Repo for Deep learning approach to Patient Case Similarity
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.
- Nachiket Naganure
- Ashwin Nayak