Implementation for Improving Clinical Outcome Predictions Using Convolution over Medical Entities with Multimodal Learning

Usage

1. Clone the code to local.

https://github.com/tanlab/ConvolutionMedicalNer.git
cd ConvolutionMedicalNer

2. Run MIMIC-Extract Pipeline as explained in https://github.com/MLforHealth/MIMIC_Extract.

3. Copy the output file of MIMIC-Extract Pipeline named all_hourly_data.h5 to data folder.

4. Run 01-Extract-Timseries-Features.ipnyb to extract first 24 hours timeseries features from MIMIC-Extract raw data.

5. Copy the ADMISSIONS.csv, NOTEEVENTS.csv, ICUSTAYS.csv files into data folder.

6. Run 02-Select-SubClinicalNotes.ipynb to select subnotes based on criteria from all MIMIC-III Notes.

7. Run 03-Prprocess-Clinical-Notes.ipnyb to prepocessing notes.

8. Run 04-Apply-med7-on-Clinical-Notes.ipynb to extract medical entities.

9. Download pretrained embeddings into embeddings folder via link in given References section.

10. Run 05-Represent-Entities-With-Different-Embeddings.ipynb to convert medical entities into word representations.

11. Run 06-Create-Timeseries-Data.ipynb to prepare the timeseries data to fed through GRU / LSTM.

12.Run 07-Timeseries-Baseline.ipynb to run timeseries baseline model to predict 4 different clinical tasks.

12.Run 08-Multimodal-Baseline.ipynb to run multimodal baseline to predict 4 different clinical tasks.

12.Run 09-Proposed-Model.ipynb to run proposed model to predict 4 different clinical tasks.

References

Download the MIMIC-III dataset via https://mimic.physionet.org/

MIMIC-Extract implementation: https://github.com/MLforHealth/MIMIC_Extract

med7 implementation: https://github.com/kormilitzin/med7

Download Pre-trained Word2Vec & FastText embeddings: https://github.com/kexinhuang12345/clinicalBERT

Preprocessing Script: https://github.com/kaggarwal/ClinicalNotesICU