Butz Markus, Lehmann Konrad, Dammasch Ingolf E, Teuchert-Noodt Gertraud
Department for Neuroanatomy, University of Bielefeld, Germany.
Neural Netw. 2006 Dec;19(10):1490-505. doi: 10.1016/j.neunet.2006.07.007. Epub 2006 Oct 2.
We describe a strongly biologically motivated artificial neural network approach to model neurogenesis and synaptic turnover as it naturally occurs for example in the hippocampal dentate gyrus (DG) of the developing and adult mammalian and human brain. The results suggest that cell proliferation (CP) has not only a functional meaning for computational tasks and learning but is also relevant for maintaining homeostatic stability of the neural activity. Moderate rates of CP buffer disturbances in input activity more effectively than networks without or very high CP. Up to a critical mark an increase of CP enhances synaptogenesis which might be beneficial for learning. However, higher rates of CP are rather ineffective as they destabilize the network: high CP rates and a disturbing input activity effect a reduced cell survival. By these results the simulation model sheds light on the recurrent interdependence of structure and function in biological neural networks especially in hippocampal circuits and the interacting morphogenetic effects of neurogenesis and synaptogenesis.
我们描述了一种具有强烈生物学动机的人工神经网络方法，用于模拟神经发生和突触更新，这是自然发生的过程，例如在发育中和成年的哺乳动物及人类大脑的海马齿状回（DG）中。结果表明，细胞增殖（CP）不仅对计算任务和学习具有功能意义，而且对于维持神经活动的稳态稳定性也很重要。适度的CP速率比没有CP或CP非常高的网络更有效地缓冲输入活动中的干扰。在达到临界值之前，CP的增加会增强突触形成，这可能对学习有益。然而，较高的CP速率相当无效，因为它们会使网络不稳定：高CP速率和干扰性输入活动会导致细胞存活率降低。通过这些结果，模拟模型揭示了生物神经网络中结构与功能的循环相互依存关系，特别是在海马回路中，以及神经发生和突触形成的相互作用形态发生效应。
