Figure 2. ACE detects neurons efficiently with high precision and records neural activity stably in deep brain tissue.

(a) Example current trace recorded showing automatic neuron detection by ACE in vivo. The pipette was advanced into the brain at 2 μm step⁻¹ sec⁻¹. (a1) At each step, a train of ten −5 mV test pulses was delivered to the pipette and the current responses were measured to calculate the average pipette resistance (R_p). R_p was then compared with values calculated at previous steps to see whether the neuron detection criteria would be met. (a2) In this penetration detection occurred within 4–6 μm. A threshould of 10% R_p increase was used. Initial drop of R_p at 540 μm indicated a rejection event (most likely the pipette passed a dendrite or capillary). (b) ACE achieves stable recording. The voltage trace (high-pass filtered at 2 Hz) of cell-attached recording of neural responses is shown for the same neuron detected in a. Spontaneous activity and visually evoked responses were both recorded. Orientation-tuning curve was mapped in this neuron with eight orientation, five spatial frequencies and one temporal frequency (Methods section). (b1) Normal spontaneous activity after about 18 min of recording indicated no apparent damage to the cytoplasma membrane integrity. (b2) Mild current injection (−100 nA) evoked rapid and reversible increase of firing. Also see Fig. 4 for detection of a spontaneously inactive neuron verified by current injections. (b3) Spike shape (left) and orientation-tuning curves (middle and right) of the same neuron. Only responses to the preferred spatial frequency (0.04 cycle per degree) were shown. (c) Our detection–verification algorithm greatly improves the performance of neuron detections. Data were compared for detection trials in which neuron detection criteria were met (that is, the detection precision), and all trials, with or without current injection to verify the detection. (d) The verification by current injection had no effect on the average trial time.