講演抄録/キーワード |
講演名 |
2015-05-15 14:00
A study on bronchus segmentation based on machine learning method from chest CT Image ○Qier Meng(Nagoya Univ.)・Takayuki Kitasaka(AIT)・Yukitaka Nimura・Yoshihoko Nakamura・Masahiro Oda・Kensaku Mori(Nagoya Univ.) SIP2015-23 IE2015-23 PRMU2015-23 MI2015-23 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
This paper presents a method for extracting bronchus regions from 3D chest CT images that uses a combination of region growing, and voxel
classification based on and machine learning methods. Most of previous methods focus on tracing the bronchial tree by region growing algorithms, it always fails to trace the tree when the abnormal appears to interrupt
such as lung tumors. Our method is mainly based on detecting the candidate voxels which have the bronchial features and implementing the SVM method to select the appropriate candidates. First, we combined two types of tube enhancement filters to detect the candidate region having line structures based on the Hessian analysis, and a modified RRF(Radial Reach Filter).
Second, we calculate bronchial features derived from both image intensity and shape. At last, we classify the candidate voxels by a classifier which is trained by the SVM using the training dataset. We applied the proposed method to four cases of 3D chest CT images and showed that it could extract the bronchial tree more accuracy than the previous method. |
キーワード |
(和) |
/ / / / / / / |
(英) |
SVM / local intensity structure / Radial Reach Filter / Hessian analysis / / / / |
文献情報 |
信学技報, vol. 115, no. 25, MI2015-23, pp. 121-126, 2015年5月. |
資料番号 |
MI2015-23 |
発行日 |
2015-05-07 (SIP, IE, PRMU, MI) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
SIP2015-23 IE2015-23 PRMU2015-23 MI2015-23 |
|