This repository is the home of the public source code releases of the DeepPhe Cancer Phenotype Extraction System. For an installation GUI of the DeepPhe system, please download our Installation Tool for Windows or Linux.
DeepPhe uses combines natural language processing built on the Apache cTAKES clinical NLP framework with an extensive domain information model to:
- extract information from plaintext documents
- summarize information for Cancers and