Paper Abstract and Keywords |
Presentation |
2020-01-30 13:25
Extracting and Visualization of Essential Features for Staining Translation of Pathological Images Ryoichi Koga, Noriaki Hashimoto, Tatsuya Yokota (NIT), Masato Nakaguro, Kei Kohno, Shigeo Nakamura (NUI), Ichiro Takeuchi, Hidekata Hontani (NIT) MI2019-116 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
In this manuscript, we propose a method for stain translation of pathology images. When one constructs a computer aided diagnosis system that can estimate the subtype of malignant lymphoma from a given H&E stained pathology image, one needs a set of training H&E stained whole-slide pathology images, in which the tumor regions are labeled because each H&E stained image includes both the tumor and non-tumor regions. It is thought not easy to collect enough number of labeled images as the labeling needs large human resources. We hence propose a stain translation method that can convert pathology images in which the tumor regions are stained by some immunostaining to virtual H&E stained images. Once we realize such the stain translation, then we can obtain the training images for training the subtype estimator straightforwardly. Our proposed method extracts image features that contain enough information for translating into any stain images, and a decoder that is specific to each immunostaining generates an virtual image stained with the specific immunostaining from the extracted image features. In addition, we visualize the extracted image features in this manuscript. In the experiments, realized a stain translation from CD20 stained images to H&E stained ones and visualized the corresponding image features. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
pathological image / staining translation / autoencoder / domain adversarial training / natural pre-image / / / |
Reference Info. |
IEICE Tech. Rep., vol. 119, no. 399, MI2019-116, pp. 215-218, Jan. 2020. |
Paper # |
MI2019-116 |
Date of Issue |
2020-01-22 (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 |
MI2019-116 |
|