We present an algorithm to identify individual neural spikes observed on high-density multi-electrode arrays (MEAs). Our method can distinguish large numbers of distinct neural units, even when spikes overlap, and accounts for intrinsic variability of spikes from each unit. As MEAs grow larger, it is important to find spike-identification methods that are scalable, that is, the computational cost of spike fitting should scale well with the number of units observed. Our algorithm accomplishes this goal, and is fast, because it exploits the spatial locality of each unit and the basic biophysics of extracellular signal propagation. Human interaction plays a key role in our method; but effort is minimized and streamlined via a graphical interface. We illustrate our method on data from guinea pig retinal ganglion cells and document its performance on simulated data consisting of spikes added to experimentally measured background noise. We present several tests demonstrating that the algorithm is highly accurate: it exhibits low error rates on fits to synthetic data, low refractory violation rates, good receptive field coverage, and consistency across users.

To explore theories of predictive coding, we presented mice with repeated sequences of images with novel images sparsely substituted. Under these conditions, mice could be rapidly trained to lick in response to a novel image, demonstrating a high level of performance on the first day of testing. Using 2-photon calcium imaging to record from layer 2/3 neurons in the primary visual cortex, we found that novel images evoked excess activity in the majority of neurons. When a new stimulus sequence was repeatedly presented, a majority of neurons had similarly elevated activity for the first few presentations, which then decayed to almost zero activity. The decay time of these transient responses was not fixed, but instead scaled with the length of the stimulus sequence. However, at the same time, we also found a small fraction of the neurons within the population (~2%) that continued to respond strongly and periodically to the repeated stimulus. Decoding analysis demonstrated that both the transient and sustained responses encoded information about stimulus identity. We conclude that the layer 2/3 population uses a two-channel predictive code: a dense transient code for novel stimuli and a sparse sustained code for familiar stimuli. These results extend and unify existing theories about the nature of predictive neural codes.

To explore how neural circuits represent novel versus familiar inputs, we presented mice with repeated sets of images with novel images sparsely substituted. Using two-photon calcium imaging to record from layer 2/3 neurons in the mouse primary visual cortex, we found that novel images evoked excess activity in the majority of neurons. This novelty response rapidly emerged, arising with a time constant of 2.6 ± 0.9 s. When a new image set was repeatedly presented, a majority of neurons had similarly elevated activity for the first few presentations, which decayed to steady state with a time constant of 1.4 ± 0.4 s. When we increased the number of images in the set, the novelty response’s amplitude decreased, defining a capacity to store ∼15 familiar images under our conditions. These results could be explained quantitatively using an adaptive subunit model in which presynaptic neurons have individual tuning and gain control. This result shows that local neural circuits can create different representations for novel versus familiar inputs using generic, widely available mechanisms.

Redundancies and correlations in the responses of sensory neurons may seem to waste neural resources, but they can also carry cues about structured stimuli and may help the brain to correct for response errors. To investigate the effect of stimulus structure on redundancy in retina, we measured simultaneous responses from populations of retinal ganglion cells presented with natural and artificial stimuli that varied greatly in correlation structure; these stimuli and recordings are publicly available online. Responding to spatio-temporally structured stimuli such as natural movies, pairs of ganglion cells were modestly more correlated than in response to white noise checkerboards, but they were much less correlated than predicted by a non-adapting functional model of retinal response. Meanwhile, responding to stimuli with purely spatial correlations, pairs of ganglion cells showed increased correlations consistent with a static, non-adapting receptive field and nonlinearity. We found that in response to spatio-temporally correlated stimuli, ganglion cells had faster temporal kernels and tended to have stronger surrounds. These properties of individual cells, along with gain changes that opposed changes in effective contrast at the ganglion cell input, largely explained the pattern of pairwise correlations across stimuli where receptive field measurements were possible.

Neurons perform computations by integrating excitatory and inhibitory synaptic inputs. Yet, it is rarely understood what computation is being performed, or how much excitation or inhibition this computation requires. Here we present evidence for a neuronal computation that maximizes the signal-to-noise power ratio (SNR). We recorded from OFF delta retinal ganglion cells in the guinea pig retina and monitored synaptic currents that were evoked by visual stimulation (flashing dark spots). These synaptic currents were mediated by a decrease in an outward current from inhibitory synapses (disinhibition) combined with an increase in an inward current from excitatory synapses. We found that the SNR of combined excitatory and disinhibitory currents was voltage sensitive, peaking at membrane potentials near resting potential. At the membrane potential for maximal SNR, the amplitude of each current, either excitatory or disinhibitory, was proportional to its SNR. Such proportionate scaling is the theoretically best strategy for combining excitatory and disinhibitory currents to maximize the SNR of their combined current. Moreover, as spot size or contrast changed, the amplitudes of excitatory and disinhibitory currents also changed but remained in proportion to their SNRs, indicating a dynamic rebalancing of excitatory and inhibitory currents to maximize SNR.