Patch-based generative shape model and MDL model selection for statistical analysis of archipelagos

TY - GEN

T1 - Patch-based generative shape model and MDL model selection for statistical analysis of archipelagos

AU - Ganz, Melanie

AU - Nielsen, Mads

AU - Brandt, Sami

N1 - Conference code: 1

PY - 2010

Y1 - 2010

N2 - We propose a statistical generative shape model for archipelago-like structures. These kind of structures occur, for instance, in medical images, where our intention is to model the appearance and shapes of calcifications in x-ray radio graphs. The generative model is constructed by (1) learning a patch-based dictionary for possible shapes, (2) building up a time-homogeneous Markov model to model the neighbourhood correlations between the patches, and (3) automatic selection of the model complexity by the minimum description length principle. The generative shape model is proposed as a probability distribution of a binary image where the model is intended to facilitate sequential simulation. Our results show that a relatively simple model is able to generate structures visually similar to calcifications. Furthermore, we used the shape model as a shape prior in the statistical segmentation of calcifications, where the area overlap with the ground truth shapes improved significantly compared to the case where the prior was not used.

AB - We propose a statistical generative shape model for archipelago-like structures. These kind of structures occur, for instance, in medical images, where our intention is to model the appearance and shapes of calcifications in x-ray radio graphs. The generative model is constructed by (1) learning a patch-based dictionary for possible shapes, (2) building up a time-homogeneous Markov model to model the neighbourhood correlations between the patches, and (3) automatic selection of the model complexity by the minimum description length principle. The generative shape model is proposed as a probability distribution of a binary image where the model is intended to facilitate sequential simulation. Our results show that a relatively simple model is able to generate structures visually similar to calcifications. Furthermore, we used the shape model as a shape prior in the statistical segmentation of calcifications, where the area overlap with the ground truth shapes improved significantly compared to the case where the prior was not used.

U2 - 10.1007/978-3-642-15948-0_5

DO - 10.1007/978-3-642-15948-0_5

M3 - Article in proceedings

SN - 978-3-642-15947-3

T3 - Lecture notes in computer science

SP - 34

EP - 41

BT - Machine Learning in Medical Imaging

A2 - Wang, Fei

A2 - Yan, Pingkun

A2 - Suzuki, Kenji

A2 - Shen, Dinggang

PB - Springer

T2 - 1st International Workshop on Machine Learning in Medical Imaging

Y2 - 20 September 2010 through 20 September 2010

ER -