Data science
Aura NGO dedicates one of its work areas to the collection and analysis of physiological data related to epilepsy. These data are the entry point for the development and evaluation of systems for the detection or even forecasting of seizures in real life.
You can find all our most recent productions below.
ECG signal quality analysis library
Octobre 2022 aux JFE
The evaluation of the ECG signal quality is essential for real life monitoring as it allows to anticipate the loss, the bad positioning of an electrode or an anomaly that would make the seizure detection system inoperative. We have developed an open source software brick that monitors this signal quality in real time.
Visualisation de données multidimensionnelles permettant un suivi personnalisé de personnes épileptiques à l'Institut La Teppe
Octobre 2022 aux JFE
La visualisation simple, interactive, rapide et actualisée de l’ensemble des données médicales est un enjeu majeur pour assurer un suivi personnalisé et dynamique d’une personne souffrant d'épilepsie. Le travail collaboratif Aura / La Teppe de recueil, de standardisation (format FHIR), et de visualisation de données (outil Grafana) a été présenté lors des JFE.
Des outils de l'Intelligence Artificielle explicable au service de la détection des crises d'épilepsie
2022
Comment l'explicabilité peut contribuer au développement d'une solution de machine learning et à la validation de sa robustesse ? Voici la question à laquelle des étudiants de datascience de CentraleSupélec Paris ont tenté de répondre avec nous. Ils ont réalisé un poster de recherche décrivant un premier prototype d'outil.
Librairie d'analyse de la qualité du signal ECG
2020 - 2021
L'évaluation de la qualité du signal ECG est indispensable pour des suivis dans des conditions de vie réelle car elle permet d'anticiper la perte, le mauvais positionnement d'une électrode ou bien une anomalie qui rendrait inopérant le système de détection de crise. Nous avons développé une brique logicielle open source qui monitore en temps réel la qualité du signal.
Exploration of different Machine Learning & Deep Learning models for seizure detection
2020
As part of their final year project, several groups of students from the Jehda datascience course, under the guidance of mentors, evaluated models for the detection of epileptic seizures based on heart rate variability. This work rely on the open data database of Temple University.
Benchmark of QRS complex detection algorithms on the ECG signal
2020
The detection of QRS complexes is an essential step to calculate the heart rate of an individual from an ECG signal. We evaluated 15 different open source implementations of QRS detectors based on 6 ECG databases labelled by cardiologists.
Automatic search for epileptic seizures in a video-EEG database
October 2019
Automatic search for epileptic seizures in a video-EEG database
October 2019
« les Journées Françaises de l'Épilepsie», (the annual French conference of neurologists specializing in epilepsy) were held in Paris at the beginning of October 2019 and we took this opportunity to present the first results of our work. This first poster concerns the search for epileptic seizures in a video-EEG database based on annotations.
Heart rate variability analysis library
October 2018
A number of scientific publications mention the cardiac signal as a relevant physiological marker for detecting or even forecasting seizures. In particular, the variability of the heart rate seems to be of particular interest. We have developed a Python library for the analysis of heart signal variability.
Physiological data collection platform
2018
AURA aims to help patients in their everyday life. It is therefore necessary, in addition to studies conducted in the hospital, to collect data in real life, referred to as "ecological environment" by the medical profession. We have developed a real life data collection and visualization platform, based on open source components.