Abstract
Alzheimer’s disease (AD) is a chronic progressive neurodegenerative disorder afflicting millions worldwide. Abnormal N400 Event-Related Potentials (ERP) are biomarkers indicative of AD progression [1]. Specifically, in the semantic category judgment task, N400 congruency and repetition effects diminish with AD progression in Mild Cognitive Impairment Patients. Aberrant neuronal properties in AD such as Calcium (Ca2+) concentrations and N-methyl-D-aspartate (NMDA) receptor dysfunction could be the underlying cause of these ERP abnormalities [2, 3]. However, there is no consensus in the literature on the cognitive functions or specific neural generators of the N400 nor detailed neuronal models that account for these factors.
Here we propose, to our knowledge, the first biologically detailed and plausible connectionist spiking neural network architecture to model the semantic category judgment task. The architecture’s neuronal characteristics are based on the spiking Selection over Time and Space (sSoTS) model that encompasses gamma aminobutyric acid (GABA), alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), NMDA, and the spike frequency adaptation currents [4]. AMPA&GABA account for fast excitatory and inhibitor currents, respectively. NMDA accounts for slow magnesium ion dependent currents. The spike frequency adaptation current is a Ca2+ ion dependent after-hyperpolarization mechanism. The architecture has groups of neurons divided into pools arranged in layers as seen in Fig. 1. Each layer represents a specific feature type i.e. auditory, visual, or semantic. Pools within a layer represent stimulus properties. The connectivity between and within pools and layers is approximated by converting the architecture into a population coded model using the Mean Field Approach, permitting the exploration of a large parameter space.
Here we propose, to our knowledge, the first biologically detailed and plausible connectionist spiking neural network architecture to model the semantic category judgment task. The architecture’s neuronal characteristics are based on the spiking Selection over Time and Space (sSoTS) model that encompasses gamma aminobutyric acid (GABA), alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA), NMDA, and the spike frequency adaptation currents [4]. AMPA&GABA account for fast excitatory and inhibitor currents, respectively. NMDA accounts for slow magnesium ion dependent currents. The spike frequency adaptation current is a Ca2+ ion dependent after-hyperpolarization mechanism. The architecture has groups of neurons divided into pools arranged in layers as seen in Fig. 1. Each layer represents a specific feature type i.e. auditory, visual, or semantic. Pools within a layer represent stimulus properties. The connectivity between and within pools and layers is approximated by converting the architecture into a population coded model using the Mean Field Approach, permitting the exploration of a large parameter space.
Original language | English |
---|---|
Pages (from-to) | 111 |
Journal | BMC Neuroscience |
Volume | 20 |
DOIs | |
Publication status | Published (VoR) - 2019 |