Presentation | 2011-01-20 Proposal of a novel boosting algorithm regularized by a global shape and its performance evaluation Kiyo SHINDO, Akinobu SHIMIZU, Hidefumi KOBATAKE, Shigeru NAWANO, Kenji SHINOZAKI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | A conventional ensemble learning based segmentation often outputs unnatural shape, because of voxel-wise discrimination independent to its neighboring voxels. This study proposes an ensemble learning algorithm that minimizes not only error-related loss but also global shape-related loss. Actually the shape-related loss is defined as a summation of squared difference over an image between an extracted result in each learning stage and its reconstructed image from a shape sub-space. It is expected that an extracted shape becomes natural by minimizing a loss function with the shape-related loss. This paper shows effectiveness of the proposed algorithm by using a synthetic image. In addition, it presents results of applying this method to actual three dimensional CT images followed by discussion about the superiority against a conventional loss function without any shape-related loss. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Boosting / ensemble / shape / CT / Spleen / Segmentation |
Paper # | MI2010-116 |
Date of Issue |
Conference Information | |
Committee | MI |
---|---|
Conference Date | 2011/1/12(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Medical Imaging (MI) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Proposal of a novel boosting algorithm regularized by a global shape and its performance evaluation |
Sub Title (in English) | |
Keyword(1) | Boosting |
Keyword(2) | ensemble |
Keyword(3) | shape |
Keyword(4) | CT |
Keyword(5) | Spleen |
Keyword(6) | Segmentation |
1st Author's Name | Kiyo SHINDO |
1st Author's Affiliation | Tokyo University of Agriculture and Technology() |
2nd Author's Name | Akinobu SHIMIZU |
2nd Author's Affiliation | Tokyo University of Agriculture and Technology |
3rd Author's Name | Hidefumi KOBATAKE |
3rd Author's Affiliation | Tokyo University of Agriculture and Technology |
4th Author's Name | Shigeru NAWANO |
4th Author's Affiliation | International University of Health and welfare |
5th Author's Name | Kenji SHINOZAKI |
5th Author's Affiliation | National Kyusyu Cancer Center |
Date | 2011-01-20 |
Paper # | MI2010-116 |
Volume (vol) | vol.110 |
Number (no) | 364 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |