Skip to content

Drajan/DDxNet

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 

DDxNet: A Multi-Specialty Diagnostic Model for Clinical Time-Series

DDxNet is a novel fully convolutional neural network architecture for time-varying clinical data. We demonstrate it's effectiveness for a variety of diagnostic tasks involving different modalities (ECG/EEG/EHR), required level of characterization (abnormality detection/phenotyping) and data fidelity (single-lead ECG/22-channel EEG). Using multiple challenging benchmark problems with EEG, ECG and EHR, we show that DDxNet produces high-fidelity predictive models in all cases, and more importantly provides significant performance gains over methods specifically designed for each of those problems. The architecture is depicted in the figure below:

DDxNet Architecture


Usage

Example Usage: python train.py

Benchmark Problems and Datasets

EEG-based abnormality detection - the TUH abnormal corpus can be downloaded from https://www.isip.piconepress.com/projects/tuh_eeg/downloads/tuh_eeg_abnormal/v2.0.0/ As part of pre-processing, MFCCs are extracted using the librosa library (https://librosa.github.io/librosa/)

ECG-based arrhythmia classification - the Physionet MIT-BIH arrhythmia database can be downloaded from https://physionet.org/physiobank/database/mitdb/

ECG-based myocardial infarction detection - the Physionet PTBDB database can be downloaded from https://physionet.org/physiobank/database/ptbdb/

For both ECG problems, the data is pre-processed as described in the paper 'ECG Heartbeat Classification: A Deep Transferable Representation' (https://arxiv.org/pdf/1805.00794.pdf) and can be downloaded from kaggle (https://www.kaggle.com/shayanfazeli/heartbeat)

EHR-based phenotyping of ICU patients - the Physionet MIMIC-III EHR database can be downloaded from https://mimic.physionet.org/ The benchmark dataset for phenotyping was then prepared using the mimic3-benchmarks repository (https://github.com/YerevaNN/mimic3-benchmarks)

Requirements

  • python 3.6.8
  • numpy = 1.16.2
  • pandas = 0.24.1
  • torch = 1.0.1
  • scikit-learn = 0.20.2
  • matplotlib = 3.0.2
  • bokeh = 1.0.4
  • h5py = 2.9.0
  • pip = 19.0.3
  • python-dateutil = 2.8.0
  • sklearn = 0.0

Citations

If you find DDxNet useful in your research, please cite the following paper:

@article{thiagarajan2020ddxnet,
  title={DDxNet: a deep learning model for automatic interpretation of electronic health records, electrocardiograms and electroencephalograms},
  author={Thiagarajan, Jayaraman J and Rajan, Deepta and Katoch, Sameeksha and Spanias, Andreas},
  journal={Scientific reports},
  volume={10},
  number={1},
  pages={1--11},
  year={2020},
  publisher={Nature Publishing Group}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages