Paper Abstract and Keywords |
Presentation |
2021-03-17 10:15
Automatic Segmentation of bleeding from Laparoscopic Video using Cascade CNN Shota Yamamoto, Yuichiro Hayashi, Shintaro Morimitsu (Nagoya Univ.), Takayuki Kitasaka (Aichi Institute Tech.), Masahiro Oda (Nagoya Univ.), Nobuyoshi Takeshita, Masaaki Ito (NCC East), Kensaku Mori (Nagoya Univ.) MI2020-88 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
We describe a bleeding region segmentation method using cascade CNN from laparoscopic images. In order to support laparoscopic surgery, researches on recognition of the status of surgery by analyzing laparoscopic images are conducted. We focused on the bleeding during surgery and segmented the bleeding region from laparoscopic images using the U-Net. This method has the problem that detailed segmentation is difficult. Therefore, we introduce a cascade processing that uses the detection results of YOLOv3 as the information to be input to the U-Net. In our experiment, bleeding regions were segmented by applying the previous and the proposed method to the video of laparoscopic surgery. As the result, it was confirmed that the detailed bleeding regions could be segmented 10.9% of F-measure higher than the previous method by introducing the cascaded process. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Laparoscopic surgery / Surgical process analysis / Segmentation / Deep learning / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 120, no. 431, MI2020-88, pp. 172-175, March 2021. |
Paper # |
MI2020-88 |
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 |
MI2020-88 |
|