A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution

Ali Gooya, Elaheh Mousavi, Christos Davatzikos, Hongen Liao

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

Abstract

Statistical shape models (SSMs) made using point sets are important tools to capture the variations within shape populations. One popular method for construction of SSMs is based on the Expectation-Maximization (EM) algorithm which establishes probabilistic matches between the model and training points. In this paper, we propose a novel Bayesian framework to automatically determine the optimal number of the model points. We use a Dirichlet distribution as a prior to enforce sparsity on the mixture weights of Gaussians. Insignificant model points are determined and pruned out using a quadratic programming technique. We apply our method to learn a sparse SSM from 15 manually segmented caudate nuclei data sets. The generalization ability of the proposed model compares favorably to a traditional EM based model.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages144-152
Number of pages9
Volume8090 LNCS
DOIs
StatePublished - 2013
Externally publishedYes
Event6th International Workshop on Augmented Reality Environments for Medical Imaging and Computer-Assisted Interventions, MIAR 2013 and 8th International Workshop, AE-CAI 2013, Held in Conjunction with MICCAI 2013 - Nagoya, Japan
Duration: Sep 22 2013Sep 22 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8090 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other6th International Workshop on Augmented Reality Environments for Medical Imaging and Computer-Assisted Interventions, MIAR 2013 and 8th International Workshop, AE-CAI 2013, Held in Conjunction with MICCAI 2013
CountryJapan
CityNagoya
Period9/22/139/22/13

Fingerprint

Dirichlet Distribution
Bayesian Approach
Model
Expectation Maximization
Quadratic programming
Expectation-maximization Algorithm
Quadratic Programming
Sparsity
Point Sets
Nucleus

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Gooya, A., Mousavi, E., Davatzikos, C., & Liao, H. (2013). A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8090 LNCS, pp. 144-152). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8090 LNCS). https://doi.org/10.1007/978-3-642-40843-4_16

A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution. / Gooya, Ali; Mousavi, Elaheh; Davatzikos, Christos; Liao, Hongen.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8090 LNCS 2013. p. 144-152 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8090 LNCS).

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

Gooya, A, Mousavi, E, Davatzikos, C & Liao, H 2013, A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 8090 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8090 LNCS, pp. 144-152, 6th International Workshop on Augmented Reality Environments for Medical Imaging and Computer-Assisted Interventions, MIAR 2013 and 8th International Workshop, AE-CAI 2013, Held in Conjunction with MICCAI 2013, Nagoya, Japan, 9/22/13. https://doi.org/10.1007/978-3-642-40843-4_16
Gooya A, Mousavi E, Davatzikos C, Liao H. A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8090 LNCS. 2013. p. 144-152. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-40843-4_16
Gooya, Ali ; Mousavi, Elaheh ; Davatzikos, Christos ; Liao, Hongen. / A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8090 LNCS 2013. pp. 144-152 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{dbb06a435c3848c592e9d275b8f2a42e,
title = "A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution",
abstract = "Statistical shape models (SSMs) made using point sets are important tools to capture the variations within shape populations. One popular method for construction of SSMs is based on the Expectation-Maximization (EM) algorithm which establishes probabilistic matches between the model and training points. In this paper, we propose a novel Bayesian framework to automatically determine the optimal number of the model points. We use a Dirichlet distribution as a prior to enforce sparsity on the mixture weights of Gaussians. Insignificant model points are determined and pruned out using a quadratic programming technique. We apply our method to learn a sparse SSM from 15 manually segmented caudate nuclei data sets. The generalization ability of the proposed model compares favorably to a traditional EM based model.",
author = "Ali Gooya and Elaheh Mousavi and Christos Davatzikos and Hongen Liao",
year = "2013",
doi = "10.1007/978-3-642-40843-4_16",
language = "English (US)",
isbn = "9783642408427",
volume = "8090 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "144--152",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - A Bayesian approach for construction of sparse statistical shape models using Dirichlet distribution

AU - Gooya, Ali

AU - Mousavi, Elaheh

AU - Davatzikos, Christos

AU - Liao, Hongen

PY - 2013

Y1 - 2013

N2 - Statistical shape models (SSMs) made using point sets are important tools to capture the variations within shape populations. One popular method for construction of SSMs is based on the Expectation-Maximization (EM) algorithm which establishes probabilistic matches between the model and training points. In this paper, we propose a novel Bayesian framework to automatically determine the optimal number of the model points. We use a Dirichlet distribution as a prior to enforce sparsity on the mixture weights of Gaussians. Insignificant model points are determined and pruned out using a quadratic programming technique. We apply our method to learn a sparse SSM from 15 manually segmented caudate nuclei data sets. The generalization ability of the proposed model compares favorably to a traditional EM based model.

AB - Statistical shape models (SSMs) made using point sets are important tools to capture the variations within shape populations. One popular method for construction of SSMs is based on the Expectation-Maximization (EM) algorithm which establishes probabilistic matches between the model and training points. In this paper, we propose a novel Bayesian framework to automatically determine the optimal number of the model points. We use a Dirichlet distribution as a prior to enforce sparsity on the mixture weights of Gaussians. Insignificant model points are determined and pruned out using a quadratic programming technique. We apply our method to learn a sparse SSM from 15 manually segmented caudate nuclei data sets. The generalization ability of the proposed model compares favorably to a traditional EM based model.

UR - http://www.scopus.com/inward/record.url?scp=84890929113&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84890929113&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-40843-4_16

DO - 10.1007/978-3-642-40843-4_16

M3 - Conference contribution

SN - 9783642408427

VL - 8090 LNCS

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

SP - 144

EP - 152

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

ER -