• No results found

Addressing the dataset density problem through additional i-Raster

The i-Raster Visualisation

9 Overview of the i-Raster Visualisation

9.1 Drawbacks of Somerville’s i-Raster Visualisation with modern neuroscience

9.1.2 Addressing the dataset density problem through additional i-Raster

In addition to applying the new VPL to manage modern neural science data sets it is also possible to introduce additional functionality to the i-Raster visualisation. As with the VPL the aim of this functionality is to efficiently sort, zoom and filter sections of the dataset. This additional functionality should not be viewed in isolation as it both extends Somerville’s i-Raster functionality and complements the processing performed by the VPL.

Two key pieces of functionality have been added to achieve the goal of managing increased neural dataset size:

1. A time filter that allows the idea of time segmenting a raster chart (as seen in the VPL) to become an interactive user experience and

Chapter 9: The i-Raster Visualisation

Page | 139

2. A graphical representation of the multi-electrode array (MEA) used to record the spike trains. Individual electrodes or groups of electrodes can be selected and the raster chart interactively reconfigures itself to show only the spike trains recorded by those electrodes.

9.1.2.1 The interactive time filter for i-Raster

The usefulness of visualisation as a tool is enhanced by the modern computers ability to allow the user to interact with the visualisation. This fact is at the core of the Visual Information Seeking Mantra in the form of its advice to use filtering and zooming to explore datasets. Somerville’s i-Raster implementation was primarily concerned with sorting and filtering data in the y-axis of the raster chart (by electrode id number). Using the VPL this has been extended by sorting and grouping spike trains. Within i-Raster individual spike trains may now be filtered into or out of groups and moved between groups. However the x-axis (or time dimension) offered no such filtering or zooming abilities. This deficiency has been addressed through the introduction of an interactive time filter.

In principle the operation of the interactive time filter is relatively simple. The user is presented with a scale bar representing the entire x-axis of the raster chart. A start and end slider define the beginning and end of a time range. Only spike events that fall between the start and end of this time range will be shown on the raster chart. Figure 9-19 shows the concept in action. The 1411 burst sorted and grouped spike trains seen in Figure 9-13 have been time filtered to show only the first 60 seconds of spiking activity.

Page | 140

Figure 9-19: Time filtered raster chart extracted from a 1411 spike train dataset

As can be seen in Figure 9-19 the time filtering achieves on the x-axis of the raster chart what grouping does for the y-axis. By filtering out the spike train outside of the time range the scale of the raster chart is changed. This has a zooming effect allowing for a more detailed view of a smaller section of the dataset. Figure 9-20 shows the key components of the time filtering tool.

Chapter 9: The i-Raster Visualisation

Page | 141

Component Function

Menu bar push button toggles the time range selection tool on / off.

Time range selection tool represents the x- axis of the raster chart.

Time range selections start slider. Colour coded green in a traffic light style.

Time range selections end slider. Colour coded red in a traffic light style.

Time point entry box permits entry of an exact time point value for either time range slider. Double click start or end slider to directly type a time point value for that slider.

Figure 9-20: Components of the interactive time filtering tool

Taken together all the components in Figure 9-20 allow the user to view any portion of the recorded dataset at any desired scale. The raster chart updates in real time as the user modifies the start and end range selection sliders. This allows the user to zoom seamlessly between the two views seen in Figure 9-19. The user is able to maintain an awareness of the overall dataset while concentrating their attention on particular time periods.

9.1.2.2 The multi-electrode array display (Electrode Display)

Accessed through i-Raster’s main interface the electrode display provides a visualisation of the multi-electrode array (MEA) that recorded the currently viewed spike train data.

Figure 9-21: Accessing the electrode display from i-Raster’s primary interface.

The increase in the number of recorded spike trains in a modern neuroscience dataset has, in part, been achieved by using ever increasing numbers of electrodes on the recording MEA devices. Neurons that are physically close to each other in a biological sample are more likely to form connections than those which are widely separated. Proximity should not be taken as a guarantee of connectivity (some neuron axon’s can be over a meter long). Proximity between neurons generating the recorded spike trains does, however, provide a starting point for exploring the dataset. The electrode display provides a means to rapidly select and filter spike trains on a raster chart to show only those recorded from specific regions of the MEA.

Page | 142

Figure 9-22: A two second burst sorted recording of 500 spike trains with the electrode display shown.

Figure 9-22 shows a two second burst sorted recording of 500 spike trains. The electrode display has been activated and since the whole dataset is being viewed all electrodes are selected. The VPL program that loaded this dataset is identical to the one seen in Figure 9-12 earlier. To filter the dataset to show spike trains from a single region the user simply drags a selection rectangle around the area. Multiple regions of the MEA can be selected by holding down the right control key while selecting. When individual spike trains are being displayed (rather than summary groups) placing the mouse over an electrode will highlight the corresponding spike train. Figure 9-23 shows the same dataset filtered to show data from three regions. One of the summary spike trains has been expanded and the electrode that recorded spike train 101 identified.

Chapter 10: The i-Grid Visualisation

Summary

This chapter describes the iGrid visualisation its implementation and the visualisation techniques add to it to manage the large datasets now being produced by recording hardware.

Chapter 10

The i-Grid Visualisation

Page | 144