Paper Abstract and Keywords |
Presentation |
2023-03-07 17:29
[Short Paper]
High Accuracy Segmentation of trachea and bronchus using 3D U-Net Tatsuya Ogasa, Rikuto Kuroda, Yoshiki Kawata, Hidenobu Suzuki (Tokushima Univ), Yuzi Matsumoto, Takaki Tsuchida, Masahiko Kusumoto (NCC), Noboru Niki (Medical Science Institute Inc.) MI2022-128 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
High-Accuracy trachea and bronchus segmentation is required for lymph node analysis in lung cancer. Since manual segmentation is time-consuming, highly accurate automatic trachea and bronchus segmentation is required. In this paper, Res3D U-Net is used to segmentation trachea, bronchus, and lymph nodes from 3D CT images. The labels of trachea, bronchus, and lymph nodes for training are created in two steps in a human-in-the-loop workflow to improve efficiency. Next, the training data is used to train Res3D U-Net, and the automatic segmentation results are evaluated to demonstrate the effectiveness of the training data creation method. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
airway / lymph-node / segmentation / deep learning / U-Net / Human-in-the-loop / / |
Reference Info. |
IEICE Tech. Rep., vol. 122, no. 417, MI2022-128, pp. 217-220, March 2023. |
Paper # |
MI2022-128 |
Date of Issue |
2023-02-27 (MI) |
ISSN |
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 |
MI2022-128 |
|