DHIF Imen

PhD student at Sorbonne University
Team : SYEL
https://lip6.fr/Imen.Dhif

Supervision : Patrick GARDA

Co-supervision : HACHICHA Khalil, PINNA Andrea

Compression, analyze and visualization of EEG signals applied to telemedecine

The field of EEG signal processing is progressing very fast in recent years. By using telemedicine, EEG recordings can be rapidly transmitted between hospitals and healthcare centers. Due to the large amount of EEG acquired over several days to detect pathologies such as epilepsy, an efficient compression technique is necessary. Besides, the lack of experts and the short duration of epileptic seizures require the automatic detection of these seizures. Furthermore, a uniform viewer is mandatory to ensure interoperability and a correct reading of transmitted EEG exams. As a certified medical image compression technique, WAAVES coder provides high compression ratios while ensuring image quality. During our thesis, three challenges are revealed : the first consists in adapting the WAAVES encoder to the compression of the EEG signals. The second challenge lies in help diagnosing epilepsy by detecting automatically epileptic seizures in an EEG signal. The last objective is to ensure the interoperability of the different displays of EEG exams. We first studied the characteristics of the EEG signals, as well as the different blocks of the WAAVES coder. One of the main challenges is that WAAVES is unable to remove spatial correlation between EEG signals, to compress directly mono-dimensional signals and to accept as input decimal values. Therefore, our approach is based on the application of Independent Component Analysis to decorrelate signals, a scaling to resize decimal values, and image construction. To keep a diagnostic quality with a PDR less than 7%, we coded the residue and added it to the coded signals. The proposed compression algorithm EEGWaaves has achieved compression ratios equal to 56. Subsequently, we aimed to couple the compression and epileptic seizure patterns by reusing blocks of EEGWaaves. We proposed a new method of EEG feature extraction based on a new calculation model of the energy expected measurement (EAM) of EEG signals. Then, statistical parameters were calculated to reduce the feature vectors and finally, Neural Networks were applied to classify and detect epileptic seizures. Our method allowed us to achieve a better sensitivity up to 100%, a specificity of 98.89% and an accuracy of 99.44%. The last chapter details the deployment of our multi-platform display of physiological signals by meeting the specifications established by doctors. The main role of this software is to ensure the interoperability and exchange of EEG exams between the different healthcare centers.

Defence : 12/13/2017

Jury members :

Roger Reyanud, Université Paris-Sud [Rapporteur]
Dan Istrate, Université de Technologie de Compiègne [Rapporteur]
Marie-Christine Jaulent, Institut national de la santé et de la recherche médicale INSERM
Sylvain Hochberg, CEO, Cira
Didier Heudes, Praticien Hospitalier Université Paris 5 René Descartes
Khalil Hachicha, Université Pierre et Marie Curie
Andrea Pinna, Université Pierre et Marie Curie
Patrick Garda, Université Pierre et Marie Curie

Departure date : 12/25/2018

2014-2017 Publications