Geometry-Invariant Abnormality Detection

Ashay Patel; Petru Daniel Tudosiu; Walter Hugo Lopez Pinaya; Olusola Adeleke; Gary Cook; Vicky Goh; Sebastien Ourselin; M. Jorge Cardoso

doi:10.1007/978-3-031-43907-0_29

Geometry-Invariant Abnormality Detection

Ashay Patel^*
, Petru Daniel Tudosiu
, Walter Hugo Lopez Pinaya
, Olusola Adeleke
, Gary Cook
, Vicky Goh
, Sebastien Ourselin
, M. Jorge Cardoso

^*Corresponding author for this work

King's College London

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

1 Citation (Scopus)

Abstract

Cancer is a highly heterogeneous condition best visualised in positron emission tomography. Due to this heterogeneity, a general-purpose cancer detection model can be built using unsupervised learning anomaly detection models. While prior work in this field has showcased the efficacy of abnormality detection methods (e.g. Transformer-based), these have shown significant vulnerabilities to differences in data geometry. Changes in image resolution or observed field of view can result in inaccurate predictions, even with significant data pre-processing and augmentation. We propose a new spatial conditioning mechanism that enables models to adapt and learn from varying data geometries, and apply it to a state-of-the-art Vector-Quantized Variational Autoencoder + Transformer abnormality detection model. We showcase that this spatial conditioning mechanism statistically-significantly improves model performance on whole-body data compared to the same model without conditioning, while allowing the model to perform inference at varying data geometries.

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings
Editors	Hayit Greenspan, Hayit Greenspan, Anant Madabhushi, Parvin Mousavi, Septimiu Salcudean, James Duncan, Tanveer Syeda-Mahmood, Russell Taylor
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	300-309
Number of pages	10
ISBN (Print)	9783031439063
DOIs	https://doi.org/10.1007/978-3-031-43907-0_29
Publication status	Published - 2023
Event	26th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2023 - Vancouver, Canada Duration: 8 Oct 2023 → 12 Oct 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14220 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	26th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2023
Country/Territory	Canada
City	Vancouver
Period	8/10/2023 → 12/10/2023

UN SDGs

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

Access to Document

10.1007/978-3-031-43907-0_29

Cite this

Patel, A., Tudosiu, P. D., Pinaya, W. H. L., Adeleke, O., Cook, G., Goh, V., Ourselin, S., & Cardoso, M. J. (2023). Geometry-Invariant Abnormality Detection. In H. Greenspan, H. Greenspan, A. Madabhushi, P. Mousavi, S. Salcudean, J. Duncan, T. Syeda-Mahmood, & R. Taylor (Eds.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings (pp. 300-309). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14220 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-43907-0_29

Patel, Ashay ; Tudosiu, Petru Daniel ; Pinaya, Walter Hugo Lopez et al. / Geometry-Invariant Abnormality Detection. Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings. editor / Hayit Greenspan ; Hayit Greenspan ; Anant Madabhushi ; Parvin Mousavi ; Septimiu Salcudean ; James Duncan ; Tanveer Syeda-Mahmood ; Russell Taylor. Springer Science and Business Media Deutschland GmbH, 2023. pp. 300-309 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inbook{b71ddda8dbc94c4c90ea0c2a1477864d,

title = "Geometry-Invariant Abnormality Detection",

abstract = "Cancer is a highly heterogeneous condition best visualised in positron emission tomography. Due to this heterogeneity, a general-purpose cancer detection model can be built using unsupervised learning anomaly detection models. While prior work in this field has showcased the efficacy of abnormality detection methods (e.g. Transformer-based), these have shown significant vulnerabilities to differences in data geometry. Changes in image resolution or observed field of view can result in inaccurate predictions, even with significant data pre-processing and augmentation. We propose a new spatial conditioning mechanism that enables models to adapt and learn from varying data geometries, and apply it to a state-of-the-art Vector-Quantized Variational Autoencoder + Transformer abnormality detection model. We showcase that this spatial conditioning mechanism statistically-significantly improves model performance on whole-body data compared to the same model without conditioning, while allowing the model to perform inference at varying data geometries.",

author = "Ashay Patel and Tudosiu, \{Petru Daniel\} and Pinaya, \{Walter Hugo Lopez\} and Olusola Adeleke and Gary Cook and Vicky Goh and Sebastien Ourselin and Cardoso, \{M. Jorge\}",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 26th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2023 ; Conference date: 08-10-2023 Through 12-10-2023",

