Signal Viewer

The central viewer displays synchronized multi-channel PSG signal and annotation data.

Top panel

The top panel displays:

• clock-times in hour intervals (24-hour format) across the top

• a representation of the hypnogram, when staging data are present: NREM, REM and wake are blues, red and green respectively

• a lower, smaller representation of the currently included (i.e. unmasked epochs) at the bottom

Pan/zoom navigation

You navigate across signals by clicking on this top panel. The white bar (here around 10pm or 22:00) dictates the interval shown in the lower panel.

As well as using the mouse, you can use the cursor keys:

• left/right : move one epoch (30-seconds) backwards or forwards in time

• shift + left/right : move multiple epochs backwards or forwards

• up/down : zoom in/out, to change the span of the lower panel; unless you've Rendered the data (see below), the maximum span is 30 seconds

• click-and-drag: if you've rendered the data, you can click-hold-drag to select larger windows of time to be shown; if you double-click within the selected region, it will return to a single epoch view; you can resize or drag the larger interview by selecting it

Zoom and pan controls enable precise inspection of signal segments.

For continuous signals (those will more than 20 discrete values), when zooming out, at a certain point Lunascope will show the min/max per "pixel-timepoint" (i.e. the interval of time that corresponds to a single pixel on the screen); when zooming out even further, to prevent saturation, Luna will show the interquartile range (25th/75th) as approximated by a robust estimate of the standard deviation.

Rendering data

A main panel also contains some widgets for altering the view, including a Render button.

Internally, the view operates in two modes:

• on first attaching an EDF, Lunascope loads signals directly from the in-memory EDF on an as needed basis to plot them; signals and annotations can be included or excluded from the view by clicking on the respective signal or annotation dock. In this mode, however, there is a restriction that no more than 30 seconds of signal data can be viewed at a time (there is no restriction for annotation-only views).

• on pressing Render, Lunascope effectively takes a snapshot of the currently selected signals and annotations, then processes those data for more efficient viewing; subsequently, all items displayed are from the cached or rendered data store.

Rendering involves generating decimated (downsampled) versions of the data, with anti-alias filtering, and allows for subsequent further downsampling to enable large intervals of time to be displayed (i.e. hours) without Lunascope trying to plot millions of sample points directly (which would slow plotting).

There are three main advantages to rendering signals:

• it allows for larger windows to be viewed efficiently

• it allows for better Y-axis scaling of signals (see below)

• it precomputes various summary statistics and other things that can enhance visualization (although these are not yet included in the alpha release of Lunascope)

Three things to be mindful of when using the rendering option:

• Although rendering speeds up subsequent viewing, the process itself can take a few moments as it has to pre-process the entire dataset. Especially for large studies with many signals and high sampling rates, only use this function if you need it.

• Rendering takes a snapshot of the currently selected signals and annotations. If you want to add a new signal to the view, you need to press Render to re-render after selecting that signal or annotation. (You can drop remove/add back in rendered signals/annotations without re-rendering however.)

• If you change the underlying signal data -- e.g. filtering, referencing or rescaling, etc via the console -- these changes will not be directly shown (unlike the initial, unrendered view, which directly pulls data from the in-memory EDF representation). Rather, one needs to press Render again to update the rendered view. To indicate this, Lunascope always sets the Render button to orange after running any Luna script, as a reminder that the underlying data may have changed. If in doubt, just press Render once more to update the view.

Y-scaling

After rendering, the button turns from white to green to indicate that display items will now be drawn from the current snapshot. As noted, rendering also allows for different y-axis scaling:

• by default, unrendered signals are plotted to autoscale within the view window - i.e. each track will be scaled such that the minimum and maximum signal values are at the bottom/top of each track; as one scrolls left or right, the scale can change; extreme values will implicitly squash the rest of the signal to a near flat line

• Empiric Y : after rendering, the bottom/top of each track is set to the 10th and 90th percentile of the signal value; this fixes the scale and gives a generally sensible view of most signals, which can make it more intuitive to see changes in signal amplitude scrolling left/right; extreme values will now be drawn below or above the actual track, i.e. meaning tracks can overlap

• Clip Y : this stops overlap across channels with empiric scaling, by clipping the signal to the top/bottom of each track

Examples are these three viewes for the same interval and set of signals:

Default (autoscaling):

Empiric scaling:

Empiric scaling with clipping:

Other viewing options

You can also:

• hide/show the labels for signals and annotations, through the Labels checkbox

• set fixed Y-axis minimum and maximum values, by selecting Fixed Y and adjusting the min./max. values; note this currently applies to all signals

• alter the scaling and spacing of each signal track;

• spacing controls whether tracks are typically spaced (value of 1.0) versus, at the other extreme, completely overlapping (value of 0.0); it can sometimes be useful to overlay two signals to see the differences betweeen them more clearly

• scaling controls how many pixels each track takes up, thereby altering the display Y-axes rather than the physical-unit scaling per se; this can be useful primarily in the hd-EEG case, where one wants to view multiple inter-related signals

• change the colors of the main view, by selecting options from the top Palette menu