Tools

We provide certain tools to encourage reproducibility and consistency of results reported in the field of automated seizure detection algorithm.

epilepsy2bids

#

A library to convert datasets to BIDS and to manipulate BIDS files.

esl-epfl/epilepsy2bids

Python library to convert EEG datasets to a BIDS compatible dataset

Python

18

4

Library for converting EEG datasets of people with epilepsy to EEG-BIDS compatible datasets. These datasets comply with the ILAE and IFCN minimum recording standards. They provide annotations that are HED-SCORE compatible. The datasets are formatted to be operated by the SzCORE seizure validation framework.

The library provides tools to:

• Convert EEG datasets to BIDS.
• Load and manipulate EDF files.
• Load and manipulate seizure annotation files.

Currently, the following datasets are supported:

• PhysioNet CHB-MIT Scalp EEG Database v1.0.0
• KULeuven SeizeIT1
• Siena Scalp EEG Database v1.0.0
• TUH EEG Seizure Corpus

timescoring

#

esl-epfl/timescoring

Lib for event and sample based performance metrics

Python

24

5

We built a library that provides different scoring methodologies to compare a reference time series with binary annotation (ground-truth annotations of the neurologist) to hypothesis binary annotations (provided by a machine learning pipeline). These different scoring methodologies provide a count of correctly identified events (True Positives) as well as missed events (False Negatives) and wrongly marked events (False positions)

In more details, we measure performance on the level of:

• Samples : Performance metric that threats every label sample independently.
• Events (e.g. epileptic seizure) : Classifies each event in both reference and hypothesis based on overlap of both.

Both methods are illustrated in the following figures :

szcore-evaluation

#

esl-epfl/szcore-evaluation

Compare szCORE compliant annotations of EEG datasets of people with epilelpsy

Python

5

2

The library provides a single function to evaluate a set of annotations.

def evaluate_dataset(
    reference: Path, hypothesis: Path, outFile: Path, avg_per_subject=True
) -> dict:
    """
    Compares two sets of seizure annotations accross a full dataset.

    Parameters:
    reference (Path): The path to the folder containing the reference TSV files.
    hypothesis (Path): The path to the folder containing the hypothesis TSV files.
    outFile (Path): The path to the output JSON file where the results are saved.
    avg_per_subject (bool): Whether to compute average scores per subject or
                            average across the full dataset.

    Returns:
    dict. return the evaluation result. The dictionary contains the following
          keys: {'sample_results': {'sensitivity', 'precision', 'f1', 'fpRate',
                    'sensitivity_std', 'precision_std', 'f1_std', 'fpRate_std'},
                 'event_results':{...}
                 }
    """

sz-validation-framework

#

esl-epfl/sz-validation-framework

Framework for the validation of EEG based automated seizure detection algorithms

Python

26

2

Example code that uses the framework for the validation of EEG based automated seizure detection algorithms.

The repository provides code to :

1. Convert EDF files from most open scalp EEG datasets of people with epilepsy to a standardized format
2. Convert seizure annotations from these datasets to a standardized format.
3. Evaluate the performance of seizure detection algorithm.

szcore

#

esl-epfl/szcore

This repository hosts an open seizure detection benchmarking platform.

JavaScript

20

6

Repository that implements a Continuous Integration pipeline for the evaluation of seizure detection algorithms. This is the repository used to populate the benchmark page.