Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans

Marc Dietrichstein; David Major; Martin Trapp; Maria Wimmer; Dimitrios Lenis; Philip Winter; Astrid Berg; Theresa Neubauer; Katja Bühler

doi:10.1007/978-3-031-18576-2_8

Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans

Marc Dietrichstein, David Major^*, Martin Trapp, Maria Wimmer, Dimitrios Lenis, Philip Winter, Astrid Berg, Theresa Neubauer, Katja Bühler

^*Corresponding author for this work

VRVis Zentrum für Virtual Reality und Visualisierung

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

29 Downloads (Pure)

Abstract

Unsupervised anomaly detection models that are trained solely by healthy data, have gained importance in recent years, as the annotation of medical data is a tedious task. Autoencoders and generative adversarial networks are the standard anomaly detection methods that are utilized to learn the data distribution. However, they fall short when it comes to inference and evaluation of the likelihood of test samples. We propose a novel combination of generative models and a probabilistic graphical model. After encoding image samples by autoencoders, the distribution of data is modeled by Random and Tensorized Sum-Product Networks ensuring exact and efficient inference at test time. We evaluate different autoencoder architectures in combination with Random and Tensorized Sum-Product Networks on mammography images using patch-wise processing and observe superior performance over utilizing the models standalone and state-of-the-art in anomaly detection for medical data.

Original language	English
Title of host publication	Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings
Editors	Anirban Mukhopadhyay, Ilkay Oksuz, Sandy Engelhardt, Dajiang Zhu, Yixuan Yuan
Publisher	Springer
Pages	77-86
Number of pages	10
ISBN (Print)	978-3-031-18575-5
DOIs	https://doi.org/10.1007/978-3-031-18576-2_8
Publication status	Published - 2022
MoE publication type	A4 Conference publication
Event	Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention - Singapore, Singapore Duration: 22 Sept 2022 → 22 Sept 2022 Conference number: 2

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer
Volume	13609 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Workshop

Workshop	Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention
Abbreviated title	DGM4MICCAI
Country/Territory	Singapore
City	Singapore
Period	22/09/2022 → 22/09/2022

Keywords

Anomaly detection
Generative models
Mammography
Sum-product networks

Access to Document

10.1007/978-3-031-18576-2_8

Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography ScansAccepted author manuscript, 409 KB

OpenUrl availability

Full text

Cite this

Dietrichstein, M., Major, D., Trapp, M., Wimmer, M., Lenis, D., Winter, P., Berg, A., Neubauer, T., & Bühler, K. (2022). Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans. In A. Mukhopadhyay, I. Oksuz, S. Engelhardt, D. Zhu, & Y. Yuan (Eds.), Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings (pp. 77-86). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13609 LNCS). Springer. https://doi.org/10.1007/978-3-031-18576-2_8

Dietrichstein, Marc ; Major, David ; Trapp, Martin et al. / Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans. Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings. editor / Anirban Mukhopadhyay ; Ilkay Oksuz ; Sandy Engelhardt ; Dajiang Zhu ; Yixuan Yuan. Springer, 2022. pp. 77-86 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{47558ee7b8d8415bb561230812a9fad8,

title = "Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans",

abstract = "Unsupervised anomaly detection models that are trained solely by healthy data, have gained importance in recent years, as the annotation of medical data is a tedious task. Autoencoders and generative adversarial networks are the standard anomaly detection methods that are utilized to learn the data distribution. However, they fall short when it comes to inference and evaluation of the likelihood of test samples. We propose a novel combination of generative models and a probabilistic graphical model. After encoding image samples by autoencoders, the distribution of data is modeled by Random and Tensorized Sum-Product Networks ensuring exact and efficient inference at test time. We evaluate different autoencoder architectures in combination with Random and Tensorized Sum-Product Networks on mammography images using patch-wise processing and observe superior performance over utilizing the models standalone and state-of-the-art in anomaly detection for medical data.",

keywords = "Anomaly detection, Generative models, Mammography, Sum-product networks",

author = "Marc Dietrichstein and David Major and Martin Trapp and Maria Wimmer and Dimitrios Lenis and Philip Winter and Astrid Berg and Theresa Neubauer and Katja B{\"u}hler",

note = "Funding Information: VRVis is funded by BMK, BMDW, Styria, SFG, Tyrol and Vienna Business Agency in the scope of COMET-Competence Centers for Excellent Technologies (879730) which is managed by FFG. Thanks go to AGFA HealthCare, project partner of VRVis, for providing valuable input. Martin Trapp acknowledges funding from the Academy of Finland (347279). ; Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention, DGM4MICCAI ; Conference date: 22-09-2022 Through 22-09-2022",

