PatientCaseSimilarityDeepLearning

Repo for Deep learning approach to Patient Case Similarity

Getting Started

We have developed a deep learning model to measure patient case similarity. The dataset used is the MIMIC-III provided by MIT Labs.

Software packages need to run the notebooks

Keras
Tensorflow-GPU
Pandas
Numpy
Gensim
Pickle

Due to enormous size of size of the prepocessed input files, we can't upload them on GitHub. Although you can view the results in the notebook

MIMIC Explore Edition Notebook - Shows the explores the basic files in MIMIC
MIMIC Explore Edition Notebook V2 - Starts with preprocessing of the NOTEEVENTS.csv
MIMIC Explore Edition Notebook V3- Preprocessing of the DIAGNOSES_ICD.csv
MIMIC Explore Edition Notebook V4 - Preprocessing to generate disease vector for each patient
ir-project, kernel(1,2) - These contain the gensim word embedding model and first test model builds.
FINAL INTEGRATED MODEL - Contains Patient Cohort selection and First iteration run/test run of our Model.
Final Notebook with all models and results - Contains all the models and test results.

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
FINAL INTEGRATED MODEL .ipynb		FINAL INTEGRATED MODEL .ipynb
Final Notebook with all models and results.ipynb		Final Notebook with all models and results.ipynb
MIMIC EXPLORE V4.ipynb		MIMIC EXPLORE V4.ipynb
MIMIC Explore Edition Notebook.ipynb		MIMIC Explore Edition Notebook.ipynb
MIMIC Explore V2.ipynb		MIMIC Explore V2.ipynb
MIMIC Explore V3.ipynb		MIMIC Explore V3.ipynb
README.md		README.md
ir-project.ipynb		ir-project.ipynb
kernel (1).ipynb		kernel (1).ipynb
kernel (6).ipynb		kernel (6).ipynb
kernel.ipynb		kernel.ipynb