Machine Learning Techniques show Sensory and Association Network Alterations in Severe Epilepsy
-
1
University of Melbourne, Florey Department of Neuroscience and Mental Health, Australia
-
2
Florey Institute of Neuroscience and Mental Health, Australia
-
3
University of Melbourne, Department of Medicine, Australia
-
4
Florey Department of Neuroscience and Mental Health, Australia
Background: Lennox-Gastaut Syndrome is a severe epilepsy syndrome where cognitive dysfunction and intellectual disability is usually present (1). Our previous work suggests that LGS is a 'network syndrome' (2). To test this hypothesis, we used two voxel-based metrics of functional connectivity to assess network alterations in LGS. The first, Regional Homogeneity, is a measure of local connectivity (3). The second, Eigenvector Centrality, is a measure of global connectivity (i.e., hubs) (4). The machine learning approach, Multivariate Pattern Analysis (MVPA) (5), was used assess between group differences.
Methods: Nine adult subjects with LGS as well as 14 controls were included. Over 20 minutes of task-free functional MRI data was used. The data was slice-timed, realigned, segmented and normalised. White matter and cerebrospinal fluid signal was regressed out. Further, the data was smoothed, detrended and filtered (0.01-0.08Hz). Images with head movement above 0.5 mm were regressed out using estimates from the Friston 24 parameters. Traditional two sample t-tests (corrected) were employed in addition to MVPA.
Results: MVPA distinguished local and global connectivity features of LGS from controls with 95.7% accuracy (22/23 subjects were correctly classified). Machine learning features identified increased global connectivity in association cortices in LGS, areas that were not detected with univariate statistics. Regions of local network decreases in LGS, compared to controls, included superior temporal, peri-central and medial occipital cortices (i.e., auditory, motor and visual cortices) as well as medial frontal areas.
Conclusion: LGS appear to be a neural network syndrome affecting sensory and association circuits. These network differences were readily discernible using machine learning.
References:
1. Dulac, O. (1993), Epilepsia,
2. Pillay, N. (2013), Neurology.
3. Zang, Y. (2004), NeuroImage.
4. Wink, A. (2012), Brain Conn.
5. Schrouff, J. (2013), Neuroinformatics.
Keywords:
Epilepsy,
machine learning,
functional MRI,
functional connectivity,
connectomics,
neural networks,
graph theory
Conference:
XII International Conference on Cognitive Neuroscience (ICON-XII), Brisbane, Queensland, Australia, 27 Jul - 31 Jul, 2014.
Presentation Type:
Poster
Topic:
Methods Development
Citation:
Pedersen
M,
Curwood
EK,
Archer
JS,
Abbott
DF and
Jackson
GD
(2015). Machine Learning Techniques show Sensory and Association Network Alterations in Severe Epilepsy.
Conference Abstract:
XII International Conference on Cognitive Neuroscience (ICON-XII).
doi: 10.3389/conf.fnhum.2015.217.00260
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
19 Feb 2015;
Published Online:
24 Apr 2015.
*
Correspondence:
Prof. Graeme D Jackson, University of Melbourne, Department of Medicine, Melbourne, Australia, g.jackson@brain.org.au