year = "2022",

doi = "10.1007/978-3-031-18576-2_8",

language = "English",

isbn = "978-3-031-18575-5",

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

publisher = "Springer",

pages = "77--86",

editor = "Anirban Mukhopadhyay and Ilkay Oksuz and Sandy Engelhardt and Dajiang Zhu and Yixuan Yuan",

booktitle = "Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings",

address = "Germany",

}

Dietrichstein, M, Major, D, Trapp, M, Wimmer, M, Lenis, D, Winter, P, Berg, A, Neubauer, T & Bühler, K 2022, Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans. in A Mukhopadhyay, I Oksuz, S Engelhardt, D Zhu & Y Yuan (eds), Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13609 LNCS, Springer, pp. 77-86, Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention, Singapore, Singapore, 22/09/2022. https://doi.org/10.1007/978-3-031-18576-2_8

Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans. / Dietrichstein, Marc; Major, David; Trapp, Martin et al.
Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings. ed. / Anirban Mukhopadhyay; Ilkay Oksuz; Sandy Engelhardt; Dajiang Zhu; Yixuan Yuan. Springer, 2022. p. 77-86 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13609 LNCS).

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

TY - GEN

T1 - Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans

AU - Dietrichstein, Marc

AU - Major, David

AU - Trapp, Martin

AU - Wimmer, Maria

AU - Lenis, Dimitrios

AU - Winter, Philip

AU - Berg, Astrid

AU - Neubauer, Theresa

AU - Bühler, Katja

N1 - Conference code: 2

PY - 2022

Y1 - 2022

N2 - Unsupervised anomaly detection models that are trained solely by healthy data, have gained importance in recent years, as the annotation of medical data is a tedious task. Autoencoders and generative adversarial networks are the standard anomaly detection methods that are utilized to learn the data distribution. However, they fall short when it comes to inference and evaluation of the likelihood of test samples. We propose a novel combination of generative models and a probabilistic graphical model. After encoding image samples by autoencoders, the distribution of data is modeled by Random and Tensorized Sum-Product Networks ensuring exact and efficient inference at test time. We evaluate different autoencoder architectures in combination with Random and Tensorized Sum-Product Networks on mammography images using patch-wise processing and observe superior performance over utilizing the models standalone and state-of-the-art in anomaly detection for medical data.

AB - Unsupervised anomaly detection models that are trained solely by healthy data, have gained importance in recent years, as the annotation of medical data is a tedious task. Autoencoders and generative adversarial networks are the standard anomaly detection methods that are utilized to learn the data distribution. However, they fall short when it comes to inference and evaluation of the likelihood of test samples. We propose a novel combination of generative models and a probabilistic graphical model. After encoding image samples by autoencoders, the distribution of data is modeled by Random and Tensorized Sum-Product Networks ensuring exact and efficient inference at test time. We evaluate different autoencoder architectures in combination with Random and Tensorized Sum-Product Networks on mammography images using patch-wise processing and observe superior performance over utilizing the models standalone and state-of-the-art in anomaly detection for medical data.

KW - Anomaly detection

KW - Generative models

KW - Mammography

KW - Sum-product networks

UR - http://www.scopus.com/inward/record.url?scp=85141753362&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-18576-2_8

DO - 10.1007/978-3-031-18576-2_8

M3 - Conference contribution

AN - SCOPUS:85141753362

SN - 978-3-031-18575-5

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

SP - 77

EP - 86

BT - Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings

A2 - Mukhopadhyay, Anirban

A2 - Oksuz, Ilkay

A2 - Engelhardt, Sandy

A2 - Zhu, Dajiang

A2 - Yuan, Yixuan

PB - Springer

T2 - Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention

Y2 - 22 September 2022 through 22 September 2022

ER -

Dietrichstein M, Major D, Trapp M, Wimmer M, Lenis D, Winter P et al. Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans. In Mukhopadhyay A, Oksuz I, Engelhardt S, Zhu D, Yuan Y, editors, Deep Generative Models - 2nd MICCAI Workshop, DGM4MICCAI 2022, Held in Conjunction with MICCAI 2022, Proceedings. Springer. 2022. p. 77-86. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-18576-2_8

Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans

Abstract

Publication series

Workshop

Keywords

Access to Document

OpenUrl availability

Other files and links

Fingerprint

Trapp Martin: Exploiting Probabilistic Circuits for Stochastic Processes and Deep Learning

Cite this

Anomaly Detection Using Generative Models and Sum-Product Networks in Mammography Scans

Abstract

Publication series

Workshop

Keywords

Access to Document

OpenUrl availability

Other files and links

Fingerprint

Projects

Trapp Martin: Exploiting Probabilistic Circuits for Stochastic Processes and Deep Learning

Cite this