Real-time, large-scale, processing in Scanbox

We have expanded the real-time processing capabilities in Scanbox to deal with large number of neurons, while offering real-time image stabilization, raw fluoresence signal extraction, non-linear deconvolution and display of the results — all at video rates.

Scanbox “Segment” button in the real-time panel can now be clicked after collecting ~2min data to automatically register and segment cells using Suite2p. The process is completely transparent to the user. After conclusion, Scanbox will display the mean, registered image that will be used for registration during the subsequent experiments. As before, the user will select a few regions where stable 2D features are present in the image so they can be used for robust registration. Right after the selection of these regions Scanbox will load the ROIs selected by Suite2p that have a “iscell” probability higher than what is specified in the box to the right of the “Segment” button. Here is the process in action:

You should remember to click the “Stabilize” button next, as this will tell scanbox that images should be registered with the mean image that resulted during registration. This is critical as any movement of the image with respect to the ROIs will result in noise. After this, you are ready to record as always:

The raw fluorescence signal will be displayed in green, the spikes in red, and the stimulus marker in blue. Sliders below the window provide a “gain” that can be used to control the modulation of each signal. The pull down menu on the right, allows the user to visualize only the raw signal, only the spikes, or both. The resulting data can be streamed over the network or shared via memory-mapped files with other processes.