A Residual CNN Model for ICD Assignment

DOI

10.2991/978-94-6463-094-7_26How to use a DOI?

Keywords

ICD codes; Residual CNN; MIMIC-III

Abstract

International Classification of Diseases (ICD) has been used as a standardized way of classifying a diagnosis or a medical procedure. ICD has also been employed to keep track of illness progression and treatment purposes. However, the assignment methods often require manual input of medical professionals and therefore time consuming and prone to human errors. By automating the assignment of ICD-9 codes to clinical notes we can effectively save time and human resources. In this light, this study proposed a residual convolution neural network leveraging label co-occurrence to measure label correlation and a label attention mechanism to capture label-dependent features. The model was fine-tuned by changing its hyper-parameters which have included dropout probabilities, CNN kernel size and its output size. The empirical findings suggested that the model has outperformed conventional approaches with 93.6% for Micro-AUC, 91.8% for Macro-AUC, 70.0% Micro-F1, and 64.6% for Macro-F1.

Copyright

© 2022 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Darryl Lin-Wei Cheng
AU  - Choo-Yee Ting
AU  - Chiung Ching Ho
PY  - 2022
DA  - 2022/12/27
TI  - A Residual CNN Model for ICD Assignment
BT  - Proceedings of the International Conference on Computer, Information Technology and Intelligent Computing (CITIC 2022)
PB  - Atlantis Press
SP  - 331
EP  - 341
SN  - 2589-4900
UR  - https://doi.org/10.2991/978-94-6463-094-7_26
DO  - 10.2991/978-94-6463-094-7_26
ID  - Cheng2022
ER  -