Automated medical coding, an essential task for healthcare operation and
delivery, makes unstructured data manageable by predicting medical codes from
clinical documents. Recent advances in deep learning models in natural language
processing have been widely applied to this task. However, it lacks a unified
view of the design of neural network architectures for medical coding. This
review proposes a unified framework to provide a general understanding of the
building blocks of medical coding models and summarizes recent advanced models
under the proposed framework. Our unified framework decomposes medical coding
into four main components, i.e., encoder modules for text feature extraction,
mechanisms for building deep encoder architectures, decoder modules for
transforming hidden representations into medical codes, and the usage of
auxiliary information. Finally, we discuss key research challenges and future
directions.
Description
A Unified Review of Deep Learning for Automated Medical Coding
%0 Generic
%1 ji2022unified
%A Ji, Shaoxiong
%A Sun, Wei
%A Dong, Hang
%A Wu, Honghan
%A Marttinen, Pekka
%D 2022
%K clinical-coding deep-learning hierarchies icd medical-coding multi-label_classification myown ontology public-health review survey text-classification
%T A Unified Review of Deep Learning for Automated Medical Coding
%U http://arxiv.org/abs/2201.02797
%X Automated medical coding, an essential task for healthcare operation and
delivery, makes unstructured data manageable by predicting medical codes from
clinical documents. Recent advances in deep learning models in natural language
processing have been widely applied to this task. However, it lacks a unified
view of the design of neural network architectures for medical coding. This
review proposes a unified framework to provide a general understanding of the
building blocks of medical coding models and summarizes recent advanced models
under the proposed framework. Our unified framework decomposes medical coding
into four main components, i.e., encoder modules for text feature extraction,
mechanisms for building deep encoder architectures, decoder modules for
transforming hidden representations into medical codes, and the usage of
auxiliary information. Finally, we discuss key research challenges and future
directions.
@misc{ji2022unified,
abstract = {Automated medical coding, an essential task for healthcare operation and
delivery, makes unstructured data manageable by predicting medical codes from
clinical documents. Recent advances in deep learning models in natural language
processing have been widely applied to this task. However, it lacks a unified
view of the design of neural network architectures for medical coding. This
review proposes a unified framework to provide a general understanding of the
building blocks of medical coding models and summarizes recent advanced models
under the proposed framework. Our unified framework decomposes medical coding
into four main components, i.e., encoder modules for text feature extraction,
mechanisms for building deep encoder architectures, decoder modules for
transforming hidden representations into medical codes, and the usage of
auxiliary information. Finally, we discuss key research challenges and future
directions.},
added-at = {2022-01-14T09:13:04.000+0100},
author = {Ji, Shaoxiong and Sun, Wei and Dong, Hang and Wu, Honghan and Marttinen, Pekka},
biburl = {https://www.bibsonomy.org/bibtex/266c539310ea4e1ecbdb60a08fd180a8b/hangdong},
description = {A Unified Review of Deep Learning for Automated Medical Coding},
interhash = {ed1a8a07d060b970d02642fd0988e61e},
intrahash = {66c539310ea4e1ecbdb60a08fd180a8b},
keywords = {clinical-coding deep-learning hierarchies icd medical-coding multi-label_classification myown ontology public-health review survey text-classification},
note = {cite arxiv:2201.02797},
timestamp = {2022-01-14T09:13:32.000+0100},
title = {A Unified Review of Deep Learning for Automated Medical Coding},
url = {http://arxiv.org/abs/2201.02797},
year = 2022
}