Metrics and curation

In the app

Curation shows the active view’s sorting in views you arrange: each is a dock panel you can drag, snap or close. The Views menu toggles each one, and the Layout menu applies a grid.

View	Shows
Unit table	Sort by any column, filter by label
Waveforms	The selected units’ waveforms
Features	PC1 against PC2
Correlograms	Auto-correlogram, and cross-correlograms between selected units
ISI	Inter-spike intervals
Amplitude	Amplitude over time
Template	The lead unit’s template
Similarity	The units most like it: merge candidates
Probe	Units on the probe

Under them, the bottom strip shows the lead unit’s channels over the shared timeline, or a raster.

Charts are interactive: wheel zooms (Shift for x, Alt for y), drag pans, and double-click resets. Each has its settings under ⋯.

Action	Keys
Label good / mua / noise	Alt+G / Alt+M / Alt+N
Merge the selected units	G
Split	Lasso on the features or amplitude view
Undo / redo	Ctrl+Z / Ctrl+Shift+Z
Save	Ctrl+S

Edits are saved as a log, curation.json, inside the .sorting.zarr, applied on read. The sorter’s arrays are never rewritten.

From Python

dspk.synapse.metrics has the analyses. Defaults follow SpikeInterface or the cited papers.

Question	Function
Is the unit refractory?	isi_violations (Hill et al., 2011), llobet_contamination (Llobet et al., 2022), acg_refractory
Is it present throughout?	presence_ratio, firing_range
Are spikes missed?	amplitude_cutoff
Is it well separated?	silhouette, isolation_distance, d_prime, snr
A single score	composite_score (EMUsort’s)
Should two units merge?	ccg_refractory, suggest_merges
Do two sortings agree?	compare_trains, compare_sortings
Timing	autocorrelogram, crosscorrelogram, firing_rate, bursts, psth, stimulus_average

sorting = dspk.synapse.io.Sorting.load("sorted")
train = sorting.spike_train(3)
dspk.synapse.metrics.isi_violations(train, sorting.total_samples, sorting.sample_rate)