Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection

Camille Ruppli; Pietro Gori; Roberto Ardon; Isabelle Bloch

doi:10.1007/978-3-031-44917-8_9

Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection

Camille Ruppli
, Pietro Gori
, Roberto Ardon
, Isabelle Bloch

Institut Polytechnique de Paris
Incepto Medical
Sorbonne Université

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

1 Citation (Scopus)

Abstract

Early diagnosis of prostate cancer is crucial for efficient treatment. Multi-parametric Magnetic Resonance Images (mp-MRI) are widely used for lesion detection. The Prostate Imaging Reporting and Data System (PI-RADS) has standardized interpretation of prostate MRI by defining a score for lesion malignancy. PI-RADS data is readily available from radiology reports but is subject to high inter-reports variability. We propose a new contrastive loss function that leverages weak metadata with multiple annotators per sample and takes advantage of inter-reports variability by defining metadata confidence. By combining metadata of varying confidence with unannotated data into a single conditional contrastive loss function, we report a 3% AUC increase on lesion detection on the public PI-CAI challenge dataset. Code is available at: https://github.com/camilleruppli/decoupled_ccl.

Original language	English
Title of host publication	Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings
Editors	Zhiyun Xue, Sameer Antani, Ghada Zamzmi, Feng Yang, Sivaramakrishnan Rajaraman, Zhaohui Liang, Sharon Xiaolei Huang, Marius George Linguraru
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	95-105
Number of pages	11
ISBN (Print)	9783031471964
DOIs	https://doi.org/10.1007/978-3-031-44917-8_9
Publication status	Published - 1 Jan 2023
Event	2nd Workshop on Medical Image Learning with Noisy and Limited Data, MILLanD 2023 - Vancouver, Canada Duration: 8 Oct 2023 → 8 Oct 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14307 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	2nd Workshop on Medical Image Learning with Noisy and Limited Data, MILLanD 2023
Country/Territory	Canada
City	Vancouver
Period	8/10/23 → 8/10/23

UN SDGs

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

SDG 3 Good Health and Well-being

Keywords

Contrastive Learning
Prostate cancer segmentation
Semi-supervised Learning

Access to Document

10.1007/978-3-031-44917-8_9

Cite this

Ruppli, C., Gori, P., Ardon, R., & Bloch, I. (2023). Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection. In Z. Xue, S. Antani, G. Zamzmi, F. Yang, S. Rajaraman, Z. Liang, S. X. Huang, & M. G. Linguraru (Eds.), Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings (pp. 95-105). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14307 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-44917-8_9

Ruppli, Camille ; Gori, Pietro ; Ardon, Roberto et al. / Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection. Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings. editor / Zhiyun Xue ; Sameer Antani ; Ghada Zamzmi ; Feng Yang ; Sivaramakrishnan Rajaraman ; Zhaohui Liang ; Sharon Xiaolei Huang ; Marius George Linguraru. Springer Science and Business Media Deutschland GmbH, 2023. pp. 95-105 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4c399af8abad4355990b616e02370de7,

title = "Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection",

abstract = "Early diagnosis of prostate cancer is crucial for efficient treatment. Multi-parametric Magnetic Resonance Images (mp-MRI) are widely used for lesion detection. The Prostate Imaging Reporting and Data System (PI-RADS) has standardized interpretation of prostate MRI by defining a score for lesion malignancy. PI-RADS data is readily available from radiology reports but is subject to high inter-reports variability. We propose a new contrastive loss function that leverages weak metadata with multiple annotators per sample and takes advantage of inter-reports variability by defining metadata confidence. By combining metadata of varying confidence with unannotated data into a single conditional contrastive loss function, we report a 3\% AUC increase on lesion detection on the public PI-CAI challenge dataset. Code is available at: https://github.com/camilleruppli/decoupled\_ccl.",

keywords = "Contrastive Learning, Prostate cancer segmentation, Semi-supervised Learning",

author = "Camille Ruppli and Pietro Gori and Roberto Ardon and Isabelle Bloch",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 2nd Workshop on Medical Image Learning with Noisy and Limited Data, MILLanD 2023 ; Conference date: 08-10-2023 Through 08-10-2023",

