EGMclassifier

Overview

EGMclassifier is a MATLAB App Designer tool for viewing, annotating, and classifying electrograms (EGMs) from datasets stored in OpenEP .mat format.

The app allows the user to:

• load one or more OpenEP cases
• browse individual EGMs
• manually classify EGMs into user-defined classes
• add annotation markers by clicking on signals
• display waveform features such as WOI and activation time
• overlay different EGM duration detection methods
• save the labelled dataset for later use

Requirements

• MATLAB
• App Designer support
• Input files in OpenEP .mat format
• Optional: a SubFunctions folder in the same directory for helper functions used by duration analysis

Input

The app expects .mat files containing a userdata structure in OpenEP format.

At minimum, the loaded file should contain:

• userdata.electric.egm

Additional OpenEP fields are used when available, for example:

• userdata.electric.annotations.woi
• userdata.electric.annotations.referenceAnnot
• userdata.electric.annotations.mapAnnot

The app can also load a previously saved session file containing:

• EGMdata

Running the app

Open MATLAB, navigate to the folder containing EGMclassifier.m, and run:

app = EGMclassifier;

This opens the graphical interface.

Basic functionality

The app includes the following main features:

Load data

• Load one or more OpenEP .mat files
• Or load a previously saved labelled session

Browse EGMs

• Select a case from the case dropdown
• Select an individual EGM from the EGM dropdown
• Move through signals using next, previous, or random buttons

Classify signals

• Define up to 8 user classes
• Assign the current EGM to a class using the class buttons
• Optionally move automatically to the next or a random signal after classification

Add annotations

• Click directly on the plotted EGM to add annotation markers
• Choose annotation types such as:
  • S1 Start / End
  • S2 Start / End
  • pacing spike
  • local activation time
  • fractionation
  • far-field signal
  • late potential
  • double potential
  • custom marker

Click sequences

• Set up a predefined sequence of annotation labels
• Step through repeated annotation workflows more quickly
• Delete sequence clicks if needed

Display signal features

Optional overlays include:

• Window of interest (WOI)
• Activation time
• Duration by deflection method
• Duration by nonlinear energy
• Duration by state machine
• Duration by sustained activity
• Duration by sliding standard deviation

Save output

The app can save the current labelled dataset as a .mat file containing:

• loaded case data
• assigned classes
• user annotations
• annotation settings
• duration settings

Output

Saved session files contain the working dataset and settings so that annotation/classification can be resumed later.

Notes

• The default sampling frequency is set to 2000 Hz
• Files that do not match the expected OpenEP structure are ignored
• Some duration methods depend on helper functions in the SubFunctions folder