Page 186 - MEGIN Book Of Abstracts - 2023

magnetic resonance imaging (MRI) findings can aid ABSTRACT Background: Epilepsy is a neurologi-
the diagnosis of intractable epilepsy caused by organic cal disorder which is characterised by recurrent and
brain lesions. involuntary seizures. Magnetoencephalography (MEG)
is clinically used as a presurgical tool in locating the
METHODS This study included 51 patients with epileptogenic zone by localising either interictal epilep-
epilepsy who had MEG clusters but whose initial MRI tic discharges (IEDs) or ictal activities. The localisation of
findings were interpreted as being negative for organic ictal onset provides reliable and more accurate seizure
lesions. Three board-certified radiologists reinterpreted onset zones rather than localising the IEDs. Ictals or
the MRI findings, utilizing the MEG findings as a guide. seizures are presently detected during MEG analysis by
The degree to which the reinterpretation of the imag- manually inspecting the recorded data. This is laborious
ing results identified an organic lesion was rated on a when the duration of recordings is longer. Methods: We
5-point scale. propose a novel method which uses statistical fea-
tures such as short-time permutation entropy (STPE),
RESULTS Reinterpretation of the MRI data with MEG gradient of STPE (GSTPE), short-time energy (STE) and
guidance helped detect an abnormality by at least one short-time mean (STM) extracted from the ictal and
radiologist in 18 of the 51 patients (35.2%) with symp- interictal MEG data of drug resistant epilepsy patients
tomatic localization-related epilepsy. A surgery was group. Since the data is heavily skewed, the RUSBoost
performed in 7 of the 51 patients, and histopathologi- algorithm with k-fold cross-validation is used to classify
cal analysis results identified focal cortical dysplasia in 5 the data into ictal and interictal by using the four fea-
patients (Ia: 1, IIa: 2, unknown: 2), hippocampal sclerosis ture vectors. This method is further used for localising
in 1 patient, and dysplastic neurons/gliosis in 1 patient. the epileptogenic region using region-specific classifi-
cations by means of the RUSBoost algorithm. Results:
CONCLUSIONS The results of this study highlight the The accuracy obtained for seizure detection is 93.4%.
potential diagnostic applications of MEG to detect The specificity and sensitivity for the same are 93%. The
organic epileptogenic lesions, particularly when radio- localisation accuracies for each lobe are in the range
logical visualization is difficult with MRI alone. of 88.1-99.1%. Discussion: Through this ictus detection
method, the current scenario of laborious inspection
Keywords: Epilepsy, Epileptogenic zone, Focal cortical dys- of the ictal MEG can be reduced. The proposed system,
plasia, Magnetic resonance imaging, Magnetoencepha- thus, can be implemented in real-time as a better and
lography more efficient method for seizure detection and further
it can prove to be highly beneficial for patients and
Epilepsy & behavior: E&B (2021), Vol. 114, No. Pt A health-care professionals during real-time MEG record-
(33323336) (1 citation) ing. Furthermore, the identification of the epileptogen-
ic lobe can provide clinicians with useful insights, and a
pre-cursor for source localisation.
Seizure detection and epileptogenic zone
localisation on heavily skewed MEG data using Keywords: EEG, MEG, RUSBoost, artificial neural networks,
RUSBoost machine learning technique (2022) classification, epileptogenic zone, ictal, inter-ictal, ma-
chine learning, permutation entropy, seizure detection
Bhanot, Nipun; Mariyappa, N; Anitha, H; Bhargava, G K;
Velmurugan, J; Sinha, Sanjib The International journal of neuroscience (2022), Vol. 132,
No. 10 (33272081) (1 citation)
Manipal Institute of Technology, Electronics and Commu-
nication, Manipal, India; Neurology, NIMHANS, Bangalore,
India; NIMHANS, Bengaluru, India

 ontents  Index 165
C