year = "2023",

month = jan,

day = "1",

doi = "10.1007/978-3-031-44917-8\_9",

language = "English",

isbn = "9783031471964",

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--105",

editor = "Zhiyun Xue and Sameer Antani and Ghada Zamzmi and Feng Yang and Sivaramakrishnan Rajaraman and Zhaohui Liang and Huang, \{Sharon Xiaolei\} and Linguraru, \{Marius George\}",

booktitle = "Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings",

}

Ruppli, C, Gori, P, Ardon, R & Bloch, I 2023, Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection. in Z Xue, S Antani, G Zamzmi, F Yang, S Rajaraman, Z Liang, SX Huang & MG Linguraru (eds), Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14307 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 95-105, 2nd Workshop on Medical Image Learning with Noisy and Limited Data, MILLanD 2023, Vancouver, Canada, 8/10/23. https://doi.org/10.1007/978-3-031-44917-8_9

Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection. / Ruppli, Camille; Gori, Pietro; Ardon, Roberto et al.
Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings. ed. / Zhiyun Xue; Sameer Antani; Ghada Zamzmi; Feng Yang; Sivaramakrishnan Rajaraman; Zhaohui Liang; Sharon Xiaolei Huang; Marius George Linguraru. Springer Science and Business Media Deutschland GmbH, 2023. p. 95-105 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14307 LNCS).

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

TY - GEN

T1 - Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection

AU - Ruppli, Camille

AU - Gori, Pietro

AU - Ardon, Roberto

AU - Bloch, Isabelle

PY - 2023/1/1

Y1 - 2023/1/1

N2 - Early diagnosis of prostate cancer is crucial for efficient treatment. Multi-parametric Magnetic Resonance Images (mp-MRI) are widely used for lesion detection. The Prostate Imaging Reporting and Data System (PI-RADS) has standardized interpretation of prostate MRI by defining a score for lesion malignancy. PI-RADS data is readily available from radiology reports but is subject to high inter-reports variability. We propose a new contrastive loss function that leverages weak metadata with multiple annotators per sample and takes advantage of inter-reports variability by defining metadata confidence. By combining metadata of varying confidence with unannotated data into a single conditional contrastive loss function, we report a 3% AUC increase on lesion detection on the public PI-CAI challenge dataset. Code is available at: https://github.com/camilleruppli/decoupled_ccl.

AB - Early diagnosis of prostate cancer is crucial for efficient treatment. Multi-parametric Magnetic Resonance Images (mp-MRI) are widely used for lesion detection. The Prostate Imaging Reporting and Data System (PI-RADS) has standardized interpretation of prostate MRI by defining a score for lesion malignancy. PI-RADS data is readily available from radiology reports but is subject to high inter-reports variability. We propose a new contrastive loss function that leverages weak metadata with multiple annotators per sample and takes advantage of inter-reports variability by defining metadata confidence. By combining metadata of varying confidence with unannotated data into a single conditional contrastive loss function, we report a 3% AUC increase on lesion detection on the public PI-CAI challenge dataset. Code is available at: https://github.com/camilleruppli/decoupled_ccl.

KW - Contrastive Learning

KW - Prostate cancer segmentation

KW - Semi-supervised Learning

U2 - 10.1007/978-3-031-44917-8_9

DO - 10.1007/978-3-031-44917-8_9

M3 - Conference contribution

AN - SCOPUS:85174724831

SN - 9783031471964

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

SP - 95

EP - 105

BT - Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings

A2 - Xue, Zhiyun

A2 - Antani, Sameer

A2 - Zamzmi, Ghada

A2 - Yang, Feng

A2 - Rajaraman, Sivaramakrishnan

A2 - Liang, Zhaohui

A2 - Huang, Sharon Xiaolei

A2 - Linguraru, Marius George

PB - Springer Science and Business Media Deutschland GmbH

T2 - 2nd Workshop on Medical Image Learning with Noisy and Limited Data, MILLanD 2023

Y2 - 8 October 2023 through 8 October 2023

ER -

Ruppli C, Gori P, Ardon R, Bloch I. Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection. In Xue Z, Antani S, Zamzmi G, Yang F, Rajaraman S, Liang Z, Huang SX, Linguraru MG, editors, Medical Image Learning with Limited and Noisy Data - 2nd International Workshop, MILLanD 2023, Held in Conjunction with MICCAI 2023, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 95-105. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-44917-8_9

Decoupled Conditional Contrastive Learning with Variable Metadata for Prostate Lesion Detection

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Fingerprint

Cite this