Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification

Xindi Wu; Yijun Mao; Haohan Wang; Xiangrui Zeng; Xin Gao; Eric P. Xing; Min Xu

doi:10.1109/BIBM47256.2019.8982954

Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification

Xindi Wu, Yijun Mao, Haohan Wang, Xiangrui Zeng, Xin Gao, Eric P. Xing, Min Xu

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

Abstract

Cellular Electron Cryo Tomography (CECT) 3D imaging has permitted biomedical community to study macromolecule structures inside single cells with deep learning approaches. Many deep learning-based methods have since been developed to classify macromolecule structures from tomograms with high accuracy. However, several recent studies have demonstrated the lack of robustness in these models against often-imperceptible, designed changes of input. Therefore, making existing subtomogram-classification models robust remains a serious challenge. In this paper, we study the robustness of the state-of-the-art subtomogram classifier on CECT images and propose a method called Regularized Adversarial Training (RAT) to defend the classifier against a wide range of designed threats. Our results show that RAT improves robustness for CECT image classification over the previous methods.

Original language	English (US)
Title of host publication	Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019
Editors	Illhoi Yoo, Jinbo Bi, Xiaohua Tony Hu
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	52-67
Number of pages	16
ISBN (Electronic)	9781728118673
DOIs	https://doi.org/10.1109/BIBM47256.2019.8982954
State	Published - Nov 2019
Externally published	Yes
Event	2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019 - San Diego, United States Duration: Nov 18 2019 → Nov 21 2019

Conference

Conference	2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019
Country/Territory	United States
City	San Diego
Period	11/18/19 → 11/21/19

Keywords

Adversarial Attacks
Adversarial Training
Cellular Electron Cryo Tomography
Classification
Robustness

ASJC Scopus subject areas

Biochemistry
Biotechnology
Molecular Medicine
Modeling and Simulation
Health Informatics
Pharmacology (medical)
Public Health, Environmental and Occupational Health

Online availability

10.1109/BIBM47256.2019.8982954

Library availability

Discover UIUC Full Text

Cite this

Wu, X., Mao, Y., Wang, H., Zeng, X., Gao, X., Xing, E. P., & Xu, M. (2019). Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification. In I. Yoo, J. Bi, & X. T. Hu (Eds.), Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019 (pp. 52-67). Article 8982954 Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/BIBM47256.2019.8982954

Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification. / Wu, Xindi; Mao, Yijun; Wang, Haohan et al.
Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019. ed. / Illhoi Yoo; Jinbo Bi; Xiaohua Tony Hu. Institute of Electrical and Electronics Engineers Inc., 2019. p. 52-67 8982954.

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

Wu, X, Mao, Y, Wang, H, Zeng, X, Gao, X, Xing, EP & Xu, M 2019, Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification. in I Yoo, J Bi & XT Hu (eds), Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019., 8982954, Institute of Electrical and Electronics Engineers Inc., pp. 52-67, 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019, San Diego, United States, 11/18/19. https://doi.org/10.1109/BIBM47256.2019.8982954

Wu X, Mao Y, Wang H, Zeng X, Gao X, Xing EP et al. Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification. In Yoo I, Bi J, Hu XT, editors, Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019. Institute of Electrical and Electronics Engineers Inc. 2019. p. 52-67. 8982954 doi: 10.1109/BIBM47256.2019.8982954

@inproceedings{244a52266f3f4d8caf0776e5f8411240,

title = "Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification",

abstract = "Cellular Electron Cryo Tomography (CECT) 3D imaging has permitted biomedical community to study macromolecule structures inside single cells with deep learning approaches. Many deep learning-based methods have since been developed to classify macromolecule structures from tomograms with high accuracy. However, several recent studies have demonstrated the lack of robustness in these models against often-imperceptible, designed changes of input. Therefore, making existing subtomogram-classification models robust remains a serious challenge. In this paper, we study the robustness of the state-of-the-art subtomogram classifier on CECT images and propose a method called Regularized Adversarial Training (RAT) to defend the classifier against a wide range of designed threats. Our results show that RAT improves robustness for CECT image classification over the previous methods.",