year = "2023",

doi = "10.1007/978-3-031-43907-0\_29",

language = "English",

isbn = "9783031439063",

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 = "300--309",

editor = "Hayit Greenspan and Hayit Greenspan and Anant Madabhushi and Parvin Mousavi and Septimiu Salcudean and James Duncan and Tanveer Syeda-Mahmood and Russell Taylor",

booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings",

address = "Germany",

}

Patel, A, Tudosiu, PD, Pinaya, WHL, Adeleke, O, Cook, G , Goh, V , Ourselin, S & Cardoso, MJ 2023, Geometry-Invariant Abnormality Detection. in H Greenspan, H Greenspan, A Madabhushi, P Mousavi, S Salcudean, J Duncan, T Syeda-Mahmood & R Taylor (eds), Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14220 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 300-309, 26th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2023, Vancouver, Canada, 8/10/2023. https://doi.org/10.1007/978-3-031-43907-0_29

Geometry-Invariant Abnormality Detection. / Patel, Ashay; Tudosiu, Petru Daniel; Pinaya, Walter Hugo Lopez et al.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings. ed. / Hayit Greenspan; Hayit Greenspan; Anant Madabhushi; Parvin Mousavi; Septimiu Salcudean; James Duncan; Tanveer Syeda-Mahmood; Russell Taylor. Springer Science and Business Media Deutschland GmbH, 2023. p. 300-309 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14220 LNCS).

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

TY - CHAP

T1 - Geometry-Invariant Abnormality Detection

AU - Patel, Ashay

AU - Tudosiu, Petru Daniel

AU - Pinaya, Walter Hugo Lopez

AU - Adeleke, Olusola

AU - Cook, Gary

AU - Goh, Vicky

AU - Ourselin, Sebastien

AU - Cardoso, M. Jorge

PY - 2023

Y1 - 2023

N2 - Cancer is a highly heterogeneous condition best visualised in positron emission tomography. Due to this heterogeneity, a general-purpose cancer detection model can be built using unsupervised learning anomaly detection models. While prior work in this field has showcased the efficacy of abnormality detection methods (e.g. Transformer-based), these have shown significant vulnerabilities to differences in data geometry. Changes in image resolution or observed field of view can result in inaccurate predictions, even with significant data pre-processing and augmentation. We propose a new spatial conditioning mechanism that enables models to adapt and learn from varying data geometries, and apply it to a state-of-the-art Vector-Quantized Variational Autoencoder + Transformer abnormality detection model. We showcase that this spatial conditioning mechanism statistically-significantly improves model performance on whole-body data compared to the same model without conditioning, while allowing the model to perform inference at varying data geometries.

AB - Cancer is a highly heterogeneous condition best visualised in positron emission tomography. Due to this heterogeneity, a general-purpose cancer detection model can be built using unsupervised learning anomaly detection models. While prior work in this field has showcased the efficacy of abnormality detection methods (e.g. Transformer-based), these have shown significant vulnerabilities to differences in data geometry. Changes in image resolution or observed field of view can result in inaccurate predictions, even with significant data pre-processing and augmentation. We propose a new spatial conditioning mechanism that enables models to adapt and learn from varying data geometries, and apply it to a state-of-the-art Vector-Quantized Variational Autoencoder + Transformer abnormality detection model. We showcase that this spatial conditioning mechanism statistically-significantly improves model performance on whole-body data compared to the same model without conditioning, while allowing the model to perform inference at varying data geometries.

UR - https://www.scopus.com/pages/publications/85174624152

U2 - 10.1007/978-3-031-43907-0_29

DO - 10.1007/978-3-031-43907-0_29

M3 - Conference paper

AN - SCOPUS:85174624152

SN - 9783031439063

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

SP - 300

EP - 309

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings

A2 - Greenspan, Hayit

A2 - Madabhushi, Anant

A2 - Mousavi, Parvin

A2 - Salcudean, Septimiu

A2 - Duncan, James

A2 - Syeda-Mahmood, Tanveer

A2 - Taylor, Russell

PB - Springer Science and Business Media Deutschland GmbH

T2 - 26th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2023

Y2 - 8 October 2023 through 12 October 2023

ER -

Patel A, Tudosiu PD, Pinaya WHL, Adeleke O, Cook G , Goh V et al. Geometry-Invariant Abnormality Detection. In Greenspan H, Greenspan H, Madabhushi A, Mousavi P, Salcudean S, Duncan J, Syeda-Mahmood T, Taylor R, editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 - 26th International Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 300-309. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-43907-0_29

Geometry-Invariant Abnormality Detection

Abstract

Publication series

Conference

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this