Artificial intelligence for classification of temporal lobe epilepsy with ROI-level MRI data: A worldwide ENIGMA-Epilepsy study

ENIGMA-Epilepsy Working Group

doi:10.1016/j.nicl.2021.102765

Artificial intelligence for classification of temporal lobe epilepsy with ROI-level MRI data: A worldwide ENIGMA-Epilepsy study

ENIGMA-Epilepsy Working Group

Research output: Contribution to journal › Article › peer-review

23 Citations (Scopus)

Abstract

Artificial intelligence has recently gained popularity across different medical fields to aid in the detection of diseases based on pathology samples or medical imaging findings. Brain magnetic resonance imaging (MRI) is a key assessment tool for patients with temporal lobe epilepsy (TLE). The role of machine learning and artificial intelligence to increase detection of brain abnormalities in TLE remains inconclusive. We used support vector machine (SV) and deep learning (DL) models based on region of interest (ROI-based) structural (n = 336) and diffusion (n = 863) brain MRI data from patients with TLE with (“lesional”) and without (“non-lesional”) radiographic features suggestive of underlying hippocampal sclerosis from the multinational (multi-center) ENIGMA-Epilepsy consortium. Our data showed that models to identify TLE performed better or similar (68–75%) compared to models to lateralize the side of TLE (56–73%, except structural-based) based on diffusion data with the opposite pattern seen for structural data (67–75% to diagnose vs. 83% to lateralize). In other aspects, structural and diffusion-based models showed similar classification accuracies. Our classification models for patients with hippocampal sclerosis were more accurate (68–76%) than models that stratified non-lesional patients (53–62%). Overall, SV and DL models performed similarly with several instances in which SV mildly outperformed DL. We discuss the relative performance of these models with ROI-level data and the implications for future applications of machine learning and artificial intelligence in epilepsy care.

Original language	English
Article number	102765
Journal	NeuroImage: Clinical
Volume	31
DOIs	https://doi.org/10.1016/j.nicl.2021.102765
Publication status	Published - Jan 2021

Keywords

Artificial inteligence
Epilepsy
Machine learning
Temporal lobe epilepsy

Access to Document

10.1016/j.nicl.2021.102765

Cite this

@article{db156f8fae914243b6aaa0b6e08f1700,

title = "Artificial intelligence for classification of temporal lobe epilepsy with ROI-level MRI data: A worldwide ENIGMA-Epilepsy study",

abstract = "Artificial intelligence has recently gained popularity across different medical fields to aid in the detection of diseases based on pathology samples or medical imaging findings. Brain magnetic resonance imaging (MRI) is a key assessment tool for patients with temporal lobe epilepsy (TLE). The role of machine learning and artificial intelligence to increase detection of brain abnormalities in TLE remains inconclusive. We used support vector machine (SV) and deep learning (DL) models based on region of interest (ROI-based) structural (n = 336) and diffusion (n = 863) brain MRI data from patients with TLE with (“lesional”) and without (“non-lesional”) radiographic features suggestive of underlying hippocampal sclerosis from the multinational (multi-center) ENIGMA-Epilepsy consortium. Our data showed that models to identify TLE performed better or similar (68–75%) compared to models to lateralize the side of TLE (56–73%, except structural-based) based on diffusion data with the opposite pattern seen for structural data (67–75% to diagnose vs. 83% to lateralize). In other aspects, structural and diffusion-based models showed similar classification accuracies. Our classification models for patients with hippocampal sclerosis were more accurate (68–76%) than models that stratified non-lesional patients (53–62%). Overall, SV and DL models performed similarly with several instances in which SV mildly outperformed DL. We discuss the relative performance of these models with ROI-level data and the implications for future applications of machine learning and artificial intelligence in epilepsy care.",

keywords = "Artificial inteligence, Epilepsy, Machine learning, Temporal lobe epilepsy",

