Information Technology Reference
In-Depth Information
guidelines integration, feature selection, distributed cases, neural networks com-
bination, and planning combination. Thirty-eight different types of system design
were identified, most of them dealing with some data mining / knowledge discov-
ery combination, and as seen above temporal abstraction / temporal reasoning.
However two categories are specific to medical domains: clinical guidelines inte-
gration, and electronic medical records integration.
Research themes. So far, the research themes are coded in free text format.
More work remains to be done to categorize this part of each paper. One pos-
sibility to simplify the problem is to select one particular research focus among
the dimensions of the paper: domain, purpose, memory, reasoning, and system
design. With this simplification, the research themes are in majority oriented
toward design (61 papers), purpose (38 papers), including 11 survey papers, and
reasoning (14 papers). It is not surprising that no paper actually focuses ex-
clusively on the application domain - this kind of publication would be better
suited for a medical journal in a medical specialty. However it is surprising that
so many papers focus on design aspects most of the time dealing with some
combination with another methodology of AI (data mining, knowledge discov-
ery, temporal reasoning and abstraction, knowledge based systems) or beyond,
such as information retrieval and databases. Only 14 papers focus on some rea-
soning step - in practice only adaptation and retrieval - using methods intrinsic
to CBR. The group of papers focusing on the paper purpose regroup such large
subgroups as decision-support capability, classification, image understanding,
survey papers, papers exploring different roles of CBR, and research support.
The memory dimension is almost always connected with either some design as-
pect such as prototype learning, some reasoning aspect such as using prototypes
to guide retrieval, or some purpose aspect, such as decision support. Indeed the
memory needs to prove useful to the system in some way, since these papers are
applied.
7 CBR versus AI in the Health Sciences
This section explores the research theme of synergies between CBR and other AI
approaches. First is a look at the impact of the health sciences on AI and of AI on
the health sciences. This is followed by a discussion of synergies with data mining
and knowledge discovery and a look at multimodal reasoning architectures.
7.1 AI in the Health Sciences
History. Since the early days of AI, the health sciences have been a favorite area
of application. The earliest systems were decision-support systems such as IN-
TERNIST in 1970 [68] and MYCIN in 1976 [69]. INTERNIST was a rule-based
expert system for the diagnosis of complex diseases. It was later commercial-
ized as Quick Medical Reference (QMR) to support internists with diagnosis.
MYCIN was also a rule-based expert system, for the diagnosis and treatment of
 
Search WWH ::




Custom Search