neuradicon is a set of models for extracting clinical information from neuroradiological reports, and deriving clinically-meaning representations of the text of the report as a whole. It currently provides models for the following tasks:
- Artefact detection
- Section segmentation
- Clinical entity extraction
- Radiological signal detection
- Report type classification
- Clinical text representation
The neuradicon project depends on two other reporting projects, neurollm and neurocluster. These must be installed before installing neuradicon. Their repositories can be found in the reporting project group.
To install neuradicon, clone this repository and run pip from the repository root.
pip install -e .Alternatively, an easier means of using the software is to pull the accompanying docker image, containing both models and the environment required to run neuradicon.
You can find an example script for using neuradicon in the examples directory.