Elsevier

Neurocomputing

Volume 61, October 2004, Pages 99-119
Neurocomputing

Networks of spiking neurons in modeling and problem solving

https://doi.org/10.1016/j.neucom.2004.03.007Get rights and content

Abstract

In this paper, we describe the networks of spiking neurons and show their applications for modeling and problem solving. We have used integrate-and-fire neuron model that closely simulates a biological neuron's behavior. First, we model the somatosensory system with Hebbian-type spike-time-dependent plasticity and show the ability of the network to self-organize. Second, we apply a network of spiking neurons for identification, via clustering, of diabetic retinopathy images using temporal correlation learning rule. Results show that the network distinguishes the diabetic objects of interest from the image background.

Section snippets

Design of a network of spiking neurons

To define a network of spiking neurons we need to choose the spiking neuron model, the plasticity rule and the network topology. The first two components are described below.

Modeling of the somatosensory system

We based our somatosensory system model on the neurophysiological description of the somatosensory pathways that process the tactile inputs from the hand to the cerebral cortex in primates. The modeling described below is partially a repetition of one originally performed by Cios and Sala [5]. The difference is that the previously used learning rule did not take into account the LTD phenomenon and only the excitatory synapse learning was used. The current model takes into account both LTP and

Diabetic retinopathy data

Diabetic retinopathy (DR) is caused by diabetes and is one of the major causes of blindness. Elevated blood sugar in diabetics causes the blood vessels in the retina to leak, close, and proliferate, damaging the retina. Nearly all patients with diabetes will eventually develop DR but they will not know that they have it until the later stages of the disease [2]. If not properly treated, DR can cause blindness, but with appropriate treatment over 90% of visual loss can be prevented. Thus

Conclusions

The network of spiking neurons performed very well in modeling of the somatosensory system and clustering of the objects of interest in diabetic retinopathy images. The network with spike time-dependent plasticity learning rule for both excitatory and inhibitory synapses learning possesses self-organizing capabilities. It correctly mapped stimulation regions onto cortical area.

The DR image clustering is an original application of the network of spiking neurons to image segmentation. The

References (19)

There are more references available in the full text version of this article.

Cited by (0)

View full text