Table 2. Currently available quality metrics in Spikeinterface.
Re-implemented by researchers at Allen Institute for Brain and by the SpikeInterface team.
Metric | Description | Reference |
---|---|---|
Signal-to-noise ratio | The signal-to-noise ratio computed on unit templates. | N/A |
Firing rate | The average firing rate over a time period. | N/A |
Presence ratio | The fraction of a time period in which spikes are present. | N/A |
Amplitude Cutoff | An estimate of the miss rate based on an amplitude histogram. | N/A |
Maximum drift | The maximum change in spike position (computed as the center of mass of the energy of the first principal component score) throughout a recording. | N/A |
Cumulative drift | The cumulative change in spike position throughout a recording. | N/A |
ISI violations | The rate of inter-spike-interval (ISI) refractory period violations. | Hill et al., 2011 |
Isolation Distance | Radius of the smallest ellipsoid that contains all the spikes from a cluster and an equal number of spikes from other clusters (centered on the specified cluster). | Harris et al., 2001 |
L-ratio | Assuming that the distribution of spike distances from a cluster center is multivariate normal, L-ratio is the average value of the tail distribution for non-member spikes of that cluster. | Schmitzer-Torbert and Redish, 2004 |
D-Prime | The classification accuracy between two units based on linear discriminant analysis (LDA) | Hill et al., 2011 |
Nearest-neighbors | A non-parametric estimate of unit contamination using nearest-neighbor classification. | Chung et al., 2017 |
Silhouette score | The ratio between cohesiveness of a cluster (distance between member spikes) and its separation from other clusters (distance to non-member spikes). | Rousseeuw, 1987 |