Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net.

Christian Payer; Darko Stern; Horst Bischof; Martin Urschler

doi:10.5220/0008975201240133

Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net.

Christian Payer, Darko Stern, Horst Bischof, Martin Urschler

Institute of Computer Graphics and Vision (7100)

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

Abstract

Localization and segmentation of vertebral bodies from spine CT volumes are crucial for pathological diagnosis, surgical planning, and postoperative assessment. However, fully automatic analysis of spine CT volumes is difficult due to the anatomical variation of pathologies, noise caused by screws and implants, and the large range of different field-of-views. We propose a fully automatic coarse to fine approach for vertebrae localization and segmentation based on fully convolutional CNNs. In a three-step approach, at first, a U-Net localizes the rough position of the spine. Then, the SpatialConfiguration-Net performs vertebrae localization and identification using heatmap regression. Finally, a U-Net performs binary segmentation of each identified vertebrae in a high resolution, before merging the individual predictions into the resulting multi-label vertebrae segmentation. The evaluation shows top performance of our approach, ranking first place and winning the MICCAI 2019 Large Scale Vertebrae Segmentation Challenge (VerSe 2019).

Original language	English
Title of host publication	VISAPP
Editors	Giovanni Maria Farinella, Petia Radeva, Jose Braz
Pages	124-133
Number of pages	10
ISBN (Electronic)	9789897584022
DOIs	https://doi.org/10.5220/0008975201240133
Publication status	Published - 1 Jan 2020
Event	16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications: VISIGRAPP 2021 - Virtuell, Austria Duration: 8 Feb 2021 → 10 Feb 2021

Conference

Conference	16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications
Country/Territory	Austria
City	Virtuell
Period	8/02/21 → 10/02/21

Keywords

SpatialConfiguration-Net
U-Net
VerSe 2019 Challenge
Vertebrae Localization
Vertebrae Segmentation

ASJC Scopus subject areas

Computer Vision and Pattern Recognition
Computer Science Applications
Computer Graphics and Computer-Aided Design

Access to Document

10.5220/0008975201240133Licence: CC BY-NC-ND 4.0

Cite this

Payer, C, Stern, D, Bischof, H & Urschler, M 2020, Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net. in GM Farinella, P Radeva & J Braz (eds), VISAPP. pp. 124-133, 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, Virtuell, Austria, 8/02/21. https://doi.org/10.5220/0008975201240133

@inproceedings{3c5f21a45d524637903e224fded5a347,

title = "Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net.",

abstract = "Localization and segmentation of vertebral bodies from spine CT volumes are crucial for pathological diagnosis, surgical planning, and postoperative assessment. However, fully automatic analysis of spine CT volumes is difficult due to the anatomical variation of pathologies, noise caused by screws and implants, and the large range of different field-of-views. We propose a fully automatic coarse to fine approach for vertebrae localization and segmentation based on fully convolutional CNNs. In a three-step approach, at first, a U-Net localizes the rough position of the spine. Then, the SpatialConfiguration-Net performs vertebrae localization and identification using heatmap regression. Finally, a U-Net performs binary segmentation of each identified vertebrae in a high resolution, before merging the individual predictions into the resulting multi-label vertebrae segmentation. The evaluation shows top performance of our approach, ranking first place and winning the MICCAI 2019 Large Scale Vertebrae Segmentation Challenge (VerSe 2019).",

keywords = "SpatialConfiguration-Net, U-Net, VerSe 2019 Challenge, Vertebrae Localization, Vertebrae Segmentation",

author = "Christian Payer and Darko Stern and Horst Bischof and Martin Urschler",

year = "2020",

month = jan,

day = "1",

doi = "10.5220/0008975201240133",

language = "English",

pages = "124--133",

editor = "Farinella, {Giovanni Maria} and Petia Radeva and Jose Braz",

booktitle = "VISAPP",

note = "16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications : VISIGRAPP 2021 ; Conference date: 08-02-2021 Through 10-02-2021",

}

TY - GEN

T1 - Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net.

AU - Payer, Christian

AU - Stern, Darko

AU - Bischof, Horst

AU - Urschler, Martin

PY - 2020/1/1

Y1 - 2020/1/1

N2 - Localization and segmentation of vertebral bodies from spine CT volumes are crucial for pathological diagnosis, surgical planning, and postoperative assessment. However, fully automatic analysis of spine CT volumes is difficult due to the anatomical variation of pathologies, noise caused by screws and implants, and the large range of different field-of-views. We propose a fully automatic coarse to fine approach for vertebrae localization and segmentation based on fully convolutional CNNs. In a three-step approach, at first, a U-Net localizes the rough position of the spine. Then, the SpatialConfiguration-Net performs vertebrae localization and identification using heatmap regression. Finally, a U-Net performs binary segmentation of each identified vertebrae in a high resolution, before merging the individual predictions into the resulting multi-label vertebrae segmentation. The evaluation shows top performance of our approach, ranking first place and winning the MICCAI 2019 Large Scale Vertebrae Segmentation Challenge (VerSe 2019).

AB - Localization and segmentation of vertebral bodies from spine CT volumes are crucial for pathological diagnosis, surgical planning, and postoperative assessment. However, fully automatic analysis of spine CT volumes is difficult due to the anatomical variation of pathologies, noise caused by screws and implants, and the large range of different field-of-views. We propose a fully automatic coarse to fine approach for vertebrae localization and segmentation based on fully convolutional CNNs. In a three-step approach, at first, a U-Net localizes the rough position of the spine. Then, the SpatialConfiguration-Net performs vertebrae localization and identification using heatmap regression. Finally, a U-Net performs binary segmentation of each identified vertebrae in a high resolution, before merging the individual predictions into the resulting multi-label vertebrae segmentation. The evaluation shows top performance of our approach, ranking first place and winning the MICCAI 2019 Large Scale Vertebrae Segmentation Challenge (VerSe 2019).

KW - SpatialConfiguration-Net

KW - U-Net

KW - VerSe 2019 Challenge

KW - Vertebrae Localization

KW - Vertebrae Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85083508673&partnerID=8YFLogxK

U2 - 10.5220/0008975201240133

DO - 10.5220/0008975201240133

M3 - Conference paper

SP - 124

EP - 133

BT - VISAPP

A2 - Farinella, Giovanni Maria

A2 - Radeva, Petia

A2 - Braz, Jose

T2 - 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications

Y2 - 8 February 2021 through 10 February 2021

ER -

Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net.

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint