Bayesian Network-Based Model for the Diagnosis of Deterioration of Semantic Content Compatible with Alzheimer’s Disease - Foundations on Natural and Artificial Computation - page 422

- Variables of interest or assumption. We calculate their a posteriori proba-

bility from the findings. These variables are: Suffers Cognitive Deterioration
(Dementia) and Suffers Alzheimer's Disease.

Discrete Bayesian Network Modelling
Alzheimer and Cognitive deterioration variables of interest have some risk fac-
tors represented by the variables Educational Level, Age and Sex. It is worth
highlighting that they are risk factors and not causes of the disease, for that rea-
son canonical models (OR/MAX) cannot be used. Furthermore, there is a causal
link between “Alzheimer” and “Cognitive Deterioration”. According to scientific
literature and epidemiological studies, the most common cause of dementia in
the European Union is Alzheimer (around 50-70% of cases), other causes of de-
mentia are: multiple cardiac arrest (around 30% of cases), Pick's disease, Lewy
bodies and others.
Intermediate variables representing conceptual-semantic-lexical deterioration
in the category domain are automatically treated, analysing and interpreting the
patient's definitions, distributed in the 11 basic conceptual blocks considered as
conceptual components underlying every organisation and representation of ob-
ject categories (taxonomical, types, parts, functional, evaluative, place/habitat,
behaviour, causes/generates, procedural, life cycle and others). Each of these
blocks has an identifying lexical label, as indicated in the work by Dr. Herminia
Peraita [12,14].
If a patient suffers AD in prodromic or incipient stage, there is a differen-
tial deterioration between the semantic categories Living Creatures and Non-
Living Creatures. According to Dr. Herminia Peraita's study[12,14], Alzheimer's
disease produces cognitive deterioration in Living Creatures before Non-Living
Creatures or artefacts. Since AD patients usually have greater damage in the
temporal limb areas in the early stages of the disease, they could show selective
deterioration for living creatures. As the disease progresses, the damage becomes
so omnipresent that mistakes occur in both domains with the same frequency.
Therefore, there is a causal relation that qualitative and quantitative models can
take into account in this study.
Another important factor that is modelled in the Bayesian Network is the bi-
directional correlations between intermediate variables and variables of interest.
For example, when a patient produces few attributes in the semantic category
of Living Creatures, the probability increases of producing few attributes in
the semantic category of Non-Living Creatures when the disease is advanced.
Similarly, when the disease is incipient and is Dementia caused by AD, the pro-
duction of few attributes in the semantic category of Living Creatures increases
the probability of producing a greater number of attributes in the semantic cate-
gory of Non-Living Creatures. In other words, there is a negative causal relation
between the two semantic categories when the disease is incipient and a positive
correlation when the disease is advanced. We should remember that Bayesian
networks do not include cycles. With the intermediate variables it will therefore
be possible to represent these circumstances. In fig.1 and fig.2, the modelling
techniques used to eliminate the cycles of de bayesian red can be observed.