Skip to content

lywey/DMKAP

Repository files navigation

DMKAP

This project was implemented using the Tensorflow framework.

Tensorflow 2.2.0, Python 3.7, cuda 10.2, cudnn 7.6.5

Running DMKAP

  1. You will first need to request access for MIMIC-III, Single-level Clinical Classification Software(CCS) for ICD-9-CM, and Multi-level CCS.

  2. Use "process_mimic.py" to process MIMIC-III dataset and generate a suitable training dataset.

  3. Use "build_trees.py" to build files that contain the ancestor information of each medical code.

  4. Use "process_treeseq.py" to generate knowledge sequences for model input. Use "process_label.py" to generate the labels.

  5. Use "process_gcnlabel.py" to generate the labels required by the GCN embedding module, and use "process_adjacencylist.py" to generate the adjacency matrix required by the GCN embedding module.

  6. Use "train.py" to generate GCN Embedding in the "gcn_embedding" folder. Before to execute this algorithm, it is necessary to install these required packages shown in the file named ' requirements.txt '. You can also install all the required packages by just using one command :

    $ pip install -r requirements.txt
    
  7. Run DMKAP by using "run.py".

About

Knowledge-Aware Representation Learning for Diagnosis Prediction

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages