Providing Case-Based Retrieval as a Decision Support Strategy in Time Dependent Medical Domains - Computational Intelligence in Healthcare: Advanced Methodologies - page 213

D S D W

2h

step 1

2h

D D

step 2

4h

D

Fig. 7. Generalization steps for the HV query

tree shown in figure 6) as a support for a quick query answering. Matching
the steps in the generalization process to the nodes in the index structure (in
the time direction), we can descend through the node DD .Now,wecanmove
“horizontally” in the symbol taxonomy direction, to reach the node D S D W ,
which matches exactly our query. As a result, we can retrieve all the cases indexed
by such a node.
We may want to generalize the required behaviour, in order to retrieve a
larger number of cases, maybe not perfectly matching the positive HV model.
Interactive and progressive query relaxation and refinement are supported in our
framework. For instance, we can allow any combination of decreasing episodes
in the session. This can be obtained by relaxing the query in the direction of the
symbols, using e.g. the sequence DD in a 2 hours granularity. Query relaxation
(as well as refinement) can be repeated several times, until the user is satisfied
with the obtained results.
Finally, the user may want to retrieve a generalized case, i.e. to stop the search
at a proper internal node in the index structure. This node subsumes a set of
cases, but the user may just be interested in calculating the distance between
the query and the node, which summarizes the retrieval set, without entering
the details of all the elements composing it. This functionality is available too
in our framework.
The interested reader may find additional technical details about the frame-
work in [31].

4 Conclusions

Time series retrieval is a critical issue in all medical domains in which the ob-
served phenomenon dynamics have to be dealt with.
Case-based Reasoning (CBR), and more specifically case-based retrieval, is
recently being recognized as a valuable decision support methodology in these
domains, as testified by the growing number of works in the field.
In this chapter, we have analyzed in depth the methodologies used to imple-
ment case-based retrieval as a decision support strategy in these applications,
dealing with the complexity of case representation and retrieval when most fea-
tures are in the form of time series.
Classical techniques as well as more innovative ones have been presented.