Bayesian Network-Based Model for the Diagnosis of Deterioration of Semantic Content Compatible with Alzheimer’s Disease - Foundations on Natural and Artificial Computation - page 419

discover new relations between the variables or new conditional probabilities, as
new cases appear. [10,2]
It has been demonstrated that there is a semantic deterioration in the abil-
ity to differentiate between living creatures (biological entities) and non-living
creatures (non-biological entities) in people suffering certain neurodegenerative
pathologies (Alzheimer, Semantic Dementia, Dementia with Lewy bodies, etc.),
traumatic pathologies (cranial traumatism), and infectious pathologies (herpes
encephalitis). Semantic categories are derived from classifications that are car-
ried out in the world around us and that treat essentially different objects as the
same. Thanks to the fact that our semantic memory is organised according to
these categories, we can perform a series of important cognitive functions, such
as inferring, establishing relations between examples, attributing properties to
objects that we do not know, reasoning, all of which is based on a cognitive
economy principle. People who suffer specific category deficits execute tasks af-
fecting totally or partially the category domain knowledge of living creatures
worse, whereas the object or artefact domain -non-living creatures- is totally or
almost totally conserved. There are also a small number of cases where the pat-
tern is the reverse; there is more deterioration in the object or artefact domain,
whereas the living creature domain is largely preserved [13].
Thus, the Bayesian Network model constructed in this article aims to diagnose
whether the patient suffers cognitive deterioration compatible with AD. This di-
agnosis is done from a corpus of oral definitions as a methodological tool that
has been shown to be very useful to study pathologies in relation to the semantic
deterioration of this disease [12]. It is a causal model based on literal definitions
of certain semantic categories -of the basic level of categorisation—both of living
creatures (dog, pine and apple), and non-living creatures (chair, car, trousers).
Patients do some tests where they have to define basic objects. When a pa-
tient suffers AD, he suffers from serious cognitive deterioration. The attributes,
features or characteristics generated by each patient's definitions are analysed.
The underlying logic for analysing the suggested features is in accordance with
a model described some time ago in Peraita, Elosúa and Linares (1992) and in
keeping with other current works ([3,8,9]).
In the causal model we represent that AD causes a conceptual-semantic-lexical
deficit and therefore the Bayesian Network will be able to infer the probability
of suffering AD from the degree of conceptual-semantic-lexical deficit. This in-
ference or abductive reasoning starts from some symptoms and searches for the
causes that best explain the symptoms. In other words, we start from conceptual-
semantic-lexical deterioration and search for the probability that this deteriora-
tion is the explanation for suffering AD. Some risk and protection factors, such
as educational level, age and sex, will also be taken into account in the Bayesian
Network.
Numerous works are currently being done in the field of Bayesian Networks
like Early Diagnosis of Alzheimer's Disease. Works on Explanation in Bayesian
Networks by [7][1] and other works of interest on Dynamic Bayesian Networks
and Learning in Dynamic Bayesian Networks [5] should be highlighted.