biofeedback is aimed at reducing the average EEG frequency [4], or achieving the ‘alpha
relaxation state’. Such mental activities as mantra repetition, chanting, visualization or con-
templative absorption should lower the brain temperature, stopping the background thoughts
and other processes that may lead to the synaptic runaway. Therefore in the early stages of
AD it may be worthwhile to experiment with various relaxation techniques to slow down
the development of the disease.
More detailed therapeutic suggestions related to an optimal compensation require more
complex associative memory models. The existing models should be extended in several
directions. Human memory involves interactions between hipocampal formation, neocortex
and neuromodulatory systems, regulating plasticity of synapses depending on the emotional
contents of the situation [2],[15]-[17]. Although such models have been created few years
ago computational simulations have started only quite recently (Murre, private information).
More realistic memory models that would allow studying the influence of different neuro-
transmitters on the inter-module inhibition and between-module excitation should help to
evaluate potential benefits of new drugs. Models based on simplified spiking neurons [13]
are needed to make direct connections with neurophysiology. Many associative memory
models based on simplified models of spiking neurons have been created recently and
should be used in a near future to study the Alzheimer disease and other memory-related
diseases.
6. Conclusions
Small number of assumptions and simple neural models allow for qualitative under-
standing of experimental observations in case of Alzheimer disease and many other neuro-
logical and psychiatric disorders. Although therapeutic suggestions drawn here from AD
models are speculative they may easily be tested. Computer simulations appeared only quite
recently as tools for modeling real brain processes. In view of the great complexity of the
brain and lack of detailed understanding of its functions skepticism towards such models
may seem to be justified. There are many fundamental problems related to the convergence
of computational models, hypothesis on which they based, selection of minimal neural
models that capture relevant phenomena and are still amenable to computer simulations.
Surprisingly, even very simple neural models of associative memory show a number of fea-
tures that reflect many properties of real biological memories known from cognitive psy-
chology [14]. The neural modeling process may not be so difficult after all.
Another interesting - and perhaps easier - area of neural modeling concerns reorgani-
zation processes following focal damages of neocortex (stroke, lesions) and damages to af-
ferent pathways (amputation of limbs). Some therapeutic suggestions may be offered for
faster recovery of sensorimotor competence after stroke [22], reduction of pain in phantom