keywords = "Adversarial Attacks, Adversarial Training, Cellular Electron Cryo Tomography, Classification, Robustness",

author = "Xindi Wu and Yijun Mao and Haohan Wang and Xiangrui Zeng and Xin Gao and Xing, {Eric P.} and Min Xu",

note = "Funding Information: + Correspoinding author. This work was supported in part by U.S. National Institutes of Health (NIH) grant P41 GM103712. HW was supported by the National Institutes of Health grants R01-GM093156 and P30-DA035778. XZ was supported by a fellowship from Carnegie Mellon University{\textquoteright}s Center for Machine Learning and Health. XG was supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No. BAS/1/1624, FCC/1/1976-18, FCC/1/1976-23, FCC/1/1976-25, and FCC/1/1976-26. Publisher Copyright: {\textcopyright} 2019 IEEE.; 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019 ; Conference date: 18-11-2019 Through 21-11-2019",

year = "2019",

month = nov,

doi = "10.1109/BIBM47256.2019.8982954",

language = "English (US)",

pages = "52--67",

editor = "Illhoi Yoo and Jinbo Bi and Hu, {Xiaohua Tony}",

booktitle = "Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

}

TY - GEN

T1 - Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification

AU - Wu, Xindi

AU - Mao, Yijun

AU - Wang, Haohan

AU - Zeng, Xiangrui

AU - Gao, Xin

AU - Xing, Eric P.

AU - Xu, Min

N1 - Funding Information: + Correspoinding author. This work was supported in part by U.S. National Institutes of Health (NIH) grant P41 GM103712. HW was supported by the National Institutes of Health grants R01-GM093156 and P30-DA035778. XZ was supported by a fellowship from Carnegie Mellon University’s Center for Machine Learning and Health. XG was supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No. BAS/1/1624, FCC/1/1976-18, FCC/1/1976-23, FCC/1/1976-25, and FCC/1/1976-26. Publisher Copyright: © 2019 IEEE.

PY - 2019/11

Y1 - 2019/11

N2 - Cellular Electron Cryo Tomography (CECT) 3D imaging has permitted biomedical community to study macromolecule structures inside single cells with deep learning approaches. Many deep learning-based methods have since been developed to classify macromolecule structures from tomograms with high accuracy. However, several recent studies have demonstrated the lack of robustness in these models against often-imperceptible, designed changes of input. Therefore, making existing subtomogram-classification models robust remains a serious challenge. In this paper, we study the robustness of the state-of-the-art subtomogram classifier on CECT images and propose a method called Regularized Adversarial Training (RAT) to defend the classifier against a wide range of designed threats. Our results show that RAT improves robustness for CECT image classification over the previous methods.

AB - Cellular Electron Cryo Tomography (CECT) 3D imaging has permitted biomedical community to study macromolecule structures inside single cells with deep learning approaches. Many deep learning-based methods have since been developed to classify macromolecule structures from tomograms with high accuracy. However, several recent studies have demonstrated the lack of robustness in these models against often-imperceptible, designed changes of input. Therefore, making existing subtomogram-classification models robust remains a serious challenge. In this paper, we study the robustness of the state-of-the-art subtomogram classifier on CECT images and propose a method called Regularized Adversarial Training (RAT) to defend the classifier against a wide range of designed threats. Our results show that RAT improves robustness for CECT image classification over the previous methods.

KW - Adversarial Attacks

KW - Adversarial Training

KW - Cellular Electron Cryo Tomography

KW - Classification

KW - Robustness

UR - http://www.scopus.com/inward/record.url?scp=85084337154&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85084337154&partnerID=8YFLogxK

U2 - 10.1109/BIBM47256.2019.8982954

DO - 10.1109/BIBM47256.2019.8982954

M3 - Conference contribution

AN - SCOPUS:85084337154

SP - 52

EP - 67

BT - Proceedings - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019

A2 - Yoo, Illhoi

A2 - Bi, Jinbo

A2 - Hu, Xiaohua Tony

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2019 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2019

Y2 - 18 November 2019 through 21 November 2019

ER -

Regularized Adversarial Training (RAT) for Robust Cellular Electron Cryo Tomograms Classification

Abstract

Conference

Keywords

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this