Machine Learning for Healthcare

Joel Jaskari

Machine Learning for Healthcare

Julkaisun otsikon käännös: Koneoppiminen terveydenhuollossa

Joel Jaskari

Tietotekniikan laitos

Tutkimustuotos: Doctoral Thesis › Collection of Articles

Abstrakti

Koneoppimista on lähiaikoina ehdotettu moniin lääketieteellisiin tehtäviin. Erityisesti syvillä neuroverkoilla on saavutettu erinomaisia tuloksia luokittelutehtävissä. Monet tutkimukset ovat kuitenkin käyttäneet yksinkertaistettuja luokittelujärjestelmiä, kuten lähetteen vaativuuden ennustaminen diabeettisen retinopatian tapauksessa. Kliinisiä vakavuusasteikkoja käyttävät tutkimukset eivät ole arvioineet luokittimien epävarmuutta, joka on tärkää tietää niiden soveltamiseen lääketieteen saralla. Automaattisten segmentaatioalgoritmien kattavaa analyysiä ja vertailua useamman radiologin väliseen vaihteluun ei ole lisäksi tehty joillekkin vaativille tehtäville, kuten mandibulaarikanavan automaattiselle segmentaatiolle. Koneoppimisalgoritmeja tulisi pystyä kouluttamaan sairaalaympäristön omalla aineistolla, joka saattaa olla haastavaa aineiston määrän suhteen. Tämä väitöstyö käsittelee koneoppimisen soveltamista moniin lääketieteellisiin tehtäviin suomalaisten sairaala-aineistojen avulla. Syviä konvoluutioneuroverkkoja sovelletaan diabeettisen retinopatian ja makulaturvotuksen luokitteluun kliinisesti käytetyillä vakavuusasteikoilla. Myös bayesilaisten neuroverkkojen approksimaatioita tutkitaan epävarmuuden huomioivaan diabeettisen retinopatian luokitteluun kliinisellä aineistolla. Työssä luodaan yhteys epävarmuuspohjaisen luokittelujen hylkäämisen ja hylkäysvaihtoehtoluokituksen välille, jonka avulla kehitetään uusi epävarmuusmittari. Konvoluutioneuroverkkoa sovelletaan myös mandibulaarikanavan segmentaatioon kartiokeilatietokonetomografiakuvista ja sitä verrataan useamman radiologin väliseen vaihteluun tässä tehtävässä. Lopuksi tässä työssä tarkastellaan useita koneoppimismenetelmiä pikkukeskosten kuolleisuuden ja sairastuvuuden ennustamiseen, ja tutkitaan mitkä piirteet ovat tärkeitä tähän tehtävään. Tulosten perusteella voidaan arvioida, että jopa suhteellisen pientä suomalaista sairaala-aineistoa voidaan käyttää diabeettisen retinopatian ja makulaturvotuksen luokitteluun kliinisesti käytetyillä vakavuusasteikoilla. Bayesilaisten neuroverkkojen approksimaatiot pystyvät myös hyödyntämään kehitettyä epävarmuusmittaria tarkkaan epävarmuuden arviointiin diabeettisen retinopatian luokittelussa kliinisellä vakavuusasteikolla. Syvän konvoluutioneuroverkon näytetään saavuttavan parempia tuloksia kuin aikaisemmat lähestymistavat mandibulaarikanavan segmentaatiossa. Sen näytetään myös paikallistavan mandibulaarikanavan pienemmällä vaihtelulla kuin useamman radiologin välinen vaihtelu. Satunnaismetsä -luokitin saavutti parempia tuloksia kuin muut menetelmät pikkukeskosten kuolleisuuden ja sairastuvuuden ennustamisessa.

