Uncertainty quantification using Bayesian neural networks in classification: Application to biomedical image segmentation

Yongchan Kwon, Joong Ho Won, Beom Joon Kim, Myunghee Cho Paik

Research output: Contribution to journalArticle

Abstract

Most recent research of deep neural networks in the field of computer vision has focused on improving performances of point predictions by developing network architectures or learning algorithms. Reliable uncertainty quantification accompanied by point estimation can lead to a more informed decision, and the quality of prediction can be improved. In this paper, we invoke a Bayesian neural network and propose a natural way of quantifying uncertainties in classification problems by decomposing the moment-based predictive uncertainty into two parts: aleatoric and epistemic uncertainty. The proposed method takes into account the discrete nature of the outcome, yielding the correct interpretation of each uncertainty. We demonstrate that the proposed uncertainty quantification method provides additional insights into the point prediction using two Ischemic Stroke Lesion Segmentation Challenge datasets and the Digital Retinal Images for Vessel Extraction dataset.

Original languageEnglish
Article number106816
JournalComputational Statistics and Data Analysis
Volume142
DOIs
StatePublished - 1 Feb 2020

Fingerprint

Uncertainty Quantification
Bayesian Networks
Image segmentation
Image Segmentation
Neural Networks
Neural networks
Uncertainty
Prediction
Epistemic Uncertainty
Point Estimation
Network Architecture
Stroke
Digital Image
Classification Problems
Computer Vision
Vessel
Learning Algorithm
Segmentation
Moment
Network architecture

Keywords

  • Aleatoric and epistemic uncertainty
  • Bayesian neural network
  • Ischemic stroke lesion segmentation
  • Retinal blood vessel segmentation
  • Uncertainty quantification

Cite this

@article{851144d0242c4229a85e61508ec1f94c,
title = "Uncertainty quantification using Bayesian neural networks in classification: Application to biomedical image segmentation",
abstract = "Most recent research of deep neural networks in the field of computer vision has focused on improving performances of point predictions by developing network architectures or learning algorithms. Reliable uncertainty quantification accompanied by point estimation can lead to a more informed decision, and the quality of prediction can be improved. In this paper, we invoke a Bayesian neural network and propose a natural way of quantifying uncertainties in classification problems by decomposing the moment-based predictive uncertainty into two parts: aleatoric and epistemic uncertainty. The proposed method takes into account the discrete nature of the outcome, yielding the correct interpretation of each uncertainty. We demonstrate that the proposed uncertainty quantification method provides additional insights into the point prediction using two Ischemic Stroke Lesion Segmentation Challenge datasets and the Digital Retinal Images for Vessel Extraction dataset.",
keywords = "Aleatoric and epistemic uncertainty, Bayesian neural network, Ischemic stroke lesion segmentation, Retinal blood vessel segmentation, Uncertainty quantification",
author = "Yongchan Kwon and Won, {Joong Ho} and Kim, {Beom Joon} and Paik, {Myunghee Cho}",
year = "2020",
month = "2",
day = "1",
doi = "10.1016/j.csda.2019.106816",
language = "English",
volume = "142",
journal = "Computational Statistics and Data Analysis",
issn = "0167-9473",
publisher = "Elsevier",

}

Uncertainty quantification using Bayesian neural networks in classification : Application to biomedical image segmentation. / Kwon, Yongchan; Won, Joong Ho; Kim, Beom Joon; Paik, Myunghee Cho.

In: Computational Statistics and Data Analysis, Vol. 142, 106816, 01.02.2020.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Uncertainty quantification using Bayesian neural networks in classification

T2 - Application to biomedical image segmentation

AU - Kwon, Yongchan

AU - Won, Joong Ho

AU - Kim, Beom Joon

AU - Paik, Myunghee Cho

PY - 2020/2/1

Y1 - 2020/2/1

N2 - Most recent research of deep neural networks in the field of computer vision has focused on improving performances of point predictions by developing network architectures or learning algorithms. Reliable uncertainty quantification accompanied by point estimation can lead to a more informed decision, and the quality of prediction can be improved. In this paper, we invoke a Bayesian neural network and propose a natural way of quantifying uncertainties in classification problems by decomposing the moment-based predictive uncertainty into two parts: aleatoric and epistemic uncertainty. The proposed method takes into account the discrete nature of the outcome, yielding the correct interpretation of each uncertainty. We demonstrate that the proposed uncertainty quantification method provides additional insights into the point prediction using two Ischemic Stroke Lesion Segmentation Challenge datasets and the Digital Retinal Images for Vessel Extraction dataset.

AB - Most recent research of deep neural networks in the field of computer vision has focused on improving performances of point predictions by developing network architectures or learning algorithms. Reliable uncertainty quantification accompanied by point estimation can lead to a more informed decision, and the quality of prediction can be improved. In this paper, we invoke a Bayesian neural network and propose a natural way of quantifying uncertainties in classification problems by decomposing the moment-based predictive uncertainty into two parts: aleatoric and epistemic uncertainty. The proposed method takes into account the discrete nature of the outcome, yielding the correct interpretation of each uncertainty. We demonstrate that the proposed uncertainty quantification method provides additional insights into the point prediction using two Ischemic Stroke Lesion Segmentation Challenge datasets and the Digital Retinal Images for Vessel Extraction dataset.

KW - Aleatoric and epistemic uncertainty

KW - Bayesian neural network

KW - Ischemic stroke lesion segmentation

KW - Retinal blood vessel segmentation

KW - Uncertainty quantification

UR - http://www.scopus.com/inward/record.url?scp=85069976043&partnerID=8YFLogxK

U2 - 10.1016/j.csda.2019.106816

DO - 10.1016/j.csda.2019.106816

M3 - Article

AN - SCOPUS:85069976043

VL - 142

JO - Computational Statistics and Data Analysis

JF - Computational Statistics and Data Analysis

SN - 0167-9473

M1 - 106816

ER -