author = "{ENIGMA-Epilepsy Working Group} and Ezequiel Gleichgerrcht and Munsell, {Brent C.} and Saud Alhusaini and Alvim, {Marina K.M.} and N{\'u}ria Bargall{\'o} and Benjamin Bender and Andrea Bernasconi and Neda Bernasconi and Boris Bernhardt and Karen Blackmon and Caligiuri, {Maria Eugenia} and Fernando Cendes and Luis Concha and Desmond, {Patricia M.} and Orrin Devinsky and Doherty, {Colin P.} and Martin Domin and Duncan, {John S.} and Focke, {Niels K.} and Antonio Gambardella and Bo Gong and Renzo Guerrini and Hatton, {Sean N.} and Reetta K{\"a}lvi{\"a}inen and Keller, {Simon S.} and Peter Kochunov and Raviteja Kotikalapudi and Kreilkamp, {Barbara A.K.} and Angelo Labate and Soenke Langner and Sara Larivi{\`e}re and Matteo Lenge and Elaine Lui and Pascal Martin and Mario Mascalchi and Stefano Meletti and O'Brien, {Terence J.} and Pardoe, {Heath R.} and Pariente, {Jose C.} and {Xian Rao}, Jun and Richardson, {Mark P.} and Ra{\'u}l Rodr{\'i}guez-Cruces and Theodor R{\"u}ber and Ben Sinclair and Hamid Soltanian-Zadeh and Stein, {Dan J.} and Pasquale Striano and Taylor, {Peter N.} and Thomas, {Rhys H.} and {Elisabetta Vaudano}, Anna",

note = "Funding Information: S.La. is funded by CIHR. P.M. was supported by the PATE program (F1315030) of the University of T{\"u}bingen. S.M. is supported by Italian Ministry of Health funding grant NET-2013-02355313. T.J. is supported by NHMRC Program Grant. M.R. is supported by Medical Research Council programme grant (MR/K013998/1); Medical Research Council Centre for Neurodevelopmental Disorders (MR/N026063/1); NIHR Biomedical Research Centre at South London and Maudsely NHS Foundation Trust. R.R.-C. is supported by the Fonds de recherche du Qu{\'e}bec – Sant{\'e} (FRQS-291486). D.J.S. is supported by SA Medical Research Council. Work developed within the framework of the DINOGMI Department of Excellence of MIUR 2018–2022 (legge 232 del 2016). R.H.T. is supported by Epilepsy Research UK. C.L.Y. is supported by FAPESP - BRAINN (2013/07599-3); CNPQ (403726/2016-6). J.S.Z. is supported by National Nature Science Foundation of China (No. 61772440). C.R.M. is supported by NIH R01 NS065838; R21 NS107739. L.B. is supported by R01NS110347 (NIH/NINDS). A.A. holds an MRC eMedLab Medical Bioinformatics Career Development Fellowship; this work was partly supported by the Medical Research Council [grant number MR/L016311/ 1]. R.W. received support from the Swiss League Against Epilepsy. Core funding for ENIGMA was provided by the NIH Big Data to Knowledge (BD2K) program under consortium grant U54 EB020403. The Bern research centre was funded by Swiss National Science Foundation (grant 180365). This work was partly undertaken at UCLH/UCL, which received a proportion of funding from the Department of Health{\textquoteright}s NIHR Biomedical Research Centres funding scheme. The work was also supported by the Epilepsy Society, UK. We are grateful to the Wolfson Trust and the Epilepsy Society for supporting the Epilepsy Society MRI scanner. The UNICAMP research centre was funded by FAPESP (S{\~a}o Paulo Research Foundation); Contract grant number: 2013/07559-3. Lastly, we are grateful for software development and high-performance computing work performed by UNC Chapel Hill graduate research assistant Kyuyeon Kim. Funding Information: B.C.M. is supported by NINDS R21 NS107739-01A1. M.K.M.A. is supported by FAPESP 15/17066-0. A.B. is supported by CIHR MOP-57840. N.Be. is supported by CIHR MOP-123520; CIHR MOP-130516. B.Ber. acknowledges research support from NSERC (Discovery-1304413), CIHR (FDN-154298), Azrieli Center for Autism Research of the Montreal Neurological Institute (ACAR), SickKids Foundation (NI17-039), and salary support from FRQS (Chercheur Boursier Junior 1). L.C. is supported by Mexican Council of Science and Technology (CONACYT 181508, 232676, 251216, and 280283); UNAM-DGAPA (IB201712). O.D. is supported by Finding A Cure for Epilepsy and Seizures (FACES). J.S.D. is supported by NIHR. Funding Information: S.S.K. is supported by Medical Research Council (MR/S00355X/1 and MR/K023152/1) and Epilepsy Research UK (1085). P.K. is supported by S10OD023696; R01EB015611. Funding Information: The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: B.Ben. is the cofounder of AIRAmed GmbH, a company that offers brain segmentation. N.K.F. holds Honoraria from Bial, Eisai, Philips/EGI, UCB. D.J.S. has received research grants or consultance honoraria from Johnson & Johnson, Lundbeck, Servier, and Takeda. P.S. received speaker fees from and is on advisory boards for Biomarin, Zogenyx, GW Pharmaceuticals; research funding by ENECTA BV, GW Pharmaceuticals, Kolfarma srl., Eisai. P.M.T. received a research grant from Biogen, Inc., and was a paid consultant for Kairos Venture Capital, Inc., USA, for projects unrelated to this work. M.G. received fees and travel support from Bial Pharmaceutical and Netl{\'e} Health Science outside the submitted work. Funding Information: N.K.F. is supported by DFG FO750/5-1. R.K{\"a}. is supported by Saastamoinen Foundation. Publisher Copyright: {\textcopyright} 2021 The Authors Copyright: Copyright 2021 Elsevier B.V., All rights reserved.",

