Deep Learning from Label Proportions for Emphysema Quantification

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

Standard

Deep Learning from Label Proportions for Emphysema Quantification. / Bortsova, Gerda; Dubost, Florian; Ørting, Silas; Katramados, Ioannis; Hogeweg, Laurens; Thomsen, Laura; Wille, Mathilde; de Bruijne, Marleen.

Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: 21st International Conference, 2018, Proceedings Part II. red. / Alejandro F. Frangi; Julia A. Schnabel; Christos Davatzikos; Carlos Alberola-López; Gabor Fichtinger. Springer, 2018. s. 768-776 (Lecture notes in computer science, Bind 11071).

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

Harvard

Bortsova, G, Dubost, F, Ørting, S, Katramados, I, Hogeweg, L, Thomsen, L, Wille, M & de Bruijne, M 2018, Deep Learning from Label Proportions for Emphysema Quantification. i AF Frangi, JA Schnabel, C Davatzikos, C Alberola-López & G Fichtinger (red), Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: 21st International Conference, 2018, Proceedings Part II. Springer, Lecture notes in computer science, bind 11071, s. 768-776, 21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018, Granada, Spanien, 16/09/2018. https://doi.org/10.1007/978-3-030-00934-2_85

APA

Bortsova, G., Dubost, F., Ørting, S., Katramados, I., Hogeweg, L., Thomsen, L., Wille, M., & de Bruijne, M. (2018). Deep Learning from Label Proportions for Emphysema Quantification. I A. F. Frangi, J. A. Schnabel, C. Davatzikos, C. Alberola-López, & G. Fichtinger (red.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: 21st International Conference, 2018, Proceedings Part II (s. 768-776). Springer. Lecture notes in computer science Bind 11071 https://doi.org/10.1007/978-3-030-00934-2_85

Vancouver

Bortsova G, Dubost F, Ørting S, Katramados I, Hogeweg L, Thomsen L o.a. Deep Learning from Label Proportions for Emphysema Quantification. I Frangi AF, Schnabel JA, Davatzikos C, Alberola-López C, Fichtinger G, red., Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: 21st International Conference, 2018, Proceedings Part II. Springer. 2018. s. 768-776. (Lecture notes in computer science, Bind 11071). https://doi.org/10.1007/978-3-030-00934-2_85

Author

Bortsova, Gerda ; Dubost, Florian ; Ørting, Silas ; Katramados, Ioannis ; Hogeweg, Laurens ; Thomsen, Laura ; Wille, Mathilde ; de Bruijne, Marleen. / Deep Learning from Label Proportions for Emphysema Quantification. Medical Image Computing and Computer Assisted Intervention – MICCAI 2018: 21st International Conference, 2018, Proceedings Part II. red. / Alejandro F. Frangi ; Julia A. Schnabel ; Christos Davatzikos ; Carlos Alberola-López ; Gabor Fichtinger. Springer, 2018. s. 768-776 (Lecture notes in computer science, Bind 11071).

Bibtex

@inproceedings{9e4e047aadf046c7b4390f20251e4d5f,
title = "Deep Learning from Label Proportions for Emphysema Quantification",
abstract = "We propose an end-to-end deep learning method that learns to estimate emphysema extent from proportions of the diseased tissue. These proportions were visually estimated by experts using a standard grading system, in which grades correspond to intervals (label example: 1–5% of diseased tissue). The proposed architecture encodes the knowledge that the labels represent a volumetric proportion. A custom loss is designed to learn with intervals. Thus, during training, our network learns to segment the diseased tissue such that its proportions fit the ground truth intervals. Our architecture and loss combined improve the performance substantially (8% ICC) compared to a more conventional regression network. We outperform traditional lung densitometry and two recently published methods for emphysema quantification by a large margin (at least 7% AUC and 15% ICC), and achieve near-human-level performance. Moreover, our method generates emphysema segmentations that predict the spatial distribution of emphysema at human level.",
keywords = "Emphysema quantification, Learning from label proportions, Multiple instance learning, Weak labels",
author = "Gerda Bortsova and Florian Dubost and Silas {\O}rting and Ioannis Katramados and Laurens Hogeweg and Laura Thomsen and Mathilde Wille and {de Bruijne}, Marleen",
year = "2018",
doi = "10.1007/978-3-030-00934-2_85",
language = "English",
isbn = "9783030009335",
series = "Lecture notes in computer science",
publisher = "Springer",
pages = "768--776",
editor = "Frangi, {Alejandro F.} and Schnabel, {Julia A. } and Davatzikos, {Christos } and Alberola-L{\'o}pez, {Carlos } and Fichtinger, {Gabor }",
booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2018",
address = "Switzerland",
note = "21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018 ; Conference date: 16-09-2018 Through 20-09-2018",

}

RIS

TY - GEN

T1 - Deep Learning from Label Proportions for Emphysema Quantification

AU - Bortsova, Gerda

AU - Dubost, Florian

AU - Ørting, Silas

AU - Katramados, Ioannis

AU - Hogeweg, Laurens

AU - Thomsen, Laura

AU - Wille, Mathilde

AU - de Bruijne, Marleen

PY - 2018

Y1 - 2018

N2 - We propose an end-to-end deep learning method that learns to estimate emphysema extent from proportions of the diseased tissue. These proportions were visually estimated by experts using a standard grading system, in which grades correspond to intervals (label example: 1–5% of diseased tissue). The proposed architecture encodes the knowledge that the labels represent a volumetric proportion. A custom loss is designed to learn with intervals. Thus, during training, our network learns to segment the diseased tissue such that its proportions fit the ground truth intervals. Our architecture and loss combined improve the performance substantially (8% ICC) compared to a more conventional regression network. We outperform traditional lung densitometry and two recently published methods for emphysema quantification by a large margin (at least 7% AUC and 15% ICC), and achieve near-human-level performance. Moreover, our method generates emphysema segmentations that predict the spatial distribution of emphysema at human level.

AB - We propose an end-to-end deep learning method that learns to estimate emphysema extent from proportions of the diseased tissue. These proportions were visually estimated by experts using a standard grading system, in which grades correspond to intervals (label example: 1–5% of diseased tissue). The proposed architecture encodes the knowledge that the labels represent a volumetric proportion. A custom loss is designed to learn with intervals. Thus, during training, our network learns to segment the diseased tissue such that its proportions fit the ground truth intervals. Our architecture and loss combined improve the performance substantially (8% ICC) compared to a more conventional regression network. We outperform traditional lung densitometry and two recently published methods for emphysema quantification by a large margin (at least 7% AUC and 15% ICC), and achieve near-human-level performance. Moreover, our method generates emphysema segmentations that predict the spatial distribution of emphysema at human level.

KW - Emphysema quantification

KW - Learning from label proportions

KW - Multiple instance learning

KW - Weak labels

UR - http://www.scopus.com/inward/record.url?scp=85054065475&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-00934-2_85

DO - 10.1007/978-3-030-00934-2_85

M3 - Article in proceedings

AN - SCOPUS:85054065475

SN - 9783030009335

T3 - Lecture notes in computer science

SP - 768

EP - 776

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2018

A2 - Frangi, Alejandro F.

A2 - Schnabel, Julia A.

A2 - Davatzikos, Christos

A2 - Alberola-López, Carlos

A2 - Fichtinger, Gabor

PB - Springer

T2 - 21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018

Y2 - 16 September 2018 through 20 September 2018

ER -

ID: 203944440