Paper Abstract and Keywords |
Presentation |
2019-01-22 14:50
[Short Paper]
Towards Annotating Less Medical Images:
-- PGGAN-based MR Image Augmentation for Brain Tumor Detection -- Changhee Han (UTokyo), Hideaki Hayashi (Kyushu Univ.), Leonardo Rundo (Univ. Cambridge), Ryosuke Araki (Chubu Univ.), Yudai Nagano (UTokyo), Yujiro Furukawa (Kanto Rosai Hosp.), Giancarlo Mauri (Univ. Milano-Bicocca), Hideki Nakayama (UTokyo) MI2018-82 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
How can we tackle the lack of available annotated medical image data through Data Augmentation (DA) techniques for accurate computer-assisted diagnosis? To fill the data lack in the real image distribution, we synthesize brain contrast-enhanced Magnetic Resonance (MR) images---realistic but completely different from the original ones---using Generative Adversarial Networks (GANs). Especially, we exploit Progressive Growing of GANs (PGGANs) to generate original-sized 256 × 256 brain MR images. Our results show that this novel PGGAN-based medical DA method can achieve better performance, when combined with classical DA and GAN-based refinement, in convolutional neural network-based tumor detection and also in other medical imaging tasks. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Data Augmentation / Generative Adversarial Networks / Deep Learning / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 118, no. 412, MI2018-82, pp. 93-94, Jan. 2019. |
Paper # |
MI2018-82 |
Date of Issue |
2019-01-15 (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 |
MI2018-82 |
|