year = "2021",

month = jan,

doi = "10.1016/j.nicl.2021.102765",

language = "English",

volume = "31",

journal = "NeuroImage: Clinical",

issn = "2213-1582",

}

TY - JOUR

T1 - Artificial intelligence for classification of temporal lobe epilepsy with ROI-level MRI data

T2 - A worldwide ENIGMA-Epilepsy study

AU - ENIGMA-Epilepsy Working Group

AU - Gleichgerrcht, Ezequiel

AU - Munsell, Brent C.

AU - Alhusaini, Saud

AU - Alvim, Marina K.M.

AU - Bargalló, Núria

AU - Bender, Benjamin

AU - Bernasconi, Andrea

AU - Bernasconi, Neda

AU - Bernhardt, Boris

AU - Blackmon, Karen

AU - Caligiuri, Maria Eugenia

AU - Cendes, Fernando

AU - Concha, Luis

AU - Desmond, Patricia M.

AU - Devinsky, Orrin

AU - Doherty, Colin P.

AU - Domin, Martin

AU - Duncan, John S.

AU - Focke, Niels K.

AU - Gambardella, Antonio

AU - Gong, Bo

AU - Guerrini, Renzo

AU - Hatton, Sean N.

AU - Kälviäinen, Reetta

AU - Keller, Simon S.

AU - Kochunov, Peter

AU - Kotikalapudi, Raviteja

AU - Kreilkamp, Barbara A.K.

AU - Labate, Angelo

AU - Langner, Soenke

AU - Larivière, Sara

AU - Lenge, Matteo

AU - Lui, Elaine

AU - Martin, Pascal

AU - Mascalchi, Mario

AU - Meletti, Stefano

AU - O'Brien, Terence J.

AU - Pardoe, Heath R.

AU - Pariente, Jose C.

AU - Xian Rao, Jun

AU - Richardson, Mark P.

