Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images

Lior Frenkel; Jacob Goldberger

doi:10.1109/isbi53787.2023.10230543

Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images

Lior Frenkel, Jacob Goldberger

Bar-Ilan University - The Alexander Kofkin Faculty of Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Calibrating regression neural networks is crucial in medical imaging applications where the decision-making depends on the predicted confidence. In this study we propose a calibration procedure for regression networks that is based on scaling the predictive variance. The amount of scaling depends on the input and is a function of the predictive variance computed by the network. We thus find an entire calibration function instead of a single parameter. We report extensive experiments on a variety of image datasets and network architectures. Our approach achieves state-of-the-art results with a guarantee that the prediction accuracy is not altered.

Original language	English
Title of host publication	2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023
Publisher	IEEE Computer Society
ISBN (Electronic)	9781665473583
DOIs	https://doi.org/10.1109/isbi53787.2023.10230543
State	Published - 2023
Event	20th IEEE International Symposium on Biomedical Imaging, ISBI 2023 - Cartagena, Colombia Duration: 18 Apr 2023 → 21 Apr 2023

Publication series

Name	Proceedings - International Symposium on Biomedical Imaging
Volume	2023-April
ISSN (Print)	1945-7928
ISSN (Electronic)	1945-8452

Conference

Conference	20th IEEE International Symposium on Biomedical Imaging, ISBI 2023
Country/Territory	Colombia
City	Cartagena
Period	18/04/23 → 21/04/23

Bibliographical note

Publisher Copyright:
© 2023 IEEE.

Funding

This research was supported by the Ministry of Science & Technology, Israel.

Funders	Funder number
Ministry of science and technology, Israel

Keywords

Expected Normalized Calibration Error (ENCE)
network calibration
neural networks
regression

Access to Document

10.1109/isbi53787.2023.10230543

Cite this

@inproceedings{cc51b07609e043b9be716a60768a310f,

title = "Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images",

abstract = "Calibrating regression neural networks is crucial in medical imaging applications where the decision-making depends on the predicted confidence. In this study we propose a calibration procedure for regression networks that is based on scaling the predictive variance. The amount of scaling depends on the input and is a function of the predictive variance computed by the network. We thus find an entire calibration function instead of a single parameter. We report extensive experiments on a variety of image datasets and network architectures. Our approach achieves state-of-the-art results with a guarantee that the prediction accuracy is not altered.",

keywords = "Expected Normalized Calibration Error (ENCE), network calibration, neural networks, regression",

author = "Lior Frenkel and Jacob Goldberger",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023 ; Conference date: 18-04-2023 Through 21-04-2023",

year = "2023",

doi = "10.1109/isbi53787.2023.10230543",

language = "אנגלית",

series = "Proceedings - International Symposium on Biomedical Imaging",

publisher = "IEEE Computer Society",

booktitle = "2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023",

address = "ארצות הברית",

}

Frenkel, L & Goldberger, J 2023, Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images. in 2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023. Proceedings - International Symposium on Biomedical Imaging, vol. 2023-April, IEEE Computer Society, 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023, Cartagena, Colombia, 18/04/23. https://doi.org/10.1109/isbi53787.2023.10230543

Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images. / Frenkel, Lior; Goldberger, Jacob.
2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023. IEEE Computer Society, 2023. (Proceedings - International Symposium on Biomedical Imaging; Vol. 2023-April).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images

AU - Frenkel, Lior

AU - Goldberger, Jacob

PY - 2023

Y1 - 2023

N2 - Calibrating regression neural networks is crucial in medical imaging applications where the decision-making depends on the predicted confidence. In this study we propose a calibration procedure for regression networks that is based on scaling the predictive variance. The amount of scaling depends on the input and is a function of the predictive variance computed by the network. We thus find an entire calibration function instead of a single parameter. We report extensive experiments on a variety of image datasets and network architectures. Our approach achieves state-of-the-art results with a guarantee that the prediction accuracy is not altered.

AB - Calibrating regression neural networks is crucial in medical imaging applications where the decision-making depends on the predicted confidence. In this study we propose a calibration procedure for regression networks that is based on scaling the predictive variance. The amount of scaling depends on the input and is a function of the predictive variance computed by the network. We thus find an entire calibration function instead of a single parameter. We report extensive experiments on a variety of image datasets and network architectures. Our approach achieves state-of-the-art results with a guarantee that the prediction accuracy is not altered.

KW - Expected Normalized Calibration Error (ENCE)

KW - network calibration

KW - neural networks

KW - regression

UR - http://www.scopus.com/inward/record.url?scp=85172140971&partnerID=8YFLogxK

U2 - 10.1109/isbi53787.2023.10230543

DO - 10.1109/isbi53787.2023.10230543

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85172140971

T3 - Proceedings - International Symposium on Biomedical Imaging

BT - 2023 IEEE International Symposium on Biomedical Imaging, ISBI 2023

PB - IEEE Computer Society

T2 - 20th IEEE International Symposium on Biomedical Imaging, ISBI 2023

Y2 - 18 April 2023 through 21 April 2023

ER -

Calibration of A Regression Network Based on the Predictive Variance with Applications to Medical Images

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this