Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting

Medical Student Annotation Collaborative

doi:10.1007/978-3-030-11166-3_7

Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting

Medical Student Annotation Collaborative

The University of Manchester

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

196 Downloads (Pure)

Abstract

We propose a new method for fully automatic landmark localisation using Convolutional Neural Networks (CNNs). Training a CNN to estimate a Gaussian response (“heatmap”) around each target point is known to be effective for this task. We show that better results can be obtained by training a CNN to predict the offset to the target point at every location, then using these predictions to vote for the point position. We show the advantages of the approach, including those of using a novel loss function and weighting scheme. We evaluate on a dataset of radiographs of child hips, including both normal and severely diseased cases. We show the effect of varying the training set size. Our results show significant improvements in accuracy and robustness for the proposed method compared to a standard heatmap prediction approach and comparable results with a traditional Random Forest method.

Original language	English
Title of host publication	Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers
Editors	Guoyan Zheng, Tomaž Vrtovec, Jianhua Yao, Jose M. Pozo
Publisher	Springer Nature
Pages	73-85
Number of pages	13
Volume	11404 LNCS
ISBN (Electronic)	978-3-030-11166-3
ISBN (Print)	978-3-030-11165-6
DOIs	https://doi.org/10.1007/978-3-030-11166-3_7
Publication status	Published - 9 Jan 2019
Event	6th International Workshop on Computational Methods and Clinical Applications in Musculoskeletal Imaging, MSKI 2018 was held in conjunction with Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018 - Granada, Spain Duration: 16 Sept 2018 → 20 Sept 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11404 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	6th International Workshop on Computational Methods and Clinical Applications in Musculoskeletal Imaging, MSKI 2018 was held in conjunction with Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018
Country/Territory	Spain
City	Granada
Period	16/09/18 → 20/09/18

Keywords

Convolutional neural network (CNN)
Deep learning
Fully convolutional network (FCN)
Paediatrics
Perthes disease
Voting
X-rays

Access to Document

10.1007/978-3-030-11166-3_7

davison_miccai-mski2018_CNNAccepted author manuscript, 1.22 MB

Cite this

Medical Student Annotation Collaborative (2019). Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting. In G. Zheng, T. Vrtovec, J. Yao, & J. M. Pozo (Eds.), Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers (Vol. 11404 LNCS, pp. 73-85). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11404 LNCS). Springer Nature. https://doi.org/10.1007/978-3-030-11166-3_7

Medical Student Annotation Collaborative. / Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting. Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. editor / Guoyan Zheng ; Tomaž Vrtovec ; Jianhua Yao ; Jose M. Pozo. Vol. 11404 LNCS Springer Nature, 2019. pp. 73-85 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{a7c3710f5fae45afab9a6ef35a6b98c6,

title = "Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting",

abstract = "We propose a new method for fully automatic landmark localisation using Convolutional Neural Networks (CNNs). Training a CNN to estimate a Gaussian response (“heatmap”) around each target point is known to be effective for this task. We show that better results can be obtained by training a CNN to predict the offset to the target point at every location, then using these predictions to vote for the point position. We show the advantages of the approach, including those of using a novel loss function and weighting scheme. We evaluate on a dataset of radiographs of child hips, including both normal and severely diseased cases. We show the effect of varying the training set size. Our results show significant improvements in accuracy and robustness for the proposed method compared to a standard heatmap prediction approach and comparable results with a traditional Random Forest method.",

keywords = "Convolutional neural network (CNN), Deep learning, Fully convolutional network (FCN), Paediatrics, Perthes disease, Voting, X-rays",

author = "Davison, {Adrian K.} and {Medical Student Annotation Collaborative} and Claudia Lindner and Perry, {Daniel C.} and Weisang Luo and Cootes, {Timothy F.}",

year = "2019",

month = jan,

day = "9",

doi = "10.1007/978-3-030-11166-3_7",

language = "English",

isbn = "978-3-030-11165-6",

volume = "11404 LNCS",

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

publisher = "Springer Nature",

pages = "73--85",

editor = "Guoyan Zheng and Toma{\v z} Vrtovec and Jianhua Yao and Pozo, {Jose M.}",

booktitle = "Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers",

address = "United States",

note = "6th International Workshop on Computational Methods and Clinical Applications in Musculoskeletal Imaging, MSKI 2018 was held in conjunction with Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018 ; Conference date: 16-09-2018 Through 20-09-2018",

