Information Technology Reference
In-Depth Information
2.2 Mathematical Transforms in Medical CBR
Case-based retrieval is recently being proposed as a well suited decision support
methodology in time dependent medical domains, where (some) case features are
in the form of time series; in these works, time series dimensionality reduction is
often performed relying on mathematical transforms. Two significant examples
of mathematical transforms exploitation in medical CBR are illustrated below.
DFT in haemodialysis. Dimensionality reduction in medical CBR by resort-
ing to DFT has been proposed in [33], where the application domain is the one
of haemodialysis. Haemodialysis is the most widely used treatment for severe
chronic renal diseases, and relies on a device, called haemodialyzer, which clears
the patient's blood from catabolites, to re-establish acid-base equilibrium and to
remove water in excess. During each single haemodialysis treatment (session),
which lasts 3-4 hours on average, the haemodialyzer collects several variables,
most of which are in the form of time series (while a few are “static”, i.e. single-
valued). Considering a case as a haemodialysis session, the system in [33] re-
trieves past cases, belonging to the same patient or to different ones, in which
the collected variables had a similar behaviour: in this way the physician can
derive indications about the patient at hand's current metabolic condition, or
static features
k-NN
class C
class 1
class 2
dynamic feature 1
range query
range query
range query
dynamic feature 2
range query
range query
range query
dynamic feature n
range query
range query
range query
case intersection
case intersection
case intersection
weighted average
weighted average
weighted average
best cases
best cases
best cases
Fig. 1. The retrieval process in [33] (the picture is taken from [33]).
Search WWH ::




Custom Search