provides a graphical representation of the cardiac electrical activity which
triggers the contractions of the cardiac muscle. The electrical activity is measured by
placing sensors at specific locations of the patient body. An ECG is made of several
leads
. A lead displays the electrical signal extracted from the data recorded by one or
several sensors.
In normal conditions, the electrical signal, precisely the depolarization wave, propa-
gates from the top of the heart at the sinoatrial (SA) node, goes through the atria then
Search WWH ::
Custom Search