Probabilistic Modeling of Landmark Distances and Structure for Anomaly-proof Landmark Detection

Shouhei Hanaoka; Yoshitaka Masutani; Mitsutaka Nemoto; Yukihiro Nomura; Takeharu Yoshikawa; Naoto Hayashi; Naoki Yoshioka; Kuni Ohtomo

inria-00624109, version 1

Probabilistic Modeling of Landmark Distances and Structure for Anomaly-proof Landmark Detection

Shouhei Hanaoka ¹, Yoshitaka Masutani ^1, 2, Mitsutaka Nemoto ¹, Yukihiro Nomura ¹, Takeharu Yoshikawa ³, Naoto Hayashi ³, Naoki Yoshioka ⁴, Kuni Ohtomo ^1, 2

Proceedings of the Third International Workshop on Mathematical Foundations of Computational Anatomy - Geometrical and Statistical Methods for Modelling Biological Shape Variability (2011) 159-169

Abstract: A combinatorial optimization algorithm for detecting multiple anatomical landmarks is presented. It can determine the positions of over 100 landmarks concurrently, taking spatial correlations of all landmark pairs into account. Provided that a set of landmark candidate lists is given by sensitivity-optimized single-landmark detectors, the proposed algorithm can find the most probable combination of them through solving a MAP estimation-based combinatorial optimization problem. Additionally, it is designed to handle subjects with "segmentation anomaly of the spinal column," a common anatomical anomaly of the spine. The proposed system was evaluated with 156 landmarks in 50 datasets, using virtually created detector output sets. In the result, the algorithm achieved 97.6\% of spinal anomaly estimation accuracy even with 50 points of candidates given per landmark, as well as 96.2\% of accuracy in landmark candidate selection. From these results, usefulness of the proposed algorithm for subjects with spinal anomaly was suggested.

1: Department of Radiology, University of Tokyo
University of Tokyo
2: Division of Radiology and Biomedical Engineering, Graduate School of Medicine, University of Tokyo
University of Tokyo
3: Department of Computational Diagnostic Radiology and Preventive Medicine, University of Tokyo Hospital
University of Tokyo Hospital
4: Department of Integrated Imaging Informatics, University of Tokyo Hospital
University of Tokyo Hospital

Collaboration : Session : Poster

inria-00624109, version 1
http://hal.inria.fr/inria-00624109
oai:hal.inria.fr:inria-00624109
From: Xavier Pennec <>
Submitted on: Thursday, 15 September 2011 17:18:58
Updated on: Friday, 16 September 2011 09:31:27

See detailed view

Associated documents

PDF :

Export

Bibtex EndNote TEI RefWorks

%% inria-00624109, version 1
%% http://hal.inria.fr/inria-00624109
@inproceedings{hanaoka:inria-00624109,
    hal_id = {inria-00624109},
    url = {http://hal.inria.fr/inria-00624109},
    title = {{Probabilistic Modeling of Landmark Distances and Structure for Anomaly-proof Landmark Detection}},
    author = {Hanaoka, Shouhei and Masutani, Yoshitaka and Nemoto, Mitsutaka and Nomura, Yukihiro and Yoshikawa, Takeharu and Hayashi, Naoto and Yoshioka, Naoki and Ohtomo, Kuni},
    abstract = {{A combinatorial optimization algorithm for detecting multiple anatomical landmarks is presented. It can determine the positions of over 100 landmarks concurrently, taking spatial correlations of all landmark pairs into account. Provided that a set of landmark candidate lists is given by sensitivity-optimized single-landmark detectors, the proposed algorithm can find the most probable combination of them through solving a MAP estimation-based combinatorial optimization problem. Additionally, it is designed to handle subjects with "segmentation anomaly of the spinal column," a common anatomical anomaly of the spine. The proposed system was evaluated with 156 landmarks in 50 datasets, using virtually created detector output sets. In the result, the algorithm achieved 97.6\\% of spinal anomaly estimation accuracy even with 50 points of candidates given per landmark, as well as 96.2\\% of accuracy in landmark candidate selection. From these results, usefulness of the proposed algorithm for subjects with spinal anomaly was suggested.}},
    keywords = {Landmark ; Combinatorial optimization ; MAP estimation; Anatomical anomaly ; Computed tomography ; Spine},
    language = {English},
    affiliation = {Department of Radiology, University of Tokyo , Division of Radiology and Biomedical Engineering, Graduate School of Medicine, University of Tokyo , Department of Computational Diagnostic Radiology and Preventive Medicine, University of Tokyo Hospital , Department of Integrated Imaging Informatics, University of Tokyo Hospital},
    booktitle = {{Proceedings of the Third International Workshop on Mathematical Foundations of Computational Anatomy - Geometrical and Statistical Methods for Modelling Biological Shape Variability}},
    pages = {159-169},
    address = {Toronto, Canada},
    editor = {Pennec, Xavier and Joshi, Sarang and Nielsen, Mads },
    audience = {international },
    collaboration = {Session : Poster },
    year = {2011},
    month = Sep,
    pdf = {http://hal.inria.fr/inria-00624109/PDF/MFCA11\_P\_3.pdf},
}