}

Medical Student Annotation Collaborative 2019, Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting. in G Zheng, T Vrtovec, J Yao & JM Pozo (eds), Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. vol. 11404 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11404 LNCS, Springer Nature, pp. 73-85, 6th International Workshop on Computational Methods and Clinical Applications in Musculoskeletal Imaging, MSKI 2018 was held in conjunction with Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018, Granada, Spain, 16/09/18. https://doi.org/10.1007/978-3-030-11166-3_7

Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting. / Medical Student Annotation Collaborative.
Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. ed. / Guoyan Zheng; Tomaž Vrtovec; Jianhua Yao; Jose M. Pozo. Vol. 11404 LNCS Springer Nature, 2019. p. 73-85 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11404 LNCS).

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

TY - GEN

T1 - Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting

AU - Davison, Adrian K.

AU - Medical Student Annotation Collaborative

AU - Lindner, Claudia

AU - Perry, Daniel C.

AU - Luo, Weisang

AU - Cootes, Timothy F.

PY - 2019/1/9

Y1 - 2019/1/9

N2 - We propose a new method for fully automatic landmark localisation using Convolutional Neural Networks (CNNs). Training a CNN to estimate a Gaussian response (“heatmap”) around each target point is known to be effective for this task. We show that better results can be obtained by training a CNN to predict the offset to the target point at every location, then using these predictions to vote for the point position. We show the advantages of the approach, including those of using a novel loss function and weighting scheme. We evaluate on a dataset of radiographs of child hips, including both normal and severely diseased cases. We show the effect of varying the training set size. Our results show significant improvements in accuracy and robustness for the proposed method compared to a standard heatmap prediction approach and comparable results with a traditional Random Forest method.

AB - We propose a new method for fully automatic landmark localisation using Convolutional Neural Networks (CNNs). Training a CNN to estimate a Gaussian response (“heatmap”) around each target point is known to be effective for this task. We show that better results can be obtained by training a CNN to predict the offset to the target point at every location, then using these predictions to vote for the point position. We show the advantages of the approach, including those of using a novel loss function and weighting scheme. We evaluate on a dataset of radiographs of child hips, including both normal and severely diseased cases. We show the effect of varying the training set size. Our results show significant improvements in accuracy and robustness for the proposed method compared to a standard heatmap prediction approach and comparable results with a traditional Random Forest method.

KW - Convolutional neural network (CNN)

KW - Deep learning

KW - Fully convolutional network (FCN)

KW - Paediatrics

KW - Perthes disease

KW - Voting

KW - X-rays

UR - http://www.scopus.com/inward/record.url?scp=85060202884&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-11166-3_7

DO - 10.1007/978-3-030-11166-3_7

M3 - Conference contribution

AN - SCOPUS:85060202884

SN - 978-3-030-11165-6

VL - 11404 LNCS

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

SP - 73

EP - 85

BT - Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers

A2 - Zheng, Guoyan

A2 - Vrtovec, Tomaž

A2 - Yao, Jianhua

A2 - Pozo, Jose M.

PB - Springer Nature

T2 - 6th International Workshop on Computational Methods and Clinical Applications in Musculoskeletal Imaging, MSKI 2018 was held in conjunction with Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018

Y2 - 16 September 2018 through 20 September 2018

ER -

Medical Student Annotation Collaborative. Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting. In Zheng G, Vrtovec T, Yao J, Pozo JM, editors, Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. Vol. 11404 LNCS. Springer Nature. 2019. p. 73-85. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Epub 2019 Jan 9. doi: 10.1007/978-3-030-11166-3_7

Landmark Localisation in Radiographs Using Weighted Heatmap Displacement Voting

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this