Paper Abstract and Keywords |
Presentation |
2021-03-17 10:30
Study on automated anatomical labeling of abdominal arteries using Spectral-based Convolutional Graph Neural Networks Yuta Hibi, Yuichiro Hayashi (Nagoya Univ), Takayuki Kitasaka (Aichi Institute of Tech), Hayato Itoh, Masahiro Oda (Nagoya Univ), Kazunari Misawa (Aichi Cancer Center Hospital), Kensaku Mori (Nagoya University/NII) MI2020-89 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
In this study, we report an automated anatomical labeling method of abdominal arteries using Spectral-based Convolutional Graph Neural Networks. In laparoscopic surgery, which is widely performed today, it is difficult to understand the vascular structure due to the narrow field of laparoscope camera. Therefore, computer assistance is desired to help understanding of grasping vascular structure on surgeons by presenting the results of automated anatomical labeling of abdominal arteries. The use of a wide range of vascular features is important for learning vascular structures, and propose automated anatomical labeling of abdominal arteries by ChebNet that can handle a wide range of graph convolution. A maximum F value of 93.1% was achieved by introducing a weighted softmax cross entropy loss to reduce the imbalance in the data set. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
blood vessel / CT volume / anatomical names recognition / blood vessel structures analysis / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 120, no. 431, MI2020-89, pp. 176-181, March 2021. |
Paper # |
MI2020-89 |
Date of Issue |
2021-03-08 (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 |
MI2020-89 |
|