Paper Abstract and Keywords |
Presentation |
2013-01-25 10:35
A Study on Bronchus Segmentation based on Machine Learning and Graph Cuts from Chest CT Image Yukitaka Nimura (Nagoya Univ.), Takayuki Kitasaka (Aichi Inst. of Tech.), Hirotoshi Honma (Sapporo-Kosei General Hospital), Hirotsugu Takabatake (Sapporo Minami-Sanjo Hospital), Masaki Mori (Sapporo-Kosei General Hospital), Hiroshi Natori (Keiwa-Kai Nishioka Hospital), Kensaku Mori (Nagoya Univ.) MI2012-98 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
This paper describes a bronchus segmentation method based on scale estimation and graph cuts of bronchi from chest CT image. A lot of previous methods utilize region growing or level set algorithms based on anatomical knowledge of connectivity of bronchi. However, it is difficult to extract bronchus precisely by only using these algorithms. Because connectivity of bronchi is often lost by partial volume effects, heartbeat, image noise or tumor in actual CT images. In this paper, we propose a method of bronchus segmentation based on another anatomical knowledge about bronchus. The proposed method detects voxels of medial lines of bronchi and its radius by using local intensity structure analysis, and extracts bronchi by using graph cuts segmentation that utilizes cost function with radius information. As the result, Jaccard index was 69.9%. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
bronchus / scale estimation / graph cuts / computer-aided diagnosis / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 112, no. 411, MI2012-98, pp. 191-196, Jan. 2013. |
Paper # |
MI2012-98 |
Date of Issue |
2013-01-17 (MI) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
MI2012-98 |
|