structures and their general neuroanatomical features, in particular influence of lesions and
neuropathological changes on changes of normal behavior and cognitive performance.
It is not a priori clear that simplified neural models will be sufficient to capture such
casual relations (cf. [6] discussing hierarchical approach to modeling of brain functions at
different levels of complexity). Convergence of the modeling process could be too slow to
make them useful; for example some pathological effects could appear only in models based
on complex integrated-and-fire spiking neurons. Fortunately there are some indications that
the qualitative behavior of complex models based on spiking neurons [27] may also be ob-
tained in simplified neural models [21]. Thus there is a chance that simple neural models
may help to understand neurological and neuropsychological syndromes, providing some
insight into the source of the pathologies and understanding of the effects of therapeutic
procedures. A review article [23] and two books [18],[21] summarize results of such efforts
to understand memory and language impairments, psychiatric disorders, Alzheimer and
Parkinson disease, epilepsy and other neurological problems (see also [7]).
Brain simulations can complement traditional techniques in several ways. They pro-
vide insights into possible causal relations, allow for a full control of all aspects of experi-
ments, they are inexpensive and are not restricted by ethical considerations. Such simula-
tions are still in the initial stage of their development. They are usually based on oversimpli-
fied recurrent neural models with two-state neurons or feedforward models wit graded sig-
moidal-response neurons. More biologically faithful networks based on simple spiking neu-
rons should soon be introduced in such simulations (cf. [1], [7]).
In this article several therapeutic suggestions resulting from computational models of
Alzheimer disease are made. Although several computational models of this disease have
been published in the last decade few conclusions were drawn. Two such models are intro-
duced here and directions for their further extensions are outlined. Assuming that computa-
tional models reflect real neural mechanisms leads to some therapeutic suggestions that
should slow down the degeneration of synaptic connections and thus the development of the
disease, at least in its early stages.
2. Alzheimer disease
Alzheimer disease (AD) is the most common form of dementia gradually leading to a
global cognitive dysfunction and death. The earliest symptoms involve memory degrada-
tion, both for learning new things and recalling known facts. This is followed by degrada-
tion of language skills, poverty of thoughts and associations, intellectual rigidity, loss of
initiative and interest, disturbances in motor and executive functions. In advanced stages
judgments are impaired, psychotic features may appear (such as paranoid delusions), and