Skip to main content

Detection of brain patterns for the characterisation of epileptic networks

Periodic Reporting for period 1 - EPINET (Detection of brain patterns for the characterisation of epileptic networks)

Reporting period: 2015-10-05 to 2017-10-04

Every year 2.4 million people in the world are diagnosed with epilepsy and approximately 25% of them respond poorly to drug treatment. Selected patients are offered the opportunity to improve seizure control through a surgical procedure, with overall improvement of quality of life which is is strictly dependent on the level of seizure-freedom after surgery. A good delineation of the seizure-onset zone (SOZ), which is the area responsible for the generation of the seizures, relies on the discovery of specific biomarkers of seizure propensity of specific brain regions; among them, high-frequency oscillations (HFOs) measured in intracranial electroencephalogram (iEEG) have gained much attention in the last years due to their strict correlation to the SOZ.
The EPINET project aimed at developing a better understanding of the role of HFOs in the generation of seizure and at designing tools and algorithms for a better identification of the SOZ in patients affected by drug-resistant epilepsy. Three research objectives (RO) were identified: to develop quantitative methods for the automated detection of intracranial and extracranial HFOs; to develop a reliable and robust set of methodologies for the totally non-invasive recognition of HFOs; to evaluate the role of brain oscillations in the high frequency range to delineate the epileptogenic zone.
At the end of the project the three ROs were fully achieved with the development of a complete set of routines and algorithms, named EPINETLAB, for the detection of HFOs and the identification of the SOZ. The tool, freely available upon request, is intended to support clinicians in the presurgical work-up, being user-friendly and fully documented in each single part. Moreover, a database of iEEG data from 60 patients, collected over three different European centres and of MEG data from 13 paediatric patients, collected at the Birmingham Children’s Hospital (BCH), allowed a robust validation of the implemented algorithms both with invasive and non-invasive recording technique, which was another aim of the project.
EPINET allowed the fellow to become an independent computational neuroscientist, thanks to the highly multidisciplinary nature of the activities and the expertise of the Aston University/BCH research teams and to the secondment at Micromed, an French company whose R&D unit is in Italy with a 30-year track record of development and commercialization of solutions for neurophysiology.
The fellow developed knowledge of the ethical and practical standard to which all clinical research is conducted, thanks to the collaboration with the BCH and the Aston Brain Centre, which provided her with the skills needed to understand how a clinical protocol is conducted and to run one on her own. Moreover the fellow acquired training in ethics, safety, data protection and intellectual property, very important features for the process of becoming an independent scientist. And she improved her networking background and her exposure to the epilepsy scientific community, thanks to the collaboration with different European centres and to the attendance to many national and international epilepsy-related congresses.
Main Results
The fellow designed and developed EPINETLAB, a Matlab-based software for the detection of HFOs and the identification of the SOZ. The tool was intended as a support tool for clinicians, being designed in a user-friendly and easy-to-set way. The software is released for free upon request and constitutes an innovative tool in the epilepsy research as, besides providing a novel algorithm for the HFO detection on a subset of informative channels, it provides statistical functionalities to compare the results of the identification of the SOZ with the clinically-defined SOZ, which is considered the gold standard to date.
The fellow collected an iEEG database of more than 60 patients from three different European centres involved in epilepsy research. An additional database of 13 patients recorded with both iEEG and MEG technologies was collected from the BCH and the Aston Brain Centre, Birmingham, UK. Metadata for each patient were collected as well.
In a dataset of 6 patients, the novel algorithm based on kurtosis-driven channels selection and wavelet-based detection of HFOs had average sensitivity of 81.94% and average specificity of 96.03% in identifying the HFO area which overlapped with the SOZ as defined by clinical presurgical workup. Furthermore, the kurtosis-based channel selection resulted in an average reduction in computational time of 66.60%.
In 6 pediatric patients both implanted with iEEG and recorded with MEG, HFOs were identified in the MEG virtual electrodes in 5/6 patients, which represents 83% concordance between modalities, and the sources of HFOs coincided with the SOZ at the sub-lobar level. These results show good concordance between the source locations of HFOs detected in non-invasive MEG data and those identified in intracranial recordings, which were in close proximity to the SOZ. This means that this innovative approach can be used independently from the input signal and that non-invasive localisation of HFOs using MEG could prove a useful addition to surgical planning in paediatric patients with epilepsy.

Dissemination activities
Two posters were presented at the 32nd International Epilepsy conference, Sept. 2-6 2017, Barcelona, Spain.
An oral presentation was delivered at the ILAE British chapter 2017, Leeds, UK.
The fellow delivered a presentation on Brain-Computer Interface technologies at the Aston University’s Masterclass program, a presentation on EPINETLAB during her secondment in Micromed and a presentation on her Marie-Curie project at the Aston Brain Centre, Birmingham, UK.
The fellow's activities were presented at the Aston University Research Fellowships poster campaign to celebrate funded fellows.
The fellow published two papers on peer-reviewed, high-ranked international journals (Journal of Neural Engineering and International Journal of Neural Systems) and submitted a paper to BMC Bioinformatics. Additional journal papers are in preparation.
The EPINET project has provided the scientific community with a robust and efficient software tool which can aid clinicians in the identification of the SOZ. The availability of such tools and the results obtained in the analysis of both invasive and non-invasive data, provides a significant contribution to the assessment of patients with drug resistant epilepsy and to epilepsy research in Europe.
The project has allowed the fellow to establish a solid network of collaborations between different institutions involved in epilepsy research; this is really important for the sharing of data and knowledge dissemination and for facilitating large-scale collaborative studies within the EU and contributing to improved surgical treatment of epileptic patients.
A clinical trial involving epilepsy surgery centres in Italy, USA and UK is being developed. The development of methods to isolate HFOs in MEG signal has also given impetus to a collaboration with dr. Papadelis from Harvard University.
The collaboration with the industrial partner (Micromed) and the possibility to integrate the HFO analysis tools in their diagnostic software package has paved the way to a cross-fertilisation between academia and industry, a strategic objective across the EU and will allow consolidation of European leadership in the field.
Main functionalities provided by the EPINETLAB software