機械学習とグラフカットによる胸部CT像からの気管支抽出に関する検討

IR HANDLE Web Site Open Access

Bibliographic Information

Other Title
  • キカイ ガクシュウ ト グラフカット ニ ヨル キョウブ CTゾウ カラ ノ キカンシ チュウシュツ ニ カンスル ケントウ
  • A Study on Bronchus Segmentation based on Machine Learning and Graph Cuts from Chest CT Image

Search this article

Abstract

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%.

IEICE Technical Report;MI2012-98

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top