Julkaisun otsikon käännös	Koneoppiminen terveydenhuollossa
Alkuperäiskieli	Englanti
Pätevyys	Tohtorintutkinto
Myöntävä instituutio	Aalto-yliopisto
Valvoja/neuvonantaja	Solin, Arno, Vastuuprofessori Särkkä, Simo, Ohjaaja Kaski, Kimmo, Ohjaaja
Kustantaja	Aalto University
Painoksen ISBN	978-952-64-1008-1
Sähköinen ISBN	978-952-64-1009-8
Tila	Julkaistu - 2022
OKM-julkaisutyyppi	G5 Tohtorinväitöskirja (artikkeli)

Tutkimusalat

koneoppiminen
syväoppiminen
bayesilaisten neuroverkkojen approksimaatiot
terveydenhuolto

YK:n kestävän kehityksen tavoitteet

Tämä tuotos edistää seuraavia kestävän kehityksen tavoitteita:

Pääsy asiakirjaan

http://urn.fi/URN:ISBN:978-952-64-1009-8

OpenUrl-saatavuus

Koko teksti

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

Siteeraa tätä

@phdthesis{6556588d0ec24ebe8fffea65b3da13c8,

title = "Machine Learning for Healthcare",

abstract = "Machine learning has been recently proposed for various medical applications. Especially the deep neural network based approach has been found to achieve state-of-the-art performance in various classification tasks. However, many of these studies use simplified classification systems, for example, the referable/non-referable system in the case of diabetic retinopathy classification. Moreover, the studies that have used clinical classification systems have not considered the uncertainty of the classifiers, which is of paramount interest in the medical field. In addition, extensive analysis of automatic segmentation algorithms that includes comparison to the interobserver variability of multiple radiologists' segmentations has not yet been performed for some challenging tasks, such as the automatic segmentation of the mandibular canals. The machine learning algorithms should also be able to be trained on local hospital data, which can pose issues relating to the amount of available training data. This thesis considers machine learning for various tasks in healthcare using Finnish hospital data. Deep convolutional neural networks (CNNs) are utilized for diabetic retinopathy and macular edema classification based on clinical severity scales. In addition, approximate Bayesian deep learning approaches are systematically studied for uncertainty-aware diabetic retinopathy classification of clinical data. A connection is derived between the referral of uncertain classifications and reject option classification, and it is used to develop a novel uncertainty measure. A CNN approach will also be introduced for the segmentation of the mandibular canal in cone beam computed tomography volumes. The approach is then compared to the interobserver variability of multiple radiologists' segmentations of the canal. Lastly, this thesis will examine multiple machine learning approaches for very low birth weight neonate mortality and morbidity prediction. The results suggest that even a relatively small set of Finnish hospital data can be utilized to train deep learning classifiers for diabetic retinopathy and macular edema classification with clinical classification systems. It also turns out that approximate Bayesian neural networks and the derived novel uncertainty measure can be used to accurately estimate the uncertainty in clinical diabetic retinopathy classification. The deep learning approach is shown to set a new state-of-the-art for the mandibular canal segmentation task and it is also found to localize the canals with lower variability than the interobserver variability of four radiologists. A random forest classifier turned out to outperform other methods in neonatal mortality and morbidity prediction.",

keywords = "machine learning, deep learning, approximate Bayesian deep learning, healthcare, koneoppiminen, syv{\"a}oppiminen, bayesilaisten neuroverkkojen approksimaatiot, terveydenhuolto, machine learning, deep learning, approximate Bayesian deep learning, healthcare",

