Low-Cost Platform for Multianimal Chronic Local Field Potential Video Monitoring with Graphical User Interface (GUI) for Seizure Detection and Behavioral Scoring

Abstract

Experiments employing chronic monitoring of neurophysiological signals and video are commonly used in studies of epilepsy to characterize behavioral correlates of seizures. Our objective was to design a low-cost platform that enables chronic monitoring of several animals simultaneously, synchronizes bilateral local field potential (LFP) and video streams in real time, and parses recorded data into manageable file sizes. We present a hardware solution leveraging Intan and Open Ephys acquisition systems and a software solution implemented in Bonsai. The platform was tested in 48-h continuous recordings simultaneously from multiple mice (male and female) with chronic epilepsy. To enable seizure detection and scoring, we developed a graphical user interface (GUI) that reads the data produced by our workflow and allows a user with no coding expertise to analyze events. Our Bonsai workflow was designed to maximize flexibility for a wide variety of experimental applications, and our use of the Open Ephys acquisition board would allow for scaling recordings up to 128 channels per animal.

Keywords: Bonsai-Rx; Intan; Open Ephys; data synchronization; epilepsy; seizure.

Figure 1.

Schematic of multianimal chronic video-LFP recording and analysis pipeline. A, Components for chronic LFP recordings. Custom-made electrodes (bottom left, scalebar: 3 mm) were implanted into CA1 of the hippocampus. Custom-made cables (top left) were soldered to an Intan adapter board (middle) and further protected with 10 mm heat shrink tubing (right). B, Timeline of experiment. Red rectangles represent 48 h of continuous monitoring. C, Experimental setup. Recordings were performed from four mice simultaneously. Video-LFP data were acquired and controlled in a custom-made Bonsai workflow. Seizure analysis was performed with a custom-made seizure analysis GUI, written in MATLAB. D, LFP and video were acquired and synchronized by merging timestamps (left). Timestamp, video, and voltage data were streamed to time-dependent loops (nodes labeled with X and *X) that enabled data saving once per hour.

Figure 2.

LFP signals simultaneously recorded from different mice are independent. A, Five-second-long clips of LFP recorded simultaneously from two mice during hour 1 of monitoring (left) or hour 2 of monitoring (right). For each mouse two traces are shown corresponding to electrodes located in left (L) and right (R) hippocampus. Red lines indicate a period in which signals from both hemispheres showed synchronous signal within mouse 1. B, A matrix to visualize comparisons made to create distributions shown in C. Correlations between hour-long traces were made for comparisons that fell into three categories: (1) left versus right same mouse, same hour (gray); (2) different mice, same hour (purple); (3) different mice, different hours (cyan). Black represents self-comparisons, which were not included. White represents comparisons between the same mouse, but different hours, which were also not included. For actual distributions shown in C, comparisons were made across four mice, and for each hour across 24 h of recording. C, Cumulative density functions (CDFs) of correlations obtained from comparisons described in B. ***p < 0.001, one-way ANOVA, post hoc Tukey’s test

Figure 3.

Seizure analysis GUI. A, Schematic of the GUI. A working directory is selected and recording date and mouse are chosen to be analyzed (top, right). Channels to show and duration of LFP traces are set by the user in Plot settings (top, left). Video lag, allows the user to set playback speed. Pressing “Seizure Detection” opens a pop-up window where the user may filter and/or flip signals before applying a threshold. When the user accepts the detection results, video and filtered LFP of the first event is loaded into the Video Playback and LFP Viewer. Detected seizure events are toggled through by the user using the “Next” button. Questions regarding the LFP signal and behavior are answered for each event (left). After performing scoring, annotations are saved by pressing the “Save” button. B, Seizure detection sequence. The user selects a channel on which the detection will occur (blue box). LFP is down-sampled, bandpass-filtered, and possibly flipped according to user input. An amplitude threshold is set by inputting an integer value (green line). Finally, detected events are displayed (magenta). User can expand view to look at several individual events and then choose to accept all results and start scoring or adjust the threshold and repeat the detection. C, Example of detected subclinical seizure. Detected seizure is indicated by the red bar and additional 60 s preseizure and postseizure baseline are shown. Gray lines indicate the range of the zoomed in 2-s-long gray trace. In this example, the seizure is only observed on the right hemisphere (bottom). D, Example of Stage V seizure. Seizure can be observed on both hemispheres. Notations are the same as in C.

Figure 4.

Examples of subclinical seizures and behavioral seizures detected by the GUI. A, Segments of data from 1 h of recording that have been passed through the spectral method of seizure detection using the GUI. Early in the hour, a cluster of three subclinical seizures were detected and later in the hour, one Stage V seizure was detected. Classification as subclinical or behavioral was done by three independent scorers using the GUI. B, Enlarged versions of the events shown in A. For events s1 and s4, the first 3 s are enlarged on the right.