List of all quality metrics and parameters - Julie-Fabre/bombcell GitHub Wiki

List of all the quality metrics and parameters

Variable naming: Quality metrics are accessed as quality_metrics["metricName"] (Python) or qMetric.metricName (MATLAB). Parameters are accessed as param["paramName"] (Python) or param.paramName (MATLAB).

Quality metrics

Noise quality metrics

Brief description	Quality metric field	Calculated	Classification parameter
Number of waveform peaks	`nPeaks`	Always	`maxNPeaks`
Number of waveform troughs	`nTroughs`	Always	`maxNTroughs`
Waveform peak-to-trough duration (µs)	`waveformDuration_peakTrough`	Always	`minWvDuration`, `maxWvDuration`
Waveform spatial decay slope	`spatialDecaySlope`	If `computeSpatialDecay`	`minSpatialDecaySlope` (linear) or `minSpatialDecaySlopeExp`, `maxSpatialDecaySlopeExp` (exponential)
Waveform baseline flatness	`waveformBaselineFlatness`	Always	`maxWvBaselineFraction`
'Repolarization peak'-to-trough ratio	`scndPeakToTroughRatio`	Always	`maxScndPeakToTroughRatio_noise`

Non-somatic quality metrics

Brief description	Quality metric field	Calculated	Classification parameter
Trough-to-'repolarization peak' ratio	`troughToPeak2Ratio`	Always	`minTroughToPeak2Ratio_nonSomatic`
First peak width (samples)	`mainPeak_before_width`	Always	`minWidthFirstPeak_nonSomatic`
Main trough width (samples)	`mainTrough_width`	Always	`minWidthMainTrough_nonSomatic`
'First peak'-to-'repolarization peak' ratio	`peak1ToPeak2Ratio`	Always	`maxPeak1ToPeak2Ratio_nonSomatic`
'Main peak'-to-trough ratio	`mainPeakToTroughRatio`	Always	`maxMainPeakToTroughRatio_nonSomatic`

Multi-unit quality metrics

Brief description	Quality metric field	Calculated	Classification parameter
Number of spikes	`nSpikes`	Always	`minNumSpikes`
Percentage of spikes below detection threshold	`percentageSpikesMissing_gaussian`	Always	`maxPercSpikesMissing`
Percentage missing (symmetric estimate)	`percentageSpikesMissing_symmetric`	Always	N/A (not used for classification)
Fraction of refractory period violations	`fractionRPVs_estimatedTauR`	Always	`maxRPVviolations`
Maximum drift estimate (µm)	`maxDriftEstimate`	If `computeDrift`	`maxDrift`
Cumulative drift estimate (µm)	`cumDriftEstimate`	If `computeDrift`	N/A (not used for classification)
Presence ratio	`presenceRatio`	Always	`minPresenceRatio`
Mean raw waveform amplitude (µV)	`rawAmplitude`	If `extractRaw`	`minAmplitude`
Signal-to-noise ratio	`signalToNoiseRatio`	If `extractRaw`	`minSNR`
Isolation distance	`isolationDistance`	If `computeDistanceMetrics`	`isoDmin`
L-ratio	`Lratio`	If `computeDistanceMetrics`	`lratioMax`
Silhouette score	`silhouetteScore`	If `computeDistanceMetrics`	`ssMin` (not currently implemented)

Time chunk metrics

Brief description	Quality metric field	Calculated	Notes
Start of valid time window (s)	`useTheseTimesStart`	If `computeTimeChunks`	Used for subsequent metric calculations
End of valid time window (s)	`useTheseTimesStop`	If `computeTimeChunks`	Used for subsequent metric calculations
RPV tauR window index	`RPV_window_index`	Always	Internal: which tauR value was selected

Parameters

General parameters

Brief description	Parameter	Default	Notes
Plot per-unit detailed figures	`plotDetails`	`False`	Generates many plots, use sparingly
Plot global summary plots	`plotGlobal`	`True`	Histograms, upset plots
Save plots to disk	`savePlots`	`False`
Directory to save plots	`plotsSaveDir`	`None`	If None, saves to kilosort_path/bombcell_plots/
Display progress information	`verbose`	`True`
Re-extract raw waveforms	`reextractRaw`	`False`	Set True to regenerate cached waveforms
Save outputs as TSV	`saveAsTSV`	`True`	Useful for Phy integration
Save unit_type TSV for Phy	`unit_type_for_phy`	`True`	Creates cluster_bc_unitType.tsv
Path to Kilosort directory	`ephysKilosortPath`	(required)

Duplicate spike parameters

Brief description	Parameter	Default	Notes
Remove duplicate spikes	`removeDuplicateSpikes`	`False`
Duplicate spike window (s)	`duplicateSpikeWindow_s`	`0.000034`	~1 sample at 30kHz
Save spikes without duplicates	`saveSpikes_withoutDuplicates`	`True`
Recompute duplicate spikes	`recomputeDuplicateSpikes`	`False`

Raw waveform / amplitude parameters

Brief description	Parameter	Default	Notes
Extract raw waveforms	`extractRaw`	`True`	Required for amplitude/SNR metrics
Number of raw spikes to extract	`nRawSpikesToExtract`	`100`	Increase for UnitMatch (1000)
Detrend average waveforms	`detrendWaveform`	`True`	Linear detrend for quality metrics
Detrend raw waveforms for UnitMatch	`detrendForUnitMatch`	`False`	UnitMatch does its own detrending
Save individual raw waveforms	`saveMultipleRaw`	`False`	Required for UnitMatch
Decompress .cbin data	`decompress_data`	`False`	Enable for compressed data
Probe type	`probeType`	`1`	1 = Neuropixels 1.0, 2 = Neuropixels 2.0

