Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting

Paul Bromiley; Eleni Kariki; Judith Adams; Timothy Cootes

doi:10.1007/978-3-319-55050-3_5

Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting

Paul Bromiley, Eleni Kariki, Judith Adams, Timothy Cootes

Division of Informatics, Imaging & Data Sciences (L5)

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

Abstract

We describe a system for fully automatic vertebra localisation and segmentation in 3D CT volumes containing arbitrary regions of the spine, with the aim of detecting osteoporotic fractures. To avoid the difficulties of high-resolution manual annotation on overlapping structures in 3D, the system consists of several 2D operations. First, a Random Forest regressor is used to localise the spinal midplane in a coronal maximum intensity projection. A 2D sagittal image showing the midplane is then produced. A second set of regressors are used to localise each vertebral body in this image. Finally, a Random Forest Regression Voting Constrained Local Model is used to segment each detected vertebra.

The system was evaluated on 402 CT volumes. 83% of vertebrae between T4 and L4 were detected and, of these, 97% were segmented with a mean error of less than or equal to 1mm . A simple classifier was applied to perform a fracture/non-fracture classification for each image, achieving 69% recall at 70% precision.

Original language	English
Title of host publication	Computational Methods and Clinical Applications for Spine Imaging
Subtitle of host publication	4th International Conference and Challenge, CSI 2016
Editors	Jianhua Yao, Tomaz Vrtovec, Guoyan Zheng, Alejandro Frangi, Ben Glocker, Shuo Li
Publisher	Springer Nature
Pages	51-63
Number of pages	13
Volume	LNCS 10182
ISBN (Electronic)	9783319550503
ISBN (Print)	9783319550497, 9783319550503
DOIs	https://doi.org/10.1007/978-3-319-55050-3_5
Publication status	Published - 1 Mar 2017

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
ISSN (Print)	0302-9743

Access to Document

10.1007/978-3-319-55050-3_5

Cite this

Bromiley, P., Kariki, E., Adams, J., & Cootes, T. (2017). Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting. In J. Yao, T. Vrtovec, G. Zheng, A. Frangi, B. Glocker, & S. Li (Eds.), Computational Methods and Clinical Applications for Spine Imaging: 4th International Conference and Challenge, CSI 2016 (Vol. LNCS 10182, pp. 51-63). (Lecture Notes in Computer Science). Springer Nature. Advance online publication. https://doi.org/10.1007/978-3-319-55050-3_5

Bromiley, Paul ; Kariki, Eleni ; Adams, Judith et al. / Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting. Computational Methods and Clinical Applications for Spine Imaging: 4th International Conference and Challenge, CSI 2016. editor / Jianhua Yao ; Tomaz Vrtovec ; Guoyan Zheng ; Alejandro Frangi ; Ben Glocker ; Shuo Li. Vol. LNCS 10182 Springer Nature, 2017. pp. 51-63 (Lecture Notes in Computer Science).

@inproceedings{e488ea7fd7354955a258242841ec583e,

title = "Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting",

abstract = "We describe a system for fully automatic vertebra localisation and segmentation in 3D CT volumes containing arbitrary regions of the spine, with the aim of detecting osteoporotic fractures. To avoid the difficulties of high-resolution manual annotation on overlapping structures in 3D, the system consists of several 2D operations. First, a Random Forest regressor is used to localise the spinal midplane in a coronal maximum intensity projection. A 2D sagittal image showing the midplane is then produced. A second set of regressors are used to localise each vertebral body in this image. Finally, a Random Forest Regression Voting Constrained Local Model is used to segment each detected vertebra.The system was evaluated on 402 CT volumes. 83% of vertebrae between T4 and L4 were detected and, of these, 97% were segmented with a mean error of less than or equal to 1mm . A simple classifier was applied to perform a fracture/non-fracture classification for each image, achieving 69% recall at 70% precision.",

author = "Paul Bromiley and Eleni Kariki and Judith Adams and Timothy Cootes",

year = "2017",

month = mar,

day = "1",

doi = "10.1007/978-3-319-55050-3_5",

language = "English",

isbn = "9783319550497",

volume = "LNCS 10182",

series = "Lecture Notes in Computer Science",

publisher = "Springer Nature",

pages = "51--63",

editor = "Jianhua Yao and Tomaz Vrtovec and Guoyan Zheng and Alejandro Frangi and Ben Glocker and Shuo Li",

booktitle = "Computational Methods and Clinical Applications for Spine Imaging",

address = "United States",

}

Bromiley, P , Kariki, E, Adams, J & Cootes, T 2017, Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting. in J Yao, T Vrtovec, G Zheng, A Frangi, B Glocker & S Li (eds), Computational Methods and Clinical Applications for Spine Imaging: 4th International Conference and Challenge, CSI 2016. vol. LNCS 10182, Lecture Notes in Computer Science, Springer Nature, pp. 51-63. https://doi.org/10.1007/978-3-319-55050-3_5

Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting. / Bromiley, Paul ; Kariki, Eleni; Adams, Judith et al.
Computational Methods and Clinical Applications for Spine Imaging: 4th International Conference and Challenge, CSI 2016. ed. / Jianhua Yao; Tomaz Vrtovec; Guoyan Zheng; Alejandro Frangi; Ben Glocker; Shuo Li. Vol. LNCS 10182 Springer Nature, 2017. p. 51-63 (Lecture Notes in Computer Science).

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

TY - GEN

T1 - Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting

AU - Bromiley, Paul

AU - Kariki, Eleni

AU - Adams, Judith

AU - Cootes, Timothy

PY - 2017/3/1

Y1 - 2017/3/1

N2 - We describe a system for fully automatic vertebra localisation and segmentation in 3D CT volumes containing arbitrary regions of the spine, with the aim of detecting osteoporotic fractures. To avoid the difficulties of high-resolution manual annotation on overlapping structures in 3D, the system consists of several 2D operations. First, a Random Forest regressor is used to localise the spinal midplane in a coronal maximum intensity projection. A 2D sagittal image showing the midplane is then produced. A second set of regressors are used to localise each vertebral body in this image. Finally, a Random Forest Regression Voting Constrained Local Model is used to segment each detected vertebra.The system was evaluated on 402 CT volumes. 83% of vertebrae between T4 and L4 were detected and, of these, 97% were segmented with a mean error of less than or equal to 1mm . A simple classifier was applied to perform a fracture/non-fracture classification for each image, achieving 69% recall at 70% precision.

AB - We describe a system for fully automatic vertebra localisation and segmentation in 3D CT volumes containing arbitrary regions of the spine, with the aim of detecting osteoporotic fractures. To avoid the difficulties of high-resolution manual annotation on overlapping structures in 3D, the system consists of several 2D operations. First, a Random Forest regressor is used to localise the spinal midplane in a coronal maximum intensity projection. A 2D sagittal image showing the midplane is then produced. A second set of regressors are used to localise each vertebral body in this image. Finally, a Random Forest Regression Voting Constrained Local Model is used to segment each detected vertebra.The system was evaluated on 402 CT volumes. 83% of vertebrae between T4 and L4 were detected and, of these, 97% were segmented with a mean error of less than or equal to 1mm . A simple classifier was applied to perform a fracture/non-fracture classification for each image, achieving 69% recall at 70% precision.

UR - http://dx.doi.org/10.1007/978-3-319-55050-3_5

U2 - 10.1007/978-3-319-55050-3_5

DO - 10.1007/978-3-319-55050-3_5

M3 - Conference contribution

SN - 9783319550497

SN - 9783319550503

VL - LNCS 10182

T3 - Lecture Notes in Computer Science

SP - 51

EP - 63

BT - Computational Methods and Clinical Applications for Spine Imaging

A2 - Yao, Jianhua

A2 - Vrtovec, Tomaz

A2 - Zheng, Guoyan

A2 - Frangi, Alejandro

A2 - Glocker, Ben

A2 - Li, Shuo

PB - Springer Nature

ER -

Bromiley P , Kariki E, Adams J, Cootes T. Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting. In Yao J, Vrtovec T, Zheng G, Frangi A, Glocker B, Li S, editors, Computational Methods and Clinical Applications for Spine Imaging: 4th International Conference and Challenge, CSI 2016. Vol. LNCS 10182. Springer Nature. 2017. p. 51-63. (Lecture Notes in Computer Science). Epub 2017 Mar 1. doi: 10.1007/978-3-319-55050-3_5

Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting

Abstract

Publication series

Access to Document

Other files and links

Fingerprint

ASPIRE™: Using Machine Learning to detect undiagnosed fractures in patients with osteoporosis

Cite this

Fully Automatic Localisation of Vertebrae in CT images using Random Forest Regression Voting

Abstract

Publication series

Access to Document

Other files and links

Fingerprint

Impacts

ASPIRE™: Using Machine Learning to detect undiagnosed fractures in patients with osteoporosis

Cite this