|
In the literature, most single neuron threshold models account for spike-frequency adaptation by means of a single exponential spike-triggered current. However, it is well known that, in real neurons, adaptation occurs on multiple timescales ranging from tens of milliseconds to seconds. Here, we propose a Leaky Integrate-and-Fire model with escape noise extended with two adaptive mechanisms: each time a spike is fired, both an adaptive current and a change in the threshold are triggered. Importantly, the functional shape of these two processes are not imposed a priori but are directly extracted from the data. The parameters of our model are fitted on experimental data recorded in-vitro from rat neocortical pyramidal neurons under sine-modulated noise stimulation. Our model is able to predict up to 70% of the observed spikes with an accuracy of 2 ms. Importantly, the response predicted by our spiking model in terms of spike rate is in close agreement with the observed data. Our model reveals that the core mechanisms at the basis of multiple-timescale scale-invariant adaptation is a power-law spike-triggered current that lasts for more then 20 s. Furthermore, the kernel of the dynamic threshold is crucial to capture the observed refractory properties of the neuron. |
|