AU - Rodríguez-Cruces, Raúl

AU - Rüber, Theodor

AU - Sinclair, Ben

AU - Soltanian-Zadeh, Hamid

AU - Stein, Dan J.

AU - Striano, Pasquale

AU - Taylor, Peter N.

AU - Thomas, Rhys H.

AU - Elisabetta Vaudano, Anna

N1 - Funding Information: S.La. is funded by CIHR. P.M. was supported by the PATE program (F1315030) of the University of Tübingen. S.M. is supported by Italian Ministry of Health funding grant NET-2013-02355313. T.J. is supported by NHMRC Program Grant. M.R. is supported by Medical Research Council programme grant (MR/K013998/1); Medical Research Council Centre for Neurodevelopmental Disorders (MR/N026063/1); NIHR Biomedical Research Centre at South London and Maudsely NHS Foundation Trust. R.R.-C. is supported by the Fonds de recherche du Québec – Santé (FRQS-291486). D.J.S. is supported by SA Medical Research Council. Work developed within the framework of the DINOGMI Department of Excellence of MIUR 2018–2022 (legge 232 del 2016). R.H.T. is supported by Epilepsy Research UK. C.L.Y. is supported by FAPESP - BRAINN (2013/07599-3); CNPQ (403726/2016-6). J.S.Z. is supported by National Nature Science Foundation of China (No. 61772440). C.R.M. is supported by NIH R01 NS065838; R21 NS107739. L.B. is supported by R01NS110347 (NIH/NINDS). A.A. holds an MRC eMedLab Medical Bioinformatics Career Development Fellowship; this work was partly supported by the Medical Research Council [grant number MR/L016311/ 1]. R.W. received support from the Swiss League Against Epilepsy. Core funding for ENIGMA was provided by the NIH Big Data to Knowledge (BD2K) program under consortium grant U54 EB020403. The Bern research centre was funded by Swiss National Science Foundation (grant 180365). This work was partly undertaken at UCLH/UCL, which received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme. The work was also supported by the Epilepsy Society, UK. We are grateful to the Wolfson Trust and the Epilepsy Society for supporting the Epilepsy Society MRI scanner. The UNICAMP research centre was funded by FAPESP (São Paulo Research Foundation); Contract grant number: 2013/07559-3. Lastly, we are grateful for software development and high-performance computing work performed by UNC Chapel Hill graduate research assistant Kyuyeon Kim. Funding Information: B.C.M. is supported by NINDS R21 NS107739-01A1. M.K.M.A. is supported by FAPESP 15/17066-0. A.B. is supported by CIHR MOP-57840. N.Be. is supported by CIHR MOP-123520; CIHR MOP-130516. B.Ber. acknowledges research support from NSERC (Discovery-1304413), CIHR (FDN-154298), Azrieli Center for Autism Research of the Montreal Neurological Institute (ACAR), SickKids Foundation (NI17-039), and salary support from FRQS (Chercheur Boursier Junior 1). L.C. is supported by Mexican Council of Science and Technology (CONACYT 181508, 232676, 251216, and 280283); UNAM-DGAPA (IB201712). O.D. is supported by Finding A Cure for Epilepsy and Seizures (FACES). J.S.D. is supported by NIHR. Funding Information: S.S.K. is supported by Medical Research Council (MR/S00355X/1 and MR/K023152/1) and Epilepsy Research UK (1085). P.K. is supported by S10OD023696; R01EB015611. Funding Information: The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: B.Ben. is the cofounder of AIRAmed GmbH, a company that offers brain segmentation. N.K.F. holds Honoraria from Bial, Eisai, Philips/EGI, UCB. D.J.S. has received research grants or consultance honoraria from Johnson & Johnson, Lundbeck, Servier, and Takeda. P.S. received speaker fees from and is on advisory boards for Biomarin, Zogenyx, GW Pharmaceuticals; research funding by ENECTA BV, GW Pharmaceuticals, Kolfarma srl., Eisai. P.M.T. received a research grant from Biogen, Inc., and was a paid consultant for Kairos Venture Capital, Inc., USA, for projects unrelated to this work. M.G. received fees and travel support from Bial Pharmaceutical and Netlé Health Science outside the submitted work. Funding Information: N.K.F. is supported by DFG FO750/5-1. R.Kä. is supported by Saastamoinen Foundation. Publisher Copyright: © 2021 The Authors Copyright: Copyright 2021 Elsevier B.V., All rights reserved.

