Artificial Enlargement of a Training Set for Statistical Shape Models: Application to Cardiac Images

Lötjönen, Jyrki; Antila, Kari; Lamminmäki, E.; Koikkalainen, Juha; Lilja, Mikko; Cootes, Timothy F.

doi:10.1007/11494621_10

Cited by 13 publications

(12 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in Table 1, the performance of the method with labeled volumes follows the accuracy estimated in [10], i.e. the modeling accuracy in terms of the number of deformation modes.…”

Section: Discussionmentioning

confidence: 58%

“…The standard PCA as well as the PCA with 24, 100 and 250 artificial modes were used (columns). The errors from top to bottom are for the lower limit of segmentation accuracy estimated in [10] (theoretical), labeled volumes (NMI based optimization), modified volumes (NMI) and real MR volumes using both NMI and Mahalanobis based optimization. Bottom: Comparison of optimization algorithms.…”

Section: Discussionmentioning

confidence: 99%

“…In [10], we evaluated several methods and the best one called a non-rigid movement strategy was adopted here. The method moves non-rigidly surface points inside a deformation sphere.…”

Section: Methodsmentioning

confidence: 99%

“…This means that the degrees of freedom in the model are limited and the model can not represent accurately all cases in a population from which the training set has been sampled. At least two techniques have been proposed to the problem: for example in [7,8], the approach was to decrease the dimension of the model, whereas in [9,10], the size of the training set was artificially increased.…”

Section: Introductionmentioning

confidence: 99%

“…In addition to testing different methods to enlarge the dataset artificially in [10], we studied the relation between the size of the training set and the ability of the model to represent an unseen object. We found that a heart model with 100 modes has enough degrees of freedom to reach about 1 mm segmentation accuracy for ventricles and atria.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Artificially Enlarged Training Set in Image Segmentation

Tölli

Koikkalainen

Lauerma

et al. 2006

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2006

View full text Add to dashboard Cite

Abstract. Due to small training sets, statistical shape models constrain often too much the deformation in medical image segmentation. Hence, an artificial enlargement of the training set has been proposed as a solution for the problem. In this paper, the error sources in the statistical shape model based segmentation were analyzed and the optimization processes were improved. The method was evaluated with 3D cardiac MR volume data. The enlargement method based on non-rigid movement produced good results -with 250 artificial modes, the average error for four-chamber model was 2.11 mm when evaluated using 25 subjects.

show abstract

“…As shown in Table 1, the performance of the method with labeled volumes follows the accuracy estimated in [10], i.e. the modeling accuracy in terms of the number of deformation modes.…”

Section: Discussionmentioning

confidence: 58%

Section: Discussionmentioning

confidence: 99%

“…In [10], we evaluated several methods and the best one called a non-rigid movement strategy was adopted here. The method moves non-rigidly surface points inside a deformation sphere.…”

Section: Methodsmentioning

confidence: 99%