author = "Joel Jaskari",

year = "2022",

language = "English",

isbn = "978-952-64-1008-1",

series = "Aalto University publication series DOCTORAL THESES",

publisher = "Aalto University",

number = "161",

address = "Finland",

school = "Aalto University",

}

TY - THES

T1 - Machine Learning for Healthcare

AU - Jaskari, Joel

PY - 2022

Y1 - 2022

N2 - Machine learning has been recently proposed for various medical applications. Especially the deep neural network based approach has been found to achieve state-of-the-art performance in various classification tasks. However, many of these studies use simplified classification systems, for example, the referable/non-referable system in the case of diabetic retinopathy classification. Moreover, the studies that have used clinical classification systems have not considered the uncertainty of the classifiers, which is of paramount interest in the medical field. In addition, extensive analysis of automatic segmentation algorithms that includes comparison to the interobserver variability of multiple radiologists' segmentations has not yet been performed for some challenging tasks, such as the automatic segmentation of the mandibular canals. The machine learning algorithms should also be able to be trained on local hospital data, which can pose issues relating to the amount of available training data. This thesis considers machine learning for various tasks in healthcare using Finnish hospital data. Deep convolutional neural networks (CNNs) are utilized for diabetic retinopathy and macular edema classification based on clinical severity scales. In addition, approximate Bayesian deep learning approaches are systematically studied for uncertainty-aware diabetic retinopathy classification of clinical data. A connection is derived between the referral of uncertain classifications and reject option classification, and it is used to develop a novel uncertainty measure. A CNN approach will also be introduced for the segmentation of the mandibular canal in cone beam computed tomography volumes. The approach is then compared to the interobserver variability of multiple radiologists' segmentations of the canal. Lastly, this thesis will examine multiple machine learning approaches for very low birth weight neonate mortality and morbidity prediction. The results suggest that even a relatively small set of Finnish hospital data can be utilized to train deep learning classifiers for diabetic retinopathy and macular edema classification with clinical classification systems. It also turns out that approximate Bayesian neural networks and the derived novel uncertainty measure can be used to accurately estimate the uncertainty in clinical diabetic retinopathy classification. The deep learning approach is shown to set a new state-of-the-art for the mandibular canal segmentation task and it is also found to localize the canals with lower variability than the interobserver variability of four radiologists. A random forest classifier turned out to outperform other methods in neonatal mortality and morbidity prediction.

AB - Machine learning has been recently proposed for various medical applications. Especially the deep neural network based approach has been found to achieve state-of-the-art performance in various classification tasks. However, many of these studies use simplified classification systems, for example, the referable/non-referable system in the case of diabetic retinopathy classification. Moreover, the studies that have used clinical classification systems have not considered the uncertainty of the classifiers, which is of paramount interest in the medical field. In addition, extensive analysis of automatic segmentation algorithms that includes comparison to the interobserver variability of multiple radiologists' segmentations has not yet been performed for some challenging tasks, such as the automatic segmentation of the mandibular canals. The machine learning algorithms should also be able to be trained on local hospital data, which can pose issues relating to the amount of available training data. This thesis considers machine learning for various tasks in healthcare using Finnish hospital data. Deep convolutional neural networks (CNNs) are utilized for diabetic retinopathy and macular edema classification based on clinical severity scales. In addition, approximate Bayesian deep learning approaches are systematically studied for uncertainty-aware diabetic retinopathy classification of clinical data. A connection is derived between the referral of uncertain classifications and reject option classification, and it is used to develop a novel uncertainty measure. A CNN approach will also be introduced for the segmentation of the mandibular canal in cone beam computed tomography volumes. The approach is then compared to the interobserver variability of multiple radiologists' segmentations of the canal. Lastly, this thesis will examine multiple machine learning approaches for very low birth weight neonate mortality and morbidity prediction. The results suggest that even a relatively small set of Finnish hospital data can be utilized to train deep learning classifiers for diabetic retinopathy and macular edema classification with clinical classification systems. It also turns out that approximate Bayesian neural networks and the derived novel uncertainty measure can be used to accurately estimate the uncertainty in clinical diabetic retinopathy classification. The deep learning approach is shown to set a new state-of-the-art for the mandibular canal segmentation task and it is also found to localize the canals with lower variability than the interobserver variability of four radiologists. A random forest classifier turned out to outperform other methods in neonatal mortality and morbidity prediction.

KW - machine learning

KW - deep learning

KW - approximate Bayesian deep learning

KW - healthcare

KW - koneoppiminen

KW - syväoppiminen

KW - bayesilaisten neuroverkkojen approksimaatiot

KW - terveydenhuolto

KW - machine learning

KW - deep learning

KW - approximate Bayesian deep learning

KW - healthcare

M3 - Doctoral Thesis

SN - 978-952-64-1008-1

T3 - Aalto University publication series DOCTORAL THESES

PB - Aalto University

ER -

Machine Learning for Healthcare

Abstrakti

Tutkimusalat

YK:n kestävän kehityksen tavoitteet

Pääsy asiakirjaan

OpenUrl-saatavuus

Sormenjälki

Laitteet

Science-IT

Siteeraa tätä