https://github.com/shaoxiongji/dcan
Dilated Convolutional Attention Network for Medical Code Assignment from Clinical Text, ClinicalNLP workshop at EMNLP 2020
https://github.com/shaoxiongji/dcan
clinical-nlp deep-neural-networks icd-coding
Last synced: about 1 month ago
JSON representation
Dilated Convolutional Attention Network for Medical Code Assignment from Clinical Text, ClinicalNLP workshop at EMNLP 2020
- Host: GitHub
- URL: https://github.com/shaoxiongji/dcan
- Owner: shaoxiongji
- License: gpl-3.0
- Created: 2020-09-30T05:35:15.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2021-08-19T12:54:13.000Z (over 3 years ago)
- Last Synced: 2025-04-12T04:51:50.856Z (about 1 month ago)
- Topics: clinical-nlp, deep-neural-networks, icd-coding
- Language: Python
- Homepage:
- Size: 29.3 KB
- Stars: 12
- Watchers: 3
- Forks: 4
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# DCAN
Dilated Convolutional Attention Network (DCAN), integrating dilated convolutions, residual connections, and label attention, for medical code assignment. It adopts dilated convolutions to capture complex medical patterns with a receptive field which increases exponentially with dilation size.
## Data
Download MIMIC-III dataset from [physionet](https://mimic.physionet.org).
Organize your data using the following structure
```
data
| D_ICD_DIAGNOSES.csv
| D_ICD_PROCEDURES.csv
| ICD9_descriptions
└───mimic3/
| | NOTEEVENTS.csv
| | DIAGNOSES_ICD.csv
| | PROCEDURES_ICD.csv
| | *_hadm_ids.csv
```
`ICD9_descriptions` is avaiable [in this repo](https://github.com/jamesmullenbach/caml-mimic/blob/master/mimicdata/ICD9_descriptions), and
`*_hadm_ids.csv` are avaiable [here](https://github.com/jamesmullenbach/caml-mimic/tree/master/mimicdata/mimic3).
`MIMIC_RAW_DSUMS` is available [here](https://physionet.org/works/ICD9CodingofDischargeSummaries/), while the rest file for MIMIC2 can be generated with their code.
If you use Python3 `consctruct_datasest.py` in `ICD9_Coding_of_Discharge_Summaries` to create data files, remember to convert dict object to list (line 82&83) and use `dict.items()` instead of `dict.iteritems()`.
Assign the directories of MIMIC data using `MIMIC_3_DIR`.
## Run
``python3 main.py``
Configs available at `options.py`.
Requirements:
- python 3.7
- pytorch 1.5.0
## Citation
```
@inproceedings{ji2020dilated,
title={Dilated Convolutional Attention Network for Medical Code Assignment from Clinical Text},
author={Ji, Shaoxiong and Cambria, Erik and Marttinen, Pekka},
booktitle={3rd Clinical Natural Language Processing Workshop at EMNLP},
year={2020}
}
```
## References
- https://github.com/jamesmullenbach/caml-mimic
- https://github.com/foxlf823/Multi-Filter-Residual-Convolutional-Neural-Network