ASSYST
Automated Rodent Seizure Detection Software
Overview
ASSYST is an interactive software tool aimed to significantly automate the process of identifying seizures and/or other epileptiform events in rodent EEG recordings, in off-line operation mode.
ASSYST is now widely used in epilepsy research centres around the world. ASSYST software is a key tool adopted by the global EPIBIOS project and is referenced in several published papers.
Features & Benefits
- ASSYST is able to read, display, and process the rodent EEG recordings, and save the processed results (the timings for identified seizures and/or other events).
- ASSYST needs only a single EEG channel for seizure detection; however, if required, it can process multiple channels as well.
- ASSYST not only finds the seizures and/or other epileptiform activity, but also allows the user to classify the identified epileptiform events into different user-defined categories (e.g., short or long seizures, spike-wave discharges (SWD), interictal spikes, etc.). The classified events can be saved into a text or Excel file.
- ASSYST may be used not only for seizure, SWD or other epileptiform activity detection but also for detection of various oscillations, including high-frequency oscillations (HFO).
- ASSYST typically takes 20 minutes to process and then complete classification of >7 days of continuous multi-channel rodent EEG recordings.
- ASSYST has been validated in many rat and mouse models with >99% sensitivity.
- ASSYST is interactive: the user can intervene and make adjustments at different stages of processing to achieve desired results.
- ASSYST is powered with a new V3.0 algorithm for fully automatic identification of events with regular repetitive structure, such as spike-wave discharges (SWDs), and distinguishing them from artefacts and other irrelevant events.
- ASSYST provides remarkable flexibility in processing depending on particular data and user needs – from highly automated with minimum influence from user, to a semi-automated inspection of every ASSYST-identified ‘event’, large or small.
- ASSYST allows easy work with parameters: it has several preset parameter sets for detection of different types of epileptiform activity (e.g., seizures or SWDs) in different rodent models of epilepsy. For every client, we can prepare specific customized parameter sets depending on animal model used and types of events to be detected.
Method
- ASSYST uses an advanced time-frequency analysis technique to reveal the EEG segments with excessive activity in a certain frequency band.
- We have found that electrographic seizures and SWDs in rats and mice contain a specific frequency component that is not present in the normal, interictal EEG. This finding formed the basis of our method for identifying and extracting epileptiform activity in the EEG. Powered by our advanced algorithm for precise calculation of short-term Fourier transform with high frequency resolution, effective and reliable performance of the software is ensured.
- The spectral content of the EEG within the given frequency band is assessed by Spectral Band Index (SBI), which is derived from the power spectrum.
- For automatic SWD identification a novel technique based on running beat spectrum analysis is used.
History
ASSYST version 1.1 - New Features Added
(February, 2019)
- Ability to find the closest event (in the list of selected events) to the selected position in the SBI pane.
- In Multiple Channel processing, the channels that contain the particular event are indicated in the "Comment" column in the Events list of the Events form, as well as in the output text and excel files.
- Ability to read data from two more file formats (in addition to EDF and Compumedics Profusion):
- Ripple NSx file format
- Acknowledge ACQ file format
- Indication of whether the current event has already been added to an event category or not, in the form of a text label 'In Category' near the Current Category control that shows up if the event is in a category. The hint of this label shows, to which categories the event is added.
ASSYST version 2.0 - New Features Added
(May, 2019)
There were three major methodological developments in version 2.0:
- Significant improvements to the SBI threshold definition algorithm
- This included SBI curve and SBI histogram normalization to make the SBI curves and the defined thresholds comparable across channels in one recording, as well as across different recordings and animals.
- Improvements to the automatic threshold definition method:
- This was modified and a new parameter affecting threshold definition introduced to allow for more user control.
- Multiple Channel Processing enhanced:
- The ability to process multiple channels in one run was previously very basic and restricted. In version 2.0, due to SBI normalization, multiple channels processing is much more sophisticated and enhanced with several possible options.
ASSYST version 2.1 - New Features Added
(December, 2019)
- Ability to output the SBI curve as a text file or image (WMF or JPEG).
- Ability to save the Power spectrum of the current window as a text file.
- Ability to calculate Spectrogram - a 2D time-frequency color-coded plot. A checkbox on the Parameters (Advanced) tab is added to select option to calculate the Spectrogram.
- Ability to continuously scroll ("play") the EEG signals forward and backward (with 2 different speeds).
- Ability to output the screenshots of the events, in two possible file formats - WMF or JPEG, when outputting the selected events.
ASSYST version 3.0 - New Features Added (April, 2020)
ASSYST 3.0 has one major enhancement over previous ASSYST versions:
- ASSYST is now empowered with a new algorithm for automatic identification and extraction of events with regular repetitive structure, such as spike-wave discharges (SWDs).
- The new SWD algorithm, which is applied after the time-frequency analysis and the threshold-based event selection, can automatically distinguish spike-wave discharges from artefacts and noise with a remarkable accuracy.
- In contrast to rodent models that have a few strong seizures during several weeks, the models with SWDs (e.g., GAERS or WAG/Rij) usually exhibit hundreds of SWDs each hour. In such cases the number of events selected by ASSYST, based on SBI threshold, may be several thousands in just a few hour-long recordings. Visual examination/assessment of such a high number of ASSYST selected events in order to reject artefacts or other irrelevant events, requires significant time from the user, though still much less than visual examination of the entire raw EEG. With ASSYST 3.0, this examination/assessment process is much faster and more automated.
With ASSYST 3.0, this examination/assessment process is much faster and more automated.
ASSYST version 3.1 - New Features Added (June, 2020)
- In previous versions of the program it was possible to use a montage (i.e., linear combination of original EEG channel) by loading a special montage file created either manually or using other software. In version 3.1 it is possible to define and use a montage directly within the program.
- Another addition is the possibility to measure intervals (and their reciprocal, i.e., the frequency) between specified points of the signal. This is very handy when defining the frequency range of SWDs.
- Also, now it is possible to output some additional characteristics of the power spectrum of the selected events.
- In addition, some changes are made in the SWD identification algorithm, which should improve performance especially for shorter events.
ASSYST version 3.2 - New Features Added (October, 2020)
- In the SWD identification algorithm, a new optional module is added aimed to eliminate too weak and too strong events (artifacts) that accidentally pass the running beat spectrum criteria. The algorithm is based on analysis of line length of the selected events.
ASSYST version 3.3 - New Features Added (February, 2023)
- Possibility to cut out too strong events that are caused by artifacts, using a second threshold (upper cut-off threshold) on the SBI curve.
- Possibility to automatically refine the start and end times of the selected events.