PY - 2021/1

Y1 - 2021/1

N2 - Artificial intelligence has recently gained popularity across different medical fields to aid in the detection of diseases based on pathology samples or medical imaging findings. Brain magnetic resonance imaging (MRI) is a key assessment tool for patients with temporal lobe epilepsy (TLE). The role of machine learning and artificial intelligence to increase detection of brain abnormalities in TLE remains inconclusive. We used support vector machine (SV) and deep learning (DL) models based on region of interest (ROI-based) structural (n = 336) and diffusion (n = 863) brain MRI data from patients with TLE with (“lesional”) and without (“non-lesional”) radiographic features suggestive of underlying hippocampal sclerosis from the multinational (multi-center) ENIGMA-Epilepsy consortium. Our data showed that models to identify TLE performed better or similar (68–75%) compared to models to lateralize the side of TLE (56–73%, except structural-based) based on diffusion data with the opposite pattern seen for structural data (67–75% to diagnose vs. 83% to lateralize). In other aspects, structural and diffusion-based models showed similar classification accuracies. Our classification models for patients with hippocampal sclerosis were more accurate (68–76%) than models that stratified non-lesional patients (53–62%). Overall, SV and DL models performed similarly with several instances in which SV mildly outperformed DL. We discuss the relative performance of these models with ROI-level data and the implications for future applications of machine learning and artificial intelligence in epilepsy care.

AB - Artificial intelligence has recently gained popularity across different medical fields to aid in the detection of diseases based on pathology samples or medical imaging findings. Brain magnetic resonance imaging (MRI) is a key assessment tool for patients with temporal lobe epilepsy (TLE). The role of machine learning and artificial intelligence to increase detection of brain abnormalities in TLE remains inconclusive. We used support vector machine (SV) and deep learning (DL) models based on region of interest (ROI-based) structural (n = 336) and diffusion (n = 863) brain MRI data from patients with TLE with (“lesional”) and without (“non-lesional”) radiographic features suggestive of underlying hippocampal sclerosis from the multinational (multi-center) ENIGMA-Epilepsy consortium. Our data showed that models to identify TLE performed better or similar (68–75%) compared to models to lateralize the side of TLE (56–73%, except structural-based) based on diffusion data with the opposite pattern seen for structural data (67–75% to diagnose vs. 83% to lateralize). In other aspects, structural and diffusion-based models showed similar classification accuracies. Our classification models for patients with hippocampal sclerosis were more accurate (68–76%) than models that stratified non-lesional patients (53–62%). Overall, SV and DL models performed similarly with several instances in which SV mildly outperformed DL. We discuss the relative performance of these models with ROI-level data and the implications for future applications of machine learning and artificial intelligence in epilepsy care.

KW - Artificial inteligence

KW - Epilepsy

KW - Machine learning

KW - Temporal lobe epilepsy

UR - http://www.scopus.com/inward/record.url?scp=85111563148&partnerID=8YFLogxK

U2 - 10.1016/j.nicl.2021.102765

DO - 10.1016/j.nicl.2021.102765

M3 - Article

AN - SCOPUS:85111563148

SN - 2213-1582

VL - 31

JO - NeuroImage: Clinical

JF - NeuroImage: Clinical

M1 - 102765

ER -

Artificial intelligence for classification of temporal lobe epilepsy with ROI-level MRI data: A worldwide ENIGMA-Epilepsy study

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this