Classifying cancer pathology reports with hierarchical self-attention networks

Shang Gao; John X. Qiu; Mohammed Alawad; Jacob D. Hinkle; Noah Schaefferkoetter; Hong Jun Yoon; Blair Christian; Paul A. Fearn; Lynne Penberthy; Xiao Cheng Wu; Linda Coyle; Georgia Tourassi; Arvind Ramanathan

doi:10.1016/j.artmed.2019.101726

Classifying cancer pathology reports with hierarchical self-attention networks

Shang Gao, John X. Qiu, Mohammed Alawad, Jacob D. Hinkle, Noah Schaefferkoetter, Hong Jun Yoon, Blair Christian, Paul A. Fearn, Lynne Penberthy, Xiao Cheng Wu, Linda Coyle, Georgia Tourassi, Arvind Ramanathan

Biostatistics and Biomedical Informatics

Research output: Contribution to journal › Article › peer-review

38 Scopus citations

Abstract

We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks – site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data – Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.

Original language	English
Article number	101726
Journal	Artificial Intelligence in Medicine
Volume	101
DOIs	https://doi.org/10.1016/j.artmed.2019.101726
State	Published - Nov 2019

Funding

This work has been supported in part by the Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) program established by the U.S. Department of Energy (DOE) and the National Cancer Institute (NCI) of the National Institutes of Health. This work was performed under the auspices of the U.S. Department of Energy by Argonne National Laboratory under Contract DE-AC02-06-CH11357, Lawrence Livermore National Laboratory under Contract DEAC52-07NA27344, Los Alamos National Laboratory under Contract DE-AC5206NA25396, and Oak Ridge National Laboratory under Contract DE-AC05-00OR22725. This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725.

Funders	Funder number
National Institutes of Health
U.S. Department of Energy
National Cancer Institute
Office of Science
Argonne National Laboratory	DE-AC02-06-CH11357
Argonne National Laboratory
Lawrence Livermore National Laboratory	DEAC52-07NA27344
Lawrence Livermore National Laboratory
Oak Ridge National Laboratory	DE-AC05-00OR22725
Oak Ridge National Laboratory
Los Alamos National Laboratory	DE-AC5206NA25396
Los Alamos National Laboratory

Keywords

Cancer pathology reports
Clinical reports
Deep learning
Natural language processing
Text classification

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.artmed.2019.101726

Cite this

Gao, S., Qiu, J. X., Alawad, M., Hinkle, J. D., Schaefferkoetter, N., Yoon, H. J., Christian, B., Fearn, P. A., Penberthy, L., Wu, X. C., Coyle, L., Tourassi, G., & Ramanathan, A. (2019). Classifying cancer pathology reports with hierarchical self-attention networks. Artificial Intelligence in Medicine, 101, Article 101726. https://doi.org/10.1016/j.artmed.2019.101726

@article{bfce8899ebac4307b86dbe1548c12be9,

title = "Classifying cancer pathology reports with hierarchical self-attention networks",

abstract = "We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks – site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data – Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.",

keywords = "Cancer pathology reports, Clinical reports, Deep learning, Natural language processing, Text classification",

author = "Shang Gao and Qiu, {John X.} and Mohammed Alawad and Hinkle, {Jacob D.} and Noah Schaefferkoetter and Yoon, {Hong Jun} and Blair Christian and Fearn, {Paul A.} and Lynne Penberthy and Wu, {Xiao Cheng} and Linda Coyle and Georgia Tourassi and Arvind Ramanathan",

note = "Publisher Copyright: {\textcopyright} 2019 The Authors",

year = "2019",

month = nov,

doi = "10.1016/j.artmed.2019.101726",

language = "English",

volume = "101",

journal = "Artificial Intelligence in Medicine",

issn = "0933-3657",

publisher = "Elsevier",

}

TY - JOUR

T1 - Classifying cancer pathology reports with hierarchical self-attention networks

AU - Gao, Shang

AU - Qiu, John X.

AU - Alawad, Mohammed

AU - Hinkle, Jacob D.

AU - Schaefferkoetter, Noah

AU - Yoon, Hong Jun

AU - Christian, Blair

AU - Fearn, Paul A.

AU - Penberthy, Lynne

AU - Wu, Xiao Cheng

AU - Coyle, Linda

AU - Tourassi, Georgia

AU - Ramanathan, Arvind

PY - 2019/11

Y1 - 2019/11

N2 - We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks – site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data – Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.

AB - We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks – site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data – Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.

KW - Cancer pathology reports

KW - Clinical reports

KW - Deep learning

KW - Natural language processing

KW - Text classification

UR - http://www.scopus.com/inward/record.url?scp=85073144666&partnerID=8YFLogxK

U2 - 10.1016/j.artmed.2019.101726

DO - 10.1016/j.artmed.2019.101726

M3 - Article

C2 - 31813492

AN - SCOPUS:85073144666

SN - 0933-3657

VL - 101

JO - Artificial Intelligence in Medicine

JF - Artificial Intelligence in Medicine

M1 - 101726

ER -

Classifying cancer pathology reports with hierarchical self-attention networks

Abstract

Funding

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this