skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Hierarchical attention networks for information extraction from cancer pathology reports

Abstract

We explored how a deep learning (DL) approach based on hierarchical attention networks (HANs) can improve model performance for multiple information extraction tasks from unstructured cancer pathology reports compared to conventional methods that do not sufficiently capture syntactic and semantic contexts from free-text documents. Data for our analyses were obtained from 942 deidentified pathology reports collected by the National Cancer Institute Surveillance, Epidemiology, and End Results program. The HAN was implemented for 2 information extraction tasks: (1) primary site, matched to 12 International Classification of Diseases for Oncology topography codes (7 breast, 5 lung primary sites), and (2) histological grade classification, matched to G1–G4. Model performance metrics were compared to conventional machine learning (ML) approaches including naive Bayes, logistic regression, support vector machine, random forest, and extreme gradient boosting, and other DL models, including a recurrent neural network (RNN), a recurrent neural network with attention (RNN w/A), and a convolutional neural network. Our results demonstrate that for both information tasks, HAN performed significantly better compared to the conventional ML and DL techniques. In particular, across the 2 tasks, the mean micro and macro F-scores for the HAN with pretraining were (0.852,0.708), compared to naive Bayes (0.518, 0.213), logistic regression (0.682,more » 0.453), support vector machine (0.634, 0.434), random forest (0.698, 0.508), extreme gradient boosting (0.696, 0.522), RNN (0.505, 0.301), RNN w/A (0.637, 0.471), and convolutional neural network (0.714, 0.460). HAN-based DL models show promise in information abstraction tasks within unstructured clinical pathology reports.« less

Authors:
 [1];  [1];  [1];  [1];  [1];  [2];  [1];  [1]
  1. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States). Computational Science and Engineering Division
  2. National Cancer Inst., Bethesda, MD (United States). Surveillance Informatics Branch. Division of Cancer Control and Population Sciences
Publication Date:
Research Org.:
Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States); National Cancer Inst., Bethesda, MD (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR) (SC-21); USDOE National Nuclear Security Administration (NNSA); National Inst. of Health (NIH) (United States)
OSTI Identifier:
1430375
Alternate Identifier(s):
OSTI ID: 1474695
Grant/Contract Number:  
AC05-00OR22725; AC02-06CH11357; AC52-07NA27344; AC52-06NA25396
Resource Type:
Journal Article: Published Article
Journal Name:
Journal of the American Medical Informatics Association
Additional Journal Information:
Journal Volume: 25; Journal Issue: 3; Journal ID: ISSN 1067-5027
Publisher:
Oxford University Press
Country of Publication:
United States
Language:
English
Subject:
97 MATHEMATICS AND COMPUTING; 60 APPLIED LIFE SCIENCES; clinical pathology reports; information retrieval; recurrent neural nets; attention networks; classification

Citation Formats

Gao, Shang, Young, Michael T., Qiu, John X., Yoon, Hong-Jun, Christian, James B., Fearn, Paul A., Tourassi, Georgia D., and Ramanthan, Arvind. Hierarchical attention networks for information extraction from cancer pathology reports. United States: N. p., 2017. Web. doi:10.1093/jamia/ocx131.
Gao, Shang, Young, Michael T., Qiu, John X., Yoon, Hong-Jun, Christian, James B., Fearn, Paul A., Tourassi, Georgia D., & Ramanthan, Arvind. Hierarchical attention networks for information extraction from cancer pathology reports. United States. doi:10.1093/jamia/ocx131.
Gao, Shang, Young, Michael T., Qiu, John X., Yoon, Hong-Jun, Christian, James B., Fearn, Paul A., Tourassi, Georgia D., and Ramanthan, Arvind. Thu . "Hierarchical attention networks for information extraction from cancer pathology reports". United States. doi:10.1093/jamia/ocx131.
@article{osti_1430375,
title = {Hierarchical attention networks for information extraction from cancer pathology reports},
author = {Gao, Shang and Young, Michael T. and Qiu, John X. and Yoon, Hong-Jun and Christian, James B. and Fearn, Paul A. and Tourassi, Georgia D. and Ramanthan, Arvind},
abstractNote = {We explored how a deep learning (DL) approach based on hierarchical attention networks (HANs) can improve model performance for multiple information extraction tasks from unstructured cancer pathology reports compared to conventional methods that do not sufficiently capture syntactic and semantic contexts from free-text documents. Data for our analyses were obtained from 942 deidentified pathology reports collected by the National Cancer Institute Surveillance, Epidemiology, and End Results program. The HAN was implemented for 2 information extraction tasks: (1) primary site, matched to 12 International Classification of Diseases for Oncology topography codes (7 breast, 5 lung primary sites), and (2) histological grade classification, matched to G1–G4. Model performance metrics were compared to conventional machine learning (ML) approaches including naive Bayes, logistic regression, support vector machine, random forest, and extreme gradient boosting, and other DL models, including a recurrent neural network (RNN), a recurrent neural network with attention (RNN w/A), and a convolutional neural network. Our results demonstrate that for both information tasks, HAN performed significantly better compared to the conventional ML and DL techniques. In particular, across the 2 tasks, the mean micro and macro F-scores for the HAN with pretraining were (0.852,0.708), compared to naive Bayes (0.518, 0.213), logistic regression (0.682, 0.453), support vector machine (0.634, 0.434), random forest (0.698, 0.508), extreme gradient boosting (0.696, 0.522), RNN (0.505, 0.301), RNN w/A (0.637, 0.471), and convolutional neural network (0.714, 0.460). HAN-based DL models show promise in information abstraction tasks within unstructured clinical pathology reports.},
doi = {10.1093/jamia/ocx131},
journal = {Journal of the American Medical Informatics Association},
number = 3,
volume = 25,
place = {United States},
year = {Thu Nov 16 00:00:00 EST 2017},
month = {Thu Nov 16 00:00:00 EST 2017}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record at 10.1093/jamia/ocx131

Save / Share: