Explaining Medical Model Exceptions - Computational Intelligence in Healthcare: Advanced Methodologies - page 258

three months, then after six months and finally after a year. Unfortunately, since some
measurements did not happen, many data are missing. Therefore the patient records
often contain different sets of measured parameters.
It is necessary to note that parameter values of dialysis patients essentially differ
from those of non-dialysis patients, especially of healthy people, because dialysis in-
terferes with the natural, physiological processes in an organism. In fact, for dialysis
patients, all physiological processes behave abnormally. Therefore, the correlation
between parameters differs too.
For statistics, this means difficulties in applying statistical methods based on corre-
lation and it limits the usage of a knowledge base developed for normal people. Non-
homogeneity of observed data, many missing data, many parameters for a relatively
small sample size, all this makes the data set practically impossible for usual statisti-
cal analysis.
Since the data set is incomplete, additional or substitutional information has to be
found in other available data sources. These are databases, namely the already exis-
tent individual base and the sequentially created case base, and the medical expert as a
special source of information.
2.1 Setting up a Model
Usually, a medical problem needs to be solved based on given data. In this example
it is:
"Does a specific fitness programs improve the physiological condition of dialysis pa-
tients?"
So, the physical conditions of active and non-active patients need to be compared.
Patients are divided into two groups, depending on their activity, active and non-
active patients.
According to the assumption, active patients should feel better after some months
of participating in the fitness program, whereas non-active ones should feel rather
worse. The meaning of “feeling better” and “feeling worse” has to be defined in this
context. Therefore, a medical expert selects appropriate factors from ISOR's menu,
which contains the parameter names from the observed database. The expert selects
the following main factors
- F1: O2PT - Oxygen pulse by training
- F2: MUO2T - Maximal Uptake of Oxygen by training
- F3: WorkJ - performed Work (Joules) during control training
Subsequently the “ research time period” has to be determined. Initially, this period
was planned to be twelve months, but after a while the patients tend to give up the
fitness program. This means, the longer the time period, the more data are missing.
Therefore, a compromise between time period and sample size had to be made; a pe-
riod of six months was chosen.
The next question is whether the model should be quantitative or qualitative? The
observed data are mostly quantitative measurements. The selected factors are also