多クラスAdaBoostを用いた腹部静脈領域に対する解剖学的名称自動対応付け手法
Bibliographic Information
- Other Title
-
- タクラス AdaBoost オ モチイタ フクブ ジョウミャク リョウイキ ニ タイスル カイボウガクテキ メイショウ ジドウ タイオウ ズケ シュホウ
- A method for automated anatomical labeling of abdominal veins by using multi-class AdaBoost
Search this article
Abstract
In abdominal surgeries, understanding of structures of blood vessel is important. Surgeons utilize CT images for such purpose. In this paper, a method for automated anatomical labeling of the abdominal veins is presented. A thinning process is applied to a abdominal vein region extracted from a 3D CT image. The result of the thinning process is expressed as a tree structure. Firstly, a characteristic blood vessel is labeld with a rule-based method. For the rest of blood vessels, likelihoods are labeled by using a machine learning technique. Then, branches in the tree structure are labeld by searching a child branch whose likelihood is the maximum. In the experiment using 20 cases of abdominal vein images which are manually extracted from 3D CT images, recall rate, precision rate, and F-measure were 86.3%, 85.7%, and 86.0%, respectively.
IEICE Technical Report;MI2012-33
Journal
-
- 電子情報通信学会技術研究報告. MI, 医用画像
-
電子情報通信学会技術研究報告. MI, 医用画像 112 (142), 65-70, 2012-07
一般社団法人電子情報通信学会
