Paper Abstract and Keywords |
Presentation |
2015-09-11 11:35
Feature Words to Predict Long Post-Operatively Stay in Semi-structured Medical Records Takanori Yamashita, Haruka Kubo, Yuusuke Adachi, Yoshifumi Wakata, Brendan Flanagan (Kyushu Univ.), Hidehisa Soejima (Kumamoto Hospital), Naoki Nakashima, Sachio Hirokawa (Kyushu Univ.) NLC2015-32 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
The hospital information system has expanded, and importance of medical data utilization is increasing. Electronic medical data include structured digital data and unstructured text data. It is expected that analyzing unstructured text data contribute to clinical decision support and improve medical process.
The present paper analyzes the words that appear in semi-structured medical chart to predict the postoperative length of stay. SVM (support vector machine) were applied to extract feature words that characterize the medical charts of patients who stayed longer than the standard hospitalization. Three measures were proposed and the prediction performance for importance of the feature word was evaluated with respect to each measure. We extract the words that related clinical test, and show high frequency of the words in Objective item. Furthermore, we apply to show feature words that characterize the medical charts of long-hospitalization by Decision tree. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
medical chart / POMR / SVM(support vector machine) / Decision Tree / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 115, no. 222, NLC2015-32, pp. 75-79, Sept. 2015. |
Paper # |
NLC2015-32 |
Date of Issue |
2015-09-03 (NLC) |
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 |
NLC2015-32 |
|