Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks

Nima Tajbakhsh, Michael B. Gotway, Jianming Liang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

23 Citations (Scopus)

Abstract

Computer-aided detection (CAD) can play a major role in diagnosing pulmonary embolism (PE) at CT pulmonary angiography (CTPA). However, despite their demonstrated utility, to achieve a clinically acceptable sensitivity, existing PE CAD systems generate a high number of false positives, imposing extra burdens on radiologists to adjudicate these superfluous CAD findings. In this study, we investigate the feasibility of convolutional neural networks (CNNs) as an effective mechanism for eliminating false positives. A critical issue in successfully utilizing CNNs for detecting an object in 3D images is to develop a “right” image representation for the object. Toward this end, we have developed a vesselaligned multi-planar image representation of emboli. Our image representation offers three advantages: (1) efficiency and compactness—concisely summarizing the 3D contextual information around an embolus in only 2 image channels, (2) consistency—automatically aligning the embolus in the 2-channel images according to the orientation of the affected vessel, and (3) expandability—naturally supporting data augmentation for training CNNs. We have evaluated our CAD approach using 121 CTPA datasets with a total of 326 emboli, achieving a sensitivity of 83% at 2 false positives per volume. This performance is superior to the best performing CAD system in the literature, which achieves a sensitivity of 71% at the same level of false positives. We have further evaluated our system using the entire 20 CTPA test datasets from the PE challenge. Our system outperforms the winning system from the challenge at 0mm localization error but is outperformed by it at 2mm and 5mm localization errors. In our view, the performance at 0mm localization error is more important than those at 2mm and 5mm localization errors.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages62-69
Number of pages8
Volume9350
ISBN (Print)9783319245706, 9783319245706, 9783319245706
DOIs
StatePublished - 2015
Event18th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2015 - Munich, Germany
Duration: Oct 5 2015Oct 9 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9350
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other18th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2015
CountryGermany
CityMunich
Period10/5/1510/9/15

Fingerprint

Computer-aided Detection
Image Representation
Vessel
False Positive
Angiography
Neural Networks
Neural networks
Data Augmentation
3D Image
Entire

Keywords

  • Computer-aided detection
  • Convolutional neural networks
  • Pulmonary embolism
  • Vessel-aligned image representation

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Tajbakhsh, N., Gotway, M. B., & Liang, J. (2015). Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9350, pp. 62-69). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9350). Springer Verlag. https://doi.org/10.1007/978-3-319-24571-3_8

Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks. / Tajbakhsh, Nima; Gotway, Michael B.; Liang, Jianming.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9350 Springer Verlag, 2015. p. 62-69 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9350).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Tajbakhsh, N, Gotway, MB & Liang, J 2015, Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 9350, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9350, Springer Verlag, pp. 62-69, 18th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2015, Munich, Germany, 10/5/15. https://doi.org/10.1007/978-3-319-24571-3_8
Tajbakhsh N, Gotway MB, Liang J. Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9350. Springer Verlag. 2015. p. 62-69. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-24571-3_8
Tajbakhsh, Nima ; Gotway, Michael B. ; Liang, Jianming. / Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9350 Springer Verlag, 2015. pp. 62-69 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{de1428a758cf445688e7a4db2a64fe72,
title = "Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks",
abstract = "Computer-aided detection (CAD) can play a major role in diagnosing pulmonary embolism (PE) at CT pulmonary angiography (CTPA). However, despite their demonstrated utility, to achieve a clinically acceptable sensitivity, existing PE CAD systems generate a high number of false positives, imposing extra burdens on radiologists to adjudicate these superfluous CAD findings. In this study, we investigate the feasibility of convolutional neural networks (CNNs) as an effective mechanism for eliminating false positives. A critical issue in successfully utilizing CNNs for detecting an object in 3D images is to develop a “right” image representation for the object. Toward this end, we have developed a vesselaligned multi-planar image representation of emboli. Our image representation offers three advantages: (1) efficiency and compactness—concisely summarizing the 3D contextual information around an embolus in only 2 image channels, (2) consistency—automatically aligning the embolus in the 2-channel images according to the orientation of the affected vessel, and (3) expandability—naturally supporting data augmentation for training CNNs. We have evaluated our CAD approach using 121 CTPA datasets with a total of 326 emboli, achieving a sensitivity of 83{\%} at 2 false positives per volume. This performance is superior to the best performing CAD system in the literature, which achieves a sensitivity of 71{\%} at the same level of false positives. We have further evaluated our system using the entire 20 CTPA test datasets from the PE challenge. Our system outperforms the winning system from the challenge at 0mm localization error but is outperformed by it at 2mm and 5mm localization errors. In our view, the performance at 0mm localization error is more important than those at 2mm and 5mm localization errors.",
keywords = "Computer-aided detection, Convolutional neural networks, Pulmonary embolism, Vessel-aligned image representation",
author = "Nima Tajbakhsh and Gotway, {Michael B.} and Jianming Liang",
year = "2015",
doi = "10.1007/978-3-319-24571-3_8",
language = "English (US)",
isbn = "9783319245706",
volume = "9350",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "62--69",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Computer-aided pulmonary embolism detection using a novel vessel-aligned multi-planar image representation and convolutional neural networks

