Automated seizure detection for neonatal EEG

Karoliina Tapani

Automated seizure detection for neonatal EEG

Julkaisun otsikon käännös: Automaattinen epileptisten kohtausten tunnistus vastasyntyneiden aivosähkökäyrästä

Karoliina Tapani

Neurotieteen ja lääketieteellisen tekniikan laitos

Tutkimustuotos: Doctoral Thesis › Collection of Articles

Abstrakti

Sellaiset vastasyntyneet lapset, joilla on suuri riski epileptisille kohtauksille, tarvitsevat kriittistä hoitoa, sillä kohtausten tiedetään vahingoittavan kehittyviä aivoja. Kohtauksia havainnoidaan mittaamalla vastasyntyneen aivosähkökäyrää, josta neurofysiologit tunnistavat kohtaukselle ominaisia signaalipiirteitä. Vaadittavaa neurofysiologista asiantuntemusta ei kuitenkaan aina ole saatavilla, joten tutkimusten tulkintaan tarvitaan automaattisia menetelmiä. Vastasyntyneiden kohtauksia tunnistavien algoritmien kehitys on haastavaa monista syistä. Kohtauksia aivojen normaalista taustatoiminnasta erottelevien signaalipiirteiden tunnistus on vaativaa, algoritmin suorituskykyä monipuolisesti mittaavista suureista ei ole yhteistä näkemystä, ja lisäksi olemassa olevien algoritmien keskinäinen luotettava vertailu ei ole ollut mahdollista avoimien aineistojen puutteen vuoksi. Tässä väitöskirjassa kehitettiin menetelmiä vastasyntyneiden aivosähkökäyrän automaattiseen tulkintaan. Kompleksisten signaalipiirteiden ja koneoppimismenetelmien avulla epileptisiä kohtauksia saatiin erotettua taustasignaalista. Tutkimuksessa kehitetyt uudet signaalipiirteet yhdistettynä aiempien tutkimusten mukaisten signaalipiirteiden kanssa muodostavat algoritmin, joka tukivektorikoneita hyödyntäen luokittelee aivosähkökäyrän signaalia kohtauksiin ja taustasignaaliin. Tutkimuksessa kokeiltiin myös erilaisia luokittimia. Osatöissä tutkittiin muun muassa konvoluutioneuroverkon etuja signaalipiirteiden automaattisessa erottelussa, sekä erilaisia poikkeavien signaalipiirteiden erottelumenetelmiä. Lisäksi tutkittiin useita suorituskykyä mittaavia menetelmiä. Algoritmin suoriutumista vertailtiin myös kahden aineiston välillä. Asiantuntijan suorittama luokittelu toimi vertailukohtana algoritmin tuottamalle luokittelulle. Usean asiantuntijan luokitteluja toisiinsa vertailemalla saatiin referenssiarvo, jonka avulla mitattiin, onko algoritmin luokittelu ihmisasiantuntijoista poikkeava. Lisäksi tutkimuksessa vertailtiin algoritmin luokittelemien kohtausten määrää ja kestoa asiantuntijan vastaaviin luokitteluihin, joiden perusteella arvioitiin algoritmin kliinistä hyötyä. Kehityksessä käytetty aineisto ja alkuperäinen algoritmi julkaistiin lopuksi avoimeen käyttöön. Tutkimuksessa kehitetyt menetelmät erottelivat kohtauksia signaalista aiempia menetelmiä tarkemmin, ja lopullisen algoritmin suorituskyvyn taso oli kehitykseen ja testaamiseen käytetyillä aineistoilla yhtenevä. Algoritmin tuottama arvio kohtauksista oli kliinisesti merkittävä, ja luokittimen kehitykseen käytetyn aineiston diversiteettiä lisäämällä todettiin, että luokittelu ei merkittävästi poikkea asiantuntijan luokittelusta.

