Electrophysiology analysis

openreadout analyze ephys-features measures patch-clamp recordings (action potentials, rheobase, input resistance and membrane-test values), and openreadout analyze spikes detects spikes in extracellular recordings. The MCP tool openreadout_analyze runs the same analyses (kinds ephys-features and spikes). Both work on top of the electrophysiology readers (ABF, ATF, NWB, Neuralynx, Blackrock, SpikeGLX, Intan, Plexon). This page describes how each value is computed and lists the analysis vocabulary.

Code: crates/openreadout-signal/src/ephys/. Provenance: docs/provenance/ephys-analysis.md.

analyze ephys-features (patch clamp)

Channel and mode. Default channel: the first whose unit is a voltage (mV, V, µV), else the first current channel (pA, nA, µA, A). Voltage channel + current command = current_clamp; current channel + voltage command = voltage_clamp; same kind for both = unknown (analysed as recorded).

Stimulus. From the ABF epoch table (info → traces[].extra.outputs): the first output with its waveform enabled and epochs. Sweep timing: the first floor(samples/64) samples hold, then the epochs follow, epoch e lasting duration + sweep·duration_step samples at level + sweep·level_step (pyABF’s documented convention). The stimulus epoch is the first step epoch whose level changes between sweeps, else the first step that differs from holding. Recordings without an epoch table (gap-free, other formats) get spike and sweep features only.

Spikes (per sweep, whole sweep): an upward crossing of --peak-threshold (−20 mV) starts a spike, the next downward crossing ends it, the peak is the maximum in between (eFEL’s Spikecount definition). Onset (threshold): first sample after the previous spike’s AHP minimum from which dV/dt ≥ --dvdt-threshold (10 V/s) for 5 samples. Per spike: peak_time_s, peak_mv, threshold_time_s, threshold_mv, amplitude_mv (peak − onset), half_width_ms (at the voltage half way between onset and peak), rise_time_ms (onset → peak), decay_time_ms (peak → back to the onset voltage), ahp_mv (minimum to the next spike’s crossing or the sweep end), ahp_depth_mv (onset − AHP), upstroke_v_per_s, downstroke_v_per_s, isi_ms.

Per sweep: spike_count, spike_count_stimulus (peaks inside the stimulus), firing_rate_hz (= in-stimulus count / stimulus duration), first_spike_latency_ms (first peak − stimulus onset), isi_mean_ms, isi_cv, adaptation_index = mean of (ISIₙ₊₁ − ISIₙ)/(ISIₙ₊₁ + ISIₙ) over the in-stimulus ISIs, half_width_mean_ms, amplitude_mean_mv; passive (baseline_mv = mean of the last 10 % before the stimulus, steady_state_mv = mean of the last 10 % of the stimulus, extreme_mv, input_resistance_mohm = (steady − baseline)/ΔI, tau_ms (single exponential fitted from the onset to the voltage extreme), sag_ratio = (steady − min)/(baseline − min) (eFEL sag_ratio1), steady_fraction = (baseline − steady)/(baseline − min) (eFEL sag_ratio2)); spikes inside the stimulus void the passive measures of that sweep.

Per cell: rheobase_pa (smallest positive step with ≥ 1 in-stimulus spike) and rheobase_sweep; fi_curve; fi_slope_hz_per_pa (least squares over firing steps); max_firing_rate_hz; input_resistance_mohm (slope of steady − baseline against the step over spike-free negative steps); tau_ms (median); capacitance_pf = τ/R_in; sag_ratio (the most negative step); resting_mv (median baseline).

Voltage clamp (test_pulse, per sweep): holding_current_pa (baseline current), peak_current_pa (extreme in the first 2 ms of the step − baseline), steady_current_pa, total_resistance_mohm = ΔV/steady, tau_ms (single exponential decaying to the measured steady state, least squares, fitted from where the transient has fallen to 90 % of its height to where it first reaches the steady state, at most 50 ms or half the step), access_resistance_mohm = ΔV/I0 with I0 the fit extrapolated back to the transient’s apex, membrane_resistance_mohm = total − access, capacitance_pf = Q/ΔV·((Ra + Rm)/Rm)² with Q the charge of the transient above the steady state from the step onset to its first return to the steady state (exact for one compartment behind a series resistance); per cell the medians.

--csv sweeps|spikes|fi prints one tidy table instead of the JSON report.

analyze spikes (extracellular)

Per channel and sweep: Butterworth band-pass (--band 300:6000, --order 5; the high edge is clamped to 0.45 × the sampling rate), zero phase (forward-backward with odd-reflection padding and steady-state initial conditions); noise = median(|x|)/0.6745 of the filtered sweep; a spike is a sample beyond --threshold (5) × noise in the --sign direction (neg) that is the extreme within --exclude-ms (0.1 ms) on both sides. Output: spike_count, rate_hz, noise, threshold, duration_s and up to --max-times spike times per channel.

