Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification

Muhammad Waqas; Tomas Maul; Amr Ahmed; Iman Yi Liao

doi:10.1007/978-981-97-1417-9_9

Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification

Muhammad Waqas^*
, Tomas Maul
, Amr Ahmed
, Iman Yi Liao

^*Corresponding author for this work

University of Nottingham Malaysia
University of Malakand
School of Computer Science
University of Nottingham

Research output: Chapter in Book/Report/Conference proceeding › Conference proceeding (ISBN) › peer-review

1 Citation (Scopus)

Abstract

Methods for post-hoc interpretability are essential for understanding neural network results. Recent years have seen the emergence of numerous post-hoc techniques, but their application to certain tasks, such as histopathological image classification for breast cancer, can produce varied and unpredictable outcomes. Frameworks for quantitative assessment are essential for evaluating each method’s effectiveness. The implementation of post-hoc interpretability methodologies is however hampered by the shortcomings of current frameworks, particularly in high-risk industries. In this study, the performance levels of several common post-hoc interpretability methods are systematically evaluated and compared in the context of histopathological image classification for breast cancer. The study is based on six post-hoc interpretability methods, 3 datasets, and 3 deep neural network models, compared via a RemOve And Retrain (ROAR) approach. The results show that Shapley value sampling obtains the best overall performance in the context of the chosen breast cancer histopathological image datasets.

Original language	English
Title of host publication	Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings
Editors	Jinchang Ren, Amir Hussain, Iman Yi Liao, Rongjun Chen, Kaizhu Huang, Huimin Zhao, Xiaoyong Liu, Ping Ma, Thomas Maul
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	95-104
Number of pages	10
ISBN (Print)	9789819714162
DOIs	https://doi.org/10.1007/978-981-97-1417-9_9
Publication status	Published - 22 May 2024
Event	13th International Conference on Brain Inspired Cognitive Systems, BICS 2023 - Kuala Lumpur, Malaysia Duration: 5 Aug 2023 → 6 Aug 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14374 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	13th International Conference on Brain Inspired Cognitive Systems, BICS 2023
Country/Territory	Malaysia
City	Kuala Lumpur
Period	5/08/23 → 6/08/23

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

Breast cancer
Histopathological images
Post-hoc interpretability
ROAR

Access to Document

10.1007/978-981-97-1417-9_9

Cite this

Waqas, M., Maul, T., Ahmed, A., & Liao, I. Y. (2024). Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification. In J. Ren, A. Hussain, I. Y. Liao, R. Chen, K. Huang, H. Zhao, X. Liu, P. Ma, & T. Maul (Eds.), Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings (pp. 95-104). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14374 LNAI). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-97-1417-9_9

Waqas, Muhammad ; Maul, Tomas ; Ahmed, Amr et al. / Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification. Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings. editor / Jinchang Ren ; Amir Hussain ; Iman Yi Liao ; Rongjun Chen ; Kaizhu Huang ; Huimin Zhao ; Xiaoyong Liu ; Ping Ma ; Thomas Maul. Springer Science and Business Media Deutschland GmbH, 2024. pp. 95-104 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{cbc02c5649e34d34829e0ce317dba4ac,

title = "Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification",

abstract = "Methods for post-hoc interpretability are essential for understanding neural network results. Recent years have seen the emergence of numerous post-hoc techniques, but their application to certain tasks, such as histopathological image classification for breast cancer, can produce varied and unpredictable outcomes. Frameworks for quantitative assessment are essential for evaluating each method{\textquoteright}s effectiveness. The implementation of post-hoc interpretability methodologies is however hampered by the shortcomings of current frameworks, particularly in high-risk industries. In this study, the performance levels of several common post-hoc interpretability methods are systematically evaluated and compared in the context of histopathological image classification for breast cancer. The study is based on six post-hoc interpretability methods, 3 datasets, and 3 deep neural network models, compared via a RemOve And Retrain (ROAR) approach. The results show that Shapley value sampling obtains the best overall performance in the context of the chosen breast cancer histopathological image datasets.",

keywords = "Breast cancer, Histopathological images, Post-hoc interpretability, ROAR",

author = "Muhammad Waqas and Tomas Maul and Amr Ahmed and Liao, \{Iman Yi\}",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.; 13th International Conference on Brain Inspired Cognitive Systems, BICS 2023 ; Conference date: 05-08-2023 Through 06-08-2023",

year = "2024",

month = may,

day = "22",

doi = "10.1007/978-981-97-1417-9\_9",

language = "English",

isbn = "9789819714162",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "95--104",

