An Artificial Neural Network Model of the Milk-Ejection Reflex in the Sheep

Milk ejection reflex involves an ascending neuronal pathway from the nipples to the hypothalamus and a descending vascular limb. Anatomical data reveal that in sheep, ascending neural information passes through the medial cuneate nucleus (CUm), or the lumbar spinal segments (LSS) and the lateral cervical nucleus (LCN) before reaching the hypothalamic paraventricular nucleus (PVN). Physiological measurements suggest that although the nipples are periodically stimulated during suckling, the hypothalamic hormone oxytocin, is only released in the bloodstream once or twice. In a first attempt to develop an artificial neural network of the above pathway, we model a typical relay station, namely the CUm, LSS or the LCN, as a sparsely connected network consisting of excitatory and inhibitory populations of leaky-integrate-and-fire (LIF) neurons. The network is characterized by its connectivity, the average number of connections for each neuron, and a parameter γ determining the ratio of excitation versus inhibition. Excitatory connections have modulated synaptic weights; strengthen upon neuronal firing and slowly decaying otherwise. The network responds to periodic stimulation with one or two bursts of highly synchronised neuronal activation and then enters a chaotic regime where naturally synchronization disappears. Such a relay station could provide the appropriate input for oxytocin release.