Recording parameters

Brief description	Parameter	Default	Notes
Ephys sample rate (Hz)	`ephys_sample_rate`	`30000`	Overridden by meta file if provided
Number of channels in raw data	`nChannels`	`385`	Including sync channels
Number of sync channels	`nSyncChannels`	`1`
Path to meta file	`ephys_meta_file`	`None`	.meta (SpikeGLX) or .oebin (OpenEphys)
Path to raw data file	`raw_data_file`	`None`	.bin, .dat, or .cbin
Gain to µV conversion	`gain_to_uV`	Auto-detected	From meta file if available

Spike width parameters (Kilosort version dependent)

Brief description	Parameter	KS4 default	KS2/3 default
Spike width (samples)	`spike_width`	`61`	`82`
Baseline noise window (samples)	`waveformBaselineNoiseWindow`	`10`	`20`
Baseline window start (samples)	`waveform_baseline_window_start`	`0`	`21`
Baseline window stop (samples)	`waveform_baseline_window_stop`	`10`	`31`

Refractory period parameters

Brief description	Parameter	Default	Notes
Refractory period minimum (s)	`tauR_valuesMin`	`0.002`
Refractory period maximum (s)	`tauR_valuesMax`	`0.002`	If different from min, estimates optimal tauR
Refractory period step (s)	`tauR_valuesStep`	`0.0005`
Censored period (s)	`tauC`	`0.0001`	To prevent duplicate spikes
Use Hill or Llobet method	`hillOrLlobetMethod`	`True`	True = Hill, False = Llobet et al.

Time chunk parameters

Brief description	Parameter	Default	Notes
Compute time chunks	`computeTimeChunks`	`False`	Split recording into chunks
Time chunk size (s)	`deltaTimeChunk`	`360`	6 minutes

Presence ratio parameters

Brief description	Parameter	Default	Notes
Presence ratio bin size (s)	`presenceRatioBinSize`	`60`

Drift parameters

Brief description	Parameter	Default	Notes
Compute drift	`computeDrift`	`False`	Disabled by default (slow)
Drift bin size (s)	`driftBinSize`	`60`

Waveform shape parameters

Brief description	Parameter	Default	Notes
Peak/trough detection threshold	`minThreshDetectPeaksTroughs`	`0.2`	× max waveform value
Compute spatial decay	`computeSpatialDecay`	`True`
Use linear fit for spatial decay	`spDecayLinFit`	`False`	False = exponential (preferred)
Normalize spatial decay	`normalizeSpDecay`	`True`	Invariant to spike-sorting

Distance metric parameters

Brief description	Parameter	Default	Notes
Compute distance metrics	`computeDistanceMetrics`	`False`	Disabled by default (slow)
Number of channels for distance metrics	`nChannelsIsoDist`	`4`

Classification thresholds

Noise classification thresholds

Brief description	Parameter	Default
Maximum number of peaks	`maxNPeaks`	`2`
Maximum number of troughs	`maxNTroughs`	`1`
Minimum waveform duration (µs)	`minWvDuration`	`100`
Maximum waveform duration (µs)	`maxWvDuration`	`1150`
Maximum baseline fraction	`maxWvBaselineFraction`	`0.3`
Maximum 'repolarization peak'-to-trough ratio	`maxScndPeakToTroughRatio_noise`	`0.8`
Minimum spatial decay slope (linear fit)	`minSpatialDecaySlope`	`-0.008`
Minimum spatial decay slope (exponential fit)	`minSpatialDecaySlopeExp`	`0.01`
Maximum spatial decay slope (exponential fit)	`maxSpatialDecaySlopeExp`	`0.1`

Non-somatic classification thresholds

Brief description	Parameter	Default	Criterion
Minimum trough-to-peak2 ratio	`minTroughToPeak2Ratio_nonSomatic`	`5`	1
Minimum first peak width (samples)	`minWidthFirstPeak_nonSomatic`	`4`	2
Minimum main trough width (samples)	`minWidthMainTrough_nonSomatic`	`5`	3
Maximum peak1-to-peak2 ratio	`maxPeak1ToPeak2Ratio_nonSomatic`	`3`	4
Maximum main peak-to-trough ratio	`maxMainPeakToTroughRatio_nonSomatic`	`0.8`	5 (standalone)
Split good/MUA non-somatic	`splitGoodAndMua_NonSomatic`	`False`	Separate GOOD NON-SOMA vs MUA NON-SOMA

Non-somatic classification logic: A unit is classified as non-somatic if criteria 1-4 are ALL met, OR if criterion 5 alone is met.

Multi-unit classification thresholds

Brief description	Parameter	Default	Notes
Minimum amplitude (µV)	`minAmplitude`	`40`	Tune per dataset
Minimum SNR	`minSNR`	`5`	Tune per dataset
Maximum RPV fraction	`maxRPVviolations`	`0.1`
Maximum % spikes missing	`maxPercSpikesMissing`	`20`
Minimum number of spikes	`minNumSpikes`	`300`
Minimum presence ratio	`minPresenceRatio`	`0.7`
Maximum drift (µm)	`maxDrift`	`100`

Distance metric thresholds

Brief description	Parameter	Default	Notes
Minimum isolation distance	`isoDmin`	`20`
Maximum L-ratio	`lratioMax`	`0.3`
Minimum silhouette score	`ssMin`	`NaN`	Not currently implemented