Information Technology Reference
In-Depth Information
reasons. Fuzzy similarity matching is incorporated into the CBR system to better
accommodate the underlying uncertainty existing in the reasoning process.
1.1 Decision Support System and Case-Based Stress Diagnosis
The term Decision Support System (DSS) is defined by Little as a “model-based set
of procedures for processing data and judgments to assist a manager in his decision
making” [9]. Medical Decision Support Systems (DSS) have been defined by many
people in many different ways. According to Shortliffe a medical DSS is “any com-
puter program designed to help health professionals make clinical decisions [10].”
The early AI systems for medical decision making emerged around the 1950's mainly
built using decision trees or truth tables. After that, different methods or algorithms
have been introduced to implement medical decision support systems such as, Bayes-
ian statistics, decision-analytical models, symbolic reasoning, neural-networks, rule-
based reasoning, fuzzy logic, case-based reasoning etc.
Since the implementation of MYCIN [11] many of the early AI systems attempted
to apply rule-based systems in developing computer based diagnosis systems.
However, for a broad and complex medical domain the effort of applying rule-based
system has encountered several problems. Some of the preliminary criteria for im-
plementing a rule-based system are that the problem domain should be well under-
stood, constant over time and the domain theory should be strong i.e. well defined
[12]. In psychophysiology, the diagnosis of stress is so difficult that even an experi-
enced clinician might have difficulty in expressing his knowledge explicitly. Large
individual variations and the absence of general rules make it difficult to diagnose
stress and predict the risk of stress-related health problems. On the other hand, case-
based reasoning (CBR) works well in such domains where the domain knowledge is
not clear enough i.e. weak domain theory. Furthermore, CBR systems can learn
automatically which is very important as the medical domain is evolving with time.
Rule-based systems cannot learn automatically as new rules are usually inserted
manually. Statistical techniques are also applied successfully in medical systems. But
to apply statistical models a large amount of data is required to investigate a hypothe-
sis which is also not available in our application domain.
Several motivations for applying CBR to stress diagnosis can be identified:
-
CBR [12] [13] is inspired by the way humans reason i.e. to solve a new prob-
lem by applying previous experiences. This reasoning process is also medi-
cally accepted and the experts in diagnosing stress too rely heavily on their
past memory to solve a new case. This is the prime reason why we prefer to
use CBR.
-
Knowledge elicitation is another problem in diagnosing stress, as human be-
haviour or our response to stress is not always predictable. Even an experi-
enced clinician in this domain might have difficulty to articulate his knowledge
explicitly. Sometimes experts make assumptions and predictions based on ex-
periences or old cases. To overcome this knowledge elicitation bottleneck we
use CBR because in CBR, this elicitation can be performed with the previous
cases in the case base.
Search WWH ::
Custom Search