%F inria-00624109, version 1
%0 Conference Proceedings
%U http://hal.inria.fr/inria-00624109
%2 INFO:INFO_OH
%T Probabilistic Modeling of Landmark Distances and Structure for Anomaly-proof Landmark Detection
%A Hanaoka, Shouhei
%A Masutani, Yoshitaka
%A Nemoto, Mitsutaka
%A Nomura, Yukihiro
%A Yoshikawa, Takeharu
%A Hayashi, Naoto
%A Yoshioka, Naoki
%A Ohtomo, Kuni
%+ Department of Radiology, University of Tokyo , Division of Radiology and Biomedical Engineering, Graduate School of Medicine, University of Tokyo , Department of Computational Diagnostic Radiology and Preventive Medicine, University of Tokyo Hospital , Department of Integrated Imaging Informatics, University of Tokyo Hospital
%X A combinatorial optimization algorithm for detecting multiple anatomical landmarks is presented. It can determine the positions of over 100 landmarks concurrently, taking spatial correlations of all landmark pairs into account. Provided that a set of landmark candidate lists is given by sensitivity-optimized single-landmark detectors, the proposed algorithm can find the most probable combination of them through solving a MAP estimation-based combinatorial optimization problem. Additionally, it is designed to handle subjects with "segmentation anomaly of the spinal column," a common anatomical anomaly of the spine. The proposed system was evaluated with 156 landmarks in 50 datasets, using virtually created detector output sets. In the result, the algorithm achieved 97.6\% of spinal anomaly estimation accuracy even with 50 points of candidates given per landmark, as well as 96.2\% of accuracy in landmark candidate selection. From these results, usefulness of the proposed algorithm for subjects with spinal anomaly was suggested.
%G English
%8 2011-09-22
%2 international
%B Proceedings of the Third International Workshop on Mathematical Foundations of Computational Anatomy - Geometrical and Statistical Methods for Modelling Biological Shape Variability
%2 Toronto
%2 Canada
%8 2011-09-22
%E Pennec, Xavier and Joshi, Sarang and Nielsen, Mads
%P 159-169
%K Landmark ; Combinatorial optimization ; MAP estimation; Anatomical anomaly ; Computed tomography ; Spine
%Z Session : Poster

<?xml version="1.0" encoding="UTF-8"?>
<listBibl xmlns="http://www.tei-c.org/ns/1.0" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML">
<listBibl type="inproceedings"><head>Peer-reviewed conferences/proceedings</head>
<biblStruct type="inproceedings" xml:id="inria-00624109"><analytic>
<author><persName><forename>Shouhei</forename><surname>Hanaoka</surname></persName>
<affiliation>
<orgName type="team">Department of Radiology, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
</author>
<author><persName><forename>Yoshitaka</forename><surname>Masutani</surname></persName>
<affiliation>
<orgName type="team">Department of Radiology, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
<affiliation>
<orgName type="team">Division of Radiology and Biomedical Engineering, Graduate School of Medicine, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
</author>
<author><persName><forename>Mitsutaka</forename><surname>Nemoto</surname></persName>
<affiliation>
<orgName type="team">Department of Radiology, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
</author>
<author><persName><forename>Yukihiro</forename><surname>Nomura</surname></persName>
<affiliation>
<orgName type="team">Department of Radiology, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
</author>
<author><persName><forename>Takeharu</forename><surname>Yoshikawa</surname></persName>
<affiliation>
<orgName type="team">Department of Computational Diagnostic Radiology and Preventive Medicine, University of Tokyo Hospital</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo Hospital</orgName>
</affiliation>
</author>
<author><persName><forename>Naoto</forename><surname>Hayashi</surname></persName>
<affiliation>
<orgName type="team">Department of Computational Diagnostic Radiology and Preventive Medicine, University of Tokyo Hospital</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo Hospital</orgName>
</affiliation>
</author>
<author><persName><forename>Naoki</forename><surname>Yoshioka</surname></persName>
<affiliation>
<orgName type="team">Department of Integrated Imaging Informatics, University of Tokyo Hospital</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo Hospital</orgName>
</affiliation>
</author>
<author><persName><forename>Kuni</forename><surname>Ohtomo</surname></persName>
<affiliation>
<orgName type="team">Department of Radiology, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
<affiliation>
<orgName type="team">Division of Radiology and Biomedical Engineering, Graduate School of Medicine, University of Tokyo</orgName>
<country>JP</country>
<orgName type="institution">University of Tokyo</orgName>
</affiliation>
</author>
<title type="main" level="a">Probabilistic Modeling of Landmark Distances and Structure for Anomaly-proof Landmark Detection</title></analytic><monogr><meeting>Proceedings of the Third International Workshop on Mathematical Foundations of Computational Anatomy - Geometrical and Statistical Methods for Modelling Biological Shape Variability</meeting><imprint><pubPlace><settlement>Toronto</settlement><country>Canada</country></pubPlace><biblScope type="publicationDate">2011-09-22</biblScope><biblScope type="pp">159-169</biblScope></imprint></monogr><idno type="HALid">http://hal.inria.fr/inria-00624109</idno><idno type="HALFile">http://hal.inria.fr/inria-00624109/PDF/MFCA11_P_3.pdf</idno></biblStruct>
</listBibl>
</listBibl>