COREBEHRT

COREBEHRT

Virtual environment

For running the tests and pipelines, it is adviced to create a virtual environment, enable it, and install the requirements.

$ python -m venv .venv
$ source .venv/bin/activate
(.venv) $ pip install -r requirements.txt

Unittests

In Linux

Enable your virtual environment and run the unittests:

(.venv) $ python -m unittest

Pipeline

The pipeline can be run from the root directory by executing the following commands:

(.venv) $ python -m corebehrt.main.create_data
(.venv) $ python -m corebehrt.main.pretrain
(.venv) $ python -m corebehrt.main.create_outcomes
(.venv) $ python -m corebehrt.main.finetune_cv

1. Create Data

Creates tokenized features from the formatted data.

2. Pretrain

Pretrains the model on the tokenized features.

3. Create Outcomes

Creates the outcomes from the formatted data. Outcomes are stored as absolute positions.

4. Finetune

Finetunes the pretrained model on the outcomes.

Classes

FeatureCreator

and patient data

we create:

and include the following:

• background
• death event

Excluder

• incorrect ages
• nans
• normalize segments after exclusion
• short sequences

Results are saved in a table.

EHRTokenizer

Currently, still operates on sequences. Adds SEP and CLS tokens Create vocabulary based on pretrain_data

Azure

Use the submodule corebehrt.azure for running on Azure with SDK v2. See how-to.