Paper Abstract and Keywords |
Presentation |
2012-07-20 11:00
Preliminary study for undersampling non-lesion voxel in training of voxel-based classification for lesion detection Yukihiro Nomura, Mitsutaka Nemoto, Yoshitaka Masutani, Shouhei Hanaoka, Soichiro Miki, Takeharu Yoshikawa, Naoto Hayashi, Kuni Ohtomo (The Univ. of Tokyo) MI2012-32 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
In voxel-based classification for lesion detection, training of a classifier is time-consuming since the number of training voxels is huge. Moreover, the number of non-lesion voxels is extremely larger than that of lesion voxels. So, it is important to reduce the number of non-lesion voxels for training. In this short report, we investigated relation between undersampling non-lesion voxels in training of voxel based classifier and detection accuracy by using cerebral aneurysm detection in MR angiograms. In the experimental study, it is desirable to train the classifier using all cases/voxels, but undersampling non-lesion voxels have potential to effective when time required for training is time-consuming or the number of training sample is excessive. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
computer-assisted detection (CAD) / voxel-based classifier / undersampling / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 112, no. 142, MI2012-32, pp. 59-64, July 2012. |
Paper # |
MI2012-32 |
Date of Issue |
2012-07-12 (MI) |
ISSN |
Print edition: ISSN 0913-5685 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 |
MI2012-32 |
|