AU - Tajbakhsh, Nima

AU - Gotway, Michael B.

AU - Liang, Jianming

PY - 2015

Y1 - 2015

N2 - Computer-aided detection (CAD) can play a major role in diagnosing pulmonary embolism (PE) at CT pulmonary angiography (CTPA). However, despite their demonstrated utility, to achieve a clinically acceptable sensitivity, existing PE CAD systems generate a high number of false positives, imposing extra burdens on radiologists to adjudicate these superfluous CAD findings. In this study, we investigate the feasibility of convolutional neural networks (CNNs) as an effective mechanism for eliminating false positives. A critical issue in successfully utilizing CNNs for detecting an object in 3D images is to develop a “right” image representation for the object. Toward this end, we have developed a vesselaligned multi-planar image representation of emboli. Our image representation offers three advantages: (1) efficiency and compactness—concisely summarizing the 3D contextual information around an embolus in only 2 image channels, (2) consistency—automatically aligning the embolus in the 2-channel images according to the orientation of the affected vessel, and (3) expandability—naturally supporting data augmentation for training CNNs. We have evaluated our CAD approach using 121 CTPA datasets with a total of 326 emboli, achieving a sensitivity of 83% at 2 false positives per volume. This performance is superior to the best performing CAD system in the literature, which achieves a sensitivity of 71% at the same level of false positives. We have further evaluated our system using the entire 20 CTPA test datasets from the PE challenge. Our system outperforms the winning system from the challenge at 0mm localization error but is outperformed by it at 2mm and 5mm localization errors. In our view, the performance at 0mm localization error is more important than those at 2mm and 5mm localization errors.

AB - Computer-aided detection (CAD) can play a major role in diagnosing pulmonary embolism (PE) at CT pulmonary angiography (CTPA). However, despite their demonstrated utility, to achieve a clinically acceptable sensitivity, existing PE CAD systems generate a high number of false positives, imposing extra burdens on radiologists to adjudicate these superfluous CAD findings. In this study, we investigate the feasibility of convolutional neural networks (CNNs) as an effective mechanism for eliminating false positives. A critical issue in successfully utilizing CNNs for detecting an object in 3D images is to develop a “right” image representation for the object. Toward this end, we have developed a vesselaligned multi-planar image representation of emboli. Our image representation offers three advantages: (1) efficiency and compactness—concisely summarizing the 3D contextual information around an embolus in only 2 image channels, (2) consistency—automatically aligning the embolus in the 2-channel images according to the orientation of the affected vessel, and (3) expandability—naturally supporting data augmentation for training CNNs. We have evaluated our CAD approach using 121 CTPA datasets with a total of 326 emboli, achieving a sensitivity of 83% at 2 false positives per volume. This performance is superior to the best performing CAD system in the literature, which achieves a sensitivity of 71% at the same level of false positives. We have further evaluated our system using the entire 20 CTPA test datasets from the PE challenge. Our system outperforms the winning system from the challenge at 0mm localization error but is outperformed by it at 2mm and 5mm localization errors. In our view, the performance at 0mm localization error is more important than those at 2mm and 5mm localization errors.

KW - Computer-aided detection

KW - Convolutional neural networks

KW - Pulmonary embolism

KW - Vessel-aligned image representation

UR - http://www.scopus.com/inward/record.url?scp=84951010232&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84951010232&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-24571-3_8

DO - 10.1007/978-3-319-24571-3_8

M3 - Conference contribution

SN - 9783319245706

SN - 9783319245706

SN - 9783319245706

VL - 9350

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

SP - 62

EP - 69

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

PB - Springer Verlag

ER -