OpenMed extracts biomedical entities and de-identifies clinical text through local execution or on-premise deployment. It supports running with a local-only data boundary when using local model storage, avoiding vendor lock-in.
Project overview
The project addresses the need to handle Protected Health Information (PHI) by providing local biomedical entity extraction and de-identification capabilities designed to operate without exposing data to cloud APIs.
Project type
Model Runtime
Deployment
Refer to project documentation
License
Apache-2.0
Best for
Developers and enterprise teams who need to process clinical text and detect medical entities or PII locally without cloud APIs.
Key capabilities
Extracts biomedical entities such as diseases, drugs, and anatomical structures from clinical text.
Packages ONNX token-classification exports for in-browser inference through Transformers.js and WebGPU.
Extracts or de-identifies PII across many documents using BatchProcessor or incremental streaming helpers.
Limitations and risks
Functionality requires downloading external model weights to operate.
Users should be aware of the potential for redaction bypass or PHI/PII leaks during processing.
GPU availability is optional for running the system.
Getting started
Installation difficulty is rated easy via standard package managers (pip, npm, Swift Package Manager) or Docker build commands and pre-configured API containers. The first success path involves installing the openmed package and running the analyze_text or deidentify function.