Figure 2.

Approaches to connecting the brain to itself. (A) In spike-triggered microsimulation, the detection of an action potential generated by a neuron recorded by an implanted microelectrode triggers focal microstimulation at microelectrodes located in another area of cortex (Jackson et al., 2006). (B) Rather than triggering stimulation in a direct, linear manner, the relationship between the spike detection and microstimulation could be governed by a weighting rule meant to emulate features of synaptic plasticity. (C) The external system could be altered to incorporate spike trains from two or more neurons and use spike-time dependent synaptic plasticity learning rules, including temporal coincidence detection, in order to trigger target. (D) As the BCI were able to record and stimulate more individual neurons, the external system could deploy learning rules to update the synaptic weight between all potential pairs, shown here on an image based on William James' hypothesis about neural process interaction (James, 1890). (E) Beyond direct mapping of recording to stimulation, or an interposed simplified neural network, more realistic cortical simulations could be used with the premise that the emulated cortex would be capable of performing its own “canonical” operations and would hence add a computation beyond connectivity (Nelson, 2002; Kouh and Poggio, 2008; Miller, 2016). Figure reprinted from Serruya and Kahana (2008) with permission from Elsevier.