Julkaisun otsikon käännös	Automaattinen epileptisten kohtausten tunnistus vastasyntyneiden aivosähkökäyrästä
Alkuperäiskieli	Englanti
Pätevyys	Tohtorintutkinto
Myöntävä instituutio	Aalto-yliopisto
Valvoja/neuvonantaja	Parkkonen, Lauri, Vastuuprofessori Stevenson, Nathan, Ohjaaja, Ulkoinen henkilö Vanhatalo, Sampsa, Ohjaaja, Ulkoinen henkilö
Kustantaja	Aalto University
Painoksen ISBN	978-952-64-1080-7
Sähköinen ISBN	978-952-64-1081-4
Tila	Julkaistu - 2022
OKM-julkaisutyyppi	G5 Tohtorinväitöskirja (artikkeli)

Tutkimusalat

vastasyntyneiden EEG-monitorointi
epileptinen kohtaus
koneoppiminen
tukivektorikone
konvoluutioneuroverkko
avoin aineisto

Pääsy asiakirjaan

http://urn.fi/URN:ISBN:978-952-64-1081-4

OpenUrl-saatavuus

Koko teksti

Science-IT
Mikko Hakala (Manager)
Perustieteiden korkeakoulu
Laitteistot/tilat: Facility

Siteeraa tätä

@phdthesis{9bd9ff3332b341bd84131de061f6ad4c,

title = "Automated seizure detection for neonatal EEG",

abstract = "Epileptic seizures damage the developing brain. Therefore, newborns with high risk of seizures must be continuously monitored to ensure that seizures are detected as quickly as possible. Reliable detection of seizures requires multichannel electroencephalography recordings to be analyzed by experienced clinical experts, such as neurophysiologists. Since this expertise is not always available, there is a demand for automatic seizure detection methods. The challenge in the development of neonatal seizure detection algorithms has been threefold: 1) to define features that best represent the characteristics of neonatal seizures, 2) to find a comprehensive suite of measures to assess the goodness of the algorithm and define its potential utility in clinical practice, and 3) to overcome the lack of open-access data sets that hinders algorithm development. In this thesis, features were defined for detecting the non-stationary periodic characteristics of neonatal seizures in both time and joint time–frequency domains. These features, after excluding outliers, were integrated by a support vector machine incorporated in a neonatal seizure detection algorithm. This algorithm was compared to alternative methods of seizure detection, such as the deep convolutional neural network. To accurately assess algorithm performance, measures for generalizability, non-inferiority to the annotation of the human expert, and clinical relevance were developed. Generalizability was assessed as the consistency of performance between training (through cross-validation) and unseen independent data. Non-inferiority was estimated by comparing the inter-observer agreement of human experts only to a composite human expert/algorithm agreement. Clinical relevance measures evaluated clinically relevant interpretations of algorithm annotation such as short-term seizure burden. The developed measures proved highly discriminatory for seizures, the final algorithm generalized to unseen data, was non-inferior to human annotation with increased diversity in the training data set and provided a clinically relevant interpretation of the recordings. These performance measures comprehensively assess the goodness and clinical efficacy of the algorithm. The data set used to train the proposed algorithm as well as the original algorithm have been made openly accessible to facilitate seizure detector development and comparisons in the future.",

keywords = "neonatal EEG, EEG monitoring, seizure, machine learning, support vector machine, convolutional neural network, open-access data, vastasyntyneiden EEG-monitorointi, epileptinen kohtaus, koneoppiminen, tukivektorikone, konvoluutioneuroverkko, avoin aineisto, neonatal EEG, EEG monitoring, seizure, machine learning, support vector machine, convolutional neural network, open-access data",

