スパイク時刻依存可塑性則を組み込んだ培養神経回路網の低遅延な結合強度推定

抄録

Estimating current states of neuronal networks from multi-neuron measurement helps a brain-machine interface to replace lost brain functions. Estimating brain states based on a physiological model has a huge advantage on estimation accuracy and directness to biophysical phenomenon. In this study, we propose a model-based method for accurate synaptic connection estimation with short time delay. The method merges our previous method and Song's spike-timing-dependent plasticity rule using Kalman filter to utilize past estimation data by assuming prediction rules. The proposed method achieved an accurate estimation by 5 s of recorded data, while previous method required > 20 s. Then, it was examined if the method can detect stimulation-induced plasticity in living neuronal network, using rat cortex neurons cultured on a microelectrode array. The proposed method detected synaptic plasticity immediately after induction. These results suggest that our method is suitable for estimating synaptic connection strength accurately with a short time delay.