units_table

The Neural Units Table

The neural units table, occupying the top left corner of XSort's main window, shows all distinct neural units defined in the current working directory: all units extracted from the original spike sorter file that you have not deleted, merged, or split, plus any derived units resulting from merging or splitting (see the Making Changes section of the user guide).

Each row in the table corresponds to one neural unit, and the columns display numeric metrics or other information:

1. UID - The unit ID, typically an integer N. For a Purkinje cell spike train extracted from the original spike sorter file, two units are created -- 'Nc' (complex spikes) and 'Ns' (simple spikes). Any unit derived from a merge or split operation has the UID 'Nx', where the 'x' suffix indicates it is a derived unit.
2. Channel -- The unit's primary channel, ie, the analog channel on which the best signal-to-noise ratio was measured.
3. #Spikes -- The total number of spikes in the unit's spike train.
4. Rate (Hz) -- The unit's mean firing rate.
5. SNR -- Highest observed signal-to-noise ratio for the unit across all recorded analog data channels.
6. Amp (uV) -- Peak-to-peak amplitude of the unit's mean spike waveform (template) as measured on the primary channel.
7. %ISI<1 -- Fraction of interspike intervals (ISI) in this unit's spike train that are less than 1 millisecond. An ISI less than the typical refractory period is an indication that some of the spike timestamps attributed to the unit are simply noise or should be assigned to another unit.
8. Similarity -- The degree of similarity of this unit to the first selected unit in the table -- the primary unit. For each unit in the table, a 1D sequence is formed by concatenating the unit's per-channel spike templates. The similarity metric for unit A is the cross-correlation cofficient of unit A's sequence with the analogous sequence for the primary unit. A value of 1 indicates perfect correlation; units may be negatively correlated. Of course, the similarity metric is unknown if no unit is currently selected in the table.
9. Label - An optional, short label (typically used to specify the unit's putative neuron type). See Making Changes chapter.

NOTE: Per-channel spike templates are only computed on the 16 channels "near" a unit's primary channel (XSort does not yet support probe geometry information, so "near" means channels [P-8 .. P+7], where P is the primary channel index). This is the unit's template channel set. The similarity metric for unit A vs B only includes the templates for those channels in the intersection of unit A's template channel set with unit B's. If that intersection is empty, then the similarity is 0.

Hiding columns and sorting on a given column

With the exception of the UID column, any of the columns in the units table may be hidden. To hide a column, right-click anywhere on the table header and uncheck the label of the column you wish to hide. The table updates immediately. To unhide a colum, right-click again and select any unchecked column that you wish to show.

Note that XSort "remembers" which columns are hidden, storing the column state in the user's settings -- so the same columns will be hidden the next time you run the program.

To sort the table on any column (except the Similiarity column), simply click on that column's header. Click again to sort in reverse order. The identity of the sort column is not persisted in user settings; XSort always sorts on the UID column initially.

The current display focus list

The display focus list is the subset of units currently selected for visualization in XSort. Up to 3 units may be chosen for display at any one time for comparison purposes, and a highlight color is assigned to each: blue for the first unit in the list -- the primary focus unit --, red for the secondary focus unit, and yellow for the tertiary focus unit. The same color is used to render the spike templates, correlograms, and other statistics for that unit across the various views.

When XSort initially loads the content of the current working directory, there are no units in the focus list, so most of the views contain nothing of interest. To select the primary focus unit, simply (left-)click anywhere on the corresponding row in the table; the previous selection, if any, is cleared. To select the second and third focus unit, hold down the Ctrl key (the Command key in MacOS) down while clicking on the relevant rows. You can also use the Up/Down arrow keys to change the identity of the primary focus unit without using the mouse, or use Ctrl-Up/Down to change the identity of the secondary focus unit without changing the primary unit.

You will notice that the views are updated -- sometimes after a noticeable delay -- to display the relevant metrics or statistics for the selected unit(s). Some statistics, once computed, are cached in memory and render quickly -- such as the correlograms and spike templates. The principal component analyis, however, must be redone each time the display focus list changes.

Whenever any statistics need to be computed, those calculations happen in the background and the views are updated as the background task delivers its results. If a background task is still running and you change the focus list again or actually edit the units table, that task is cancelled and a modal dialog blocks user input until that task stops, typically in a few seconds or less. (Hint: The PCA computations are slow. If you don't need PCA, hide that view and XSort will not run a background task to perform the analysis.)

While a recording session may include hundreds of analog channels, the Templates view only displays unit spike templates on a maximum of 16 channels, and the Channels view only displays a maximum of 16 individual analog channel traces. The primary focus unit determines the range of channels selected: the 16 channels in the neighborhood of that unit's primary channel.

Highlighting units most similar to the current primary focus unit

The Similarity column in the units table indicates how similar each unit is to the current primary focus unit. The most similar units may not be directly adjacent to the primary unit. However, if you check the "Highlight up to 5 most similar units..." box below the table, then XSort will list the 5 most similar units immediately below the primary unit regardless the sort order and are highlight them with a light blue background. The state of the checkbox is a user preference saved at application exit and restored at startup.