author = "Karoliina Tapani",

year = "2022",

language = "English",

isbn = "978-952-64-1080-7",

series = "Aalto University publication series DOCTORAL THESES",

publisher = "Aalto University",

number = "198",

address = "Finland",

school = "Aalto University",

}

TY - THES

T1 - Automated seizure detection for neonatal EEG

AU - Tapani, Karoliina

PY - 2022

Y1 - 2022

N2 - Epileptic seizures damage the developing brain. Therefore, newborns with high risk of seizures must be continuously monitored to ensure that seizures are detected as quickly as possible. Reliable detection of seizures requires multichannel electroencephalography recordings to be analyzed by experienced clinical experts, such as neurophysiologists. Since this expertise is not always available, there is a demand for automatic seizure detection methods. The challenge in the development of neonatal seizure detection algorithms has been threefold: 1) to define features that best represent the characteristics of neonatal seizures, 2) to find a comprehensive suite of measures to assess the goodness of the algorithm and define its potential utility in clinical practice, and 3) to overcome the lack of open-access data sets that hinders algorithm development. In this thesis, features were defined for detecting the non-stationary periodic characteristics of neonatal seizures in both time and joint time–frequency domains. These features, after excluding outliers, were integrated by a support vector machine incorporated in a neonatal seizure detection algorithm. This algorithm was compared to alternative methods of seizure detection, such as the deep convolutional neural network. To accurately assess algorithm performance, measures for generalizability, non-inferiority to the annotation of the human expert, and clinical relevance were developed. Generalizability was assessed as the consistency of performance between training (through cross-validation) and unseen independent data. Non-inferiority was estimated by comparing the inter-observer agreement of human experts only to a composite human expert/algorithm agreement. Clinical relevance measures evaluated clinically relevant interpretations of algorithm annotation such as short-term seizure burden. The developed measures proved highly discriminatory for seizures, the final algorithm generalized to unseen data, was non-inferior to human annotation with increased diversity in the training data set and provided a clinically relevant interpretation of the recordings. These performance measures comprehensively assess the goodness and clinical efficacy of the algorithm. The data set used to train the proposed algorithm as well as the original algorithm have been made openly accessible to facilitate seizure detector development and comparisons in the future.

AB - Epileptic seizures damage the developing brain. Therefore, newborns with high risk of seizures must be continuously monitored to ensure that seizures are detected as quickly as possible. Reliable detection of seizures requires multichannel electroencephalography recordings to be analyzed by experienced clinical experts, such as neurophysiologists. Since this expertise is not always available, there is a demand for automatic seizure detection methods. The challenge in the development of neonatal seizure detection algorithms has been threefold: 1) to define features that best represent the characteristics of neonatal seizures, 2) to find a comprehensive suite of measures to assess the goodness of the algorithm and define its potential utility in clinical practice, and 3) to overcome the lack of open-access data sets that hinders algorithm development. In this thesis, features were defined for detecting the non-stationary periodic characteristics of neonatal seizures in both time and joint time–frequency domains. These features, after excluding outliers, were integrated by a support vector machine incorporated in a neonatal seizure detection algorithm. This algorithm was compared to alternative methods of seizure detection, such as the deep convolutional neural network. To accurately assess algorithm performance, measures for generalizability, non-inferiority to the annotation of the human expert, and clinical relevance were developed. Generalizability was assessed as the consistency of performance between training (through cross-validation) and unseen independent data. Non-inferiority was estimated by comparing the inter-observer agreement of human experts only to a composite human expert/algorithm agreement. Clinical relevance measures evaluated clinically relevant interpretations of algorithm annotation such as short-term seizure burden. The developed measures proved highly discriminatory for seizures, the final algorithm generalized to unseen data, was non-inferior to human annotation with increased diversity in the training data set and provided a clinically relevant interpretation of the recordings. These performance measures comprehensively assess the goodness and clinical efficacy of the algorithm. The data set used to train the proposed algorithm as well as the original algorithm have been made openly accessible to facilitate seizure detector development and comparisons in the future.

KW - neonatal EEG

KW - EEG monitoring

KW - seizure

KW - machine learning

KW - support vector machine

KW - convolutional neural network

KW - open-access data

KW - vastasyntyneiden EEG-monitorointi

KW - epileptinen kohtaus

KW - koneoppiminen

KW - tukivektorikone

KW - konvoluutioneuroverkko

KW - avoin aineisto

KW - neonatal EEG

KW - EEG monitoring

KW - seizure

KW - machine learning

KW - support vector machine

KW - convolutional neural network

KW - open-access data

M3 - Doctoral Thesis

SN - 978-952-64-1080-7

T3 - Aalto University publication series DOCTORAL THESES

PB - Aalto University

ER -

Automated seizure detection for neonatal EEG

Abstrakti

Tutkimusalat

Pääsy asiakirjaan

OpenUrl-saatavuus

Sormenjälki

Laitteet

Science-IT

Siteeraa tätä