Validation (automated in crates/openreadout-corpus-tests/tests/ephys_analysis.rs)

Reference readers: oracle/signal/ephys.py → corpus/oracle/ephys-analysis/*.json (eFEL 5.7 as a black box, pyABF memtest, Neo + SpikeInterface 0.105). Results:

data	against	result
pyabf-171116sh-0018, -2019-07-24-0055-fsi, -190619b-0003 (current-clamp steps; 44 sweeps, 1182 spikes)	eFEL	spike counts per sweep and inside the stimulus identical in 44/44 sweeps; stimulus windows and step currents identical; rheobase identical (50, 25, 120 pA); input resistance within 0.03 % (104.16 vs 104.14, 176.82 vs 176.77, 224.01 vs 223.99 MΩ); sag ratio within 0.0002; baseline within 0.1 mV, steady state within 0.03 mV; 1160/1182 peaks at the same or the neighbouring sample (differences on flat-topped peaks); median onset voltage difference 0.5–1.0 mV (different onset rules); median half-width difference 3 %; time constant median difference 10–17 % (different fit windows)
pyabf-18808025-memtest, pyabf-171116sh-0011 (voltage-clamp membrane tests)	pyABF memtest	holding current within 0.2 pA; ΔV/ΔI steady-state resistance within 4 % (1922.6 vs 2002.2, 95.9 vs 97.1 MΩ); access resistance within 8 % (28.7 vs 26.8, 18.3 vs 16.9 MΩ); capacitance 24.0 vs 18.8 and 237.6 vs 150.8 pF — pyABF uses τ/Ra, which ignores Rm and the membrane current; ours integrates the transient’s charge (exact on a synthetic RC cell, unit-tested)
brk-filespec2-3001.ns5 (10 ch), intan-test-tetrode-163225.rhd (4 ch), nlx-cheetah-v5-5-1-tet3a.ncs	SpikeInterface bandpass + detect_peaks	spike counts identical on all 15 channels (3,772 spikes), every SpikeInterface spike matched within 0.1 ms, noise equal to 4 digits

Performance (release build): analyze ephys-features on pyabf-2019-07-24-0055-fsi (17 sweeps × 60,000 samples, 948 spikes) 0.02 s, 12 MB; analyze spikes on brk-filespec2-3001.ns5 (10 channels × 900,300 samples at 30 kHz) 0.47 s, 168 MB; on intan-rhd-test-1.rhd (192 channels × 30,000 samples) 0.30 s, 91 MB. Memory is one sweep of all channels plus one filtered channel; --max-seconds bounds it on long continuous recordings.

Known gaps

• Stimulus windows come from ABF epoch tables only; NWB icephys stimulus series and ATF files give spike features without a stimulus.
• Ramps, trains and user-list protocols are not used for rheobase/f–I.
• Spike detection is a voltage-crossing rule; spikes that do not reach −20 mV (depolarisation block, dendritic recordings) need --peak-threshold.
• Extracellular detection reads each sweep whole (--max-seconds limits it); no whitening, common-average referencing or sorting.

Vocabulary (public identifiers of crates/openreadout-signal/src/ephys/*.rs)

identifier	meaning
to_mv, to_pa	unit factors to mV and pA
Protocol, output, unit, holding, from_trace, sweep_epochs, stimulus	the command output’s epoch table
SweepEpoch, index, kind, start_sample, end_sample, level	one epoch placed in a sweep
ApSettings, peak_threshold_mv, dvdt_threshold_v_per_s, dvdt_window_samples	spike detection settings
Spike, sweep, peak_sample, peak_time_s, peak_mv, threshold_time_s, threshold_mv, threshold_from_dvdt, amplitude_mv, half_width_ms, rise_time_ms, decay_time_ms, ahp_mv, ahp_depth_mv, upstroke_v_per_s, downstroke_v_per_s, isi_ms	one action potential
detect, adaptation_index	spike detection, ISI adaptation
StepResponse, baseline_mv, steady_state_mv, extreme_mv, input_resistance_mohm, tau_ms, sag_ratio, steady_fraction, current_step	passive response to a current step
TestPulse, delta_mv, holding_current_pa, peak_current_pa, steady_current_pa, access_resistance_mohm, membrane_resistance_mohm, total_resistance_mohm, capacitance_pf, test_pulse	voltage-clamp test pulse
base_and_steady, fit_exponential, fit_decay	measurement windows and the exponential fits (free asymptote; decay to zero)
CellRequest, trace, channel, sweeps, ap, max_spikes	analyze ephys-features request
CellReport, path, format, clamp_mode, sample_rate_hz, settings, cell, spike_count_total, spikes, spikes_truncated, notes	analyze ephys-features output
ChannelRef, name	the channel analysed
StimulusInfo, epoch	the stimulus found
SweepFeatures, stimulus_start_s, stimulus_end_s, stimulus_pa, stimulus_mv, spike_count, spike_count_stimulus, firing_rate_hz, first_spike_latency_ms, isi_mean_ms, isi_cv, half_width_mean_ms, amplitude_mean_mv, passive	one sweep
FiPoint, current_pa, rate_hz	f–I curve point
CellFeatures, resting_mv, rheobase_pa, rheobase_sweep, fi_curve, fi_slope_hz_per_pa, max_firing_rate_hz, membrane_capacitance_pf	per-cell summary
analyze_cell, analyze_extracellular	entry points
SpikesRequest, channels, max_seconds, max_times	analyze spikes request
SpikesReport, ChannelSpikes, noise, threshold, duration_s, times, times_truncated, SpikeTime, sample, time_s, amplitude	analyze spikes output
DetectSettings, low_hz, high_hz, order, sign, exclude_ms, PeakSign { Neg, Pos, Both }	extracellular detection settings
mad_noise, bandpass, detect_peaks	extracellular detection steps
Section, bandpass_sos, filtfilt	Butterworth band-pass and zero-phase filtering
EphysQuery, SpikesQuery, NmrQuery, band_hz, baseline_mode, DEFAULT_MAX_SPIKES	JSON queries shared by the MCP tools and batch tables (api.rs): the tool arguments by name, request() builds the request; DEFAULT_MAX_SPIKES is the per-spike rows returned by default (25)