editor = "Jinchang Ren and Amir Hussain and Liao, \{Iman Yi\} and Rongjun Chen and Kaizhu Huang and Huimin Zhao and Xiaoyong Liu and Ping Ma and Thomas Maul",

booktitle = "Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings",

address = "Germany",

}

Waqas, M, Maul, T, Ahmed, A & Liao, IY 2024, Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification. in J Ren, A Hussain, IY Liao, R Chen, K Huang, H Zhao, X Liu, P Ma & T Maul (eds), Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14374 LNAI, Springer Science and Business Media Deutschland GmbH, pp. 95-104, 13th International Conference on Brain Inspired Cognitive Systems, BICS 2023, Kuala Lumpur, Malaysia, 5/08/23. https://doi.org/10.1007/978-981-97-1417-9_9

Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification. / Waqas, Muhammad; Maul, Tomas; Ahmed, Amr et al.
Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings. ed. / Jinchang Ren; Amir Hussain; Iman Yi Liao; Rongjun Chen; Kaizhu Huang; Huimin Zhao; Xiaoyong Liu; Ping Ma; Thomas Maul. Springer Science and Business Media Deutschland GmbH, 2024. p. 95-104 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14374 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference proceeding (ISBN) › peer-review

TY - GEN

T1 - Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification

AU - Waqas, Muhammad

AU - Maul, Tomas

AU - Ahmed, Amr

AU - Liao, Iman Yi

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

PY - 2024/5/22

Y1 - 2024/5/22

N2 - Methods for post-hoc interpretability are essential for understanding neural network results. Recent years have seen the emergence of numerous post-hoc techniques, but their application to certain tasks, such as histopathological image classification for breast cancer, can produce varied and unpredictable outcomes. Frameworks for quantitative assessment are essential for evaluating each method’s effectiveness. The implementation of post-hoc interpretability methodologies is however hampered by the shortcomings of current frameworks, particularly in high-risk industries. In this study, the performance levels of several common post-hoc interpretability methods are systematically evaluated and compared in the context of histopathological image classification for breast cancer. The study is based on six post-hoc interpretability methods, 3 datasets, and 3 deep neural network models, compared via a RemOve And Retrain (ROAR) approach. The results show that Shapley value sampling obtains the best overall performance in the context of the chosen breast cancer histopathological image datasets.

AB - Methods for post-hoc interpretability are essential for understanding neural network results. Recent years have seen the emergence of numerous post-hoc techniques, but their application to certain tasks, such as histopathological image classification for breast cancer, can produce varied and unpredictable outcomes. Frameworks for quantitative assessment are essential for evaluating each method’s effectiveness. The implementation of post-hoc interpretability methodologies is however hampered by the shortcomings of current frameworks, particularly in high-risk industries. In this study, the performance levels of several common post-hoc interpretability methods are systematically evaluated and compared in the context of histopathological image classification for breast cancer. The study is based on six post-hoc interpretability methods, 3 datasets, and 3 deep neural network models, compared via a RemOve And Retrain (ROAR) approach. The results show that Shapley value sampling obtains the best overall performance in the context of the chosen breast cancer histopathological image datasets.

KW - Breast cancer

KW - Histopathological images

KW - Post-hoc interpretability

KW - ROAR

UR - https://www.scopus.com/pages/publications/85195140316

UR - https://www.scopus.com/pages/publications/85195140316#tab=citedBy

U2 - 10.1007/978-981-97-1417-9_9

DO - 10.1007/978-981-97-1417-9_9

M3 - Conference proceeding (ISBN)

AN - SCOPUS:85195140316

SN - 9789819714162

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 95

EP - 104

BT - Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings

A2 - Ren, Jinchang

A2 - Hussain, Amir

A2 - Liao, Iman Yi

A2 - Chen, Rongjun

A2 - Huang, Kaizhu

A2 - Zhao, Huimin

A2 - Liu, Xiaoyong

A2 - Ma, Ping

A2 - Maul, Thomas

PB - Springer Science and Business Media Deutschland GmbH

T2 - 13th International Conference on Brain Inspired Cognitive Systems, BICS 2023

Y2 - 5 August 2023 through 6 August 2023

ER -

Waqas M, Maul T, Ahmed A, Liao IY. Evaluation of Post-hoc Interpretability Methods in Breast Cancer Histopathological Image Classification. In Ren J, Hussain A, Liao IY, Chen R, Huang K, Zhao H, Liu X, Ma P, Maul T, editors, Advances in Brain Inspired Cognitive Systems - 13th International Conference, BICS 2023, Proceedings. Springer Science and Business Media Deutschland GmbH. 2024. p. 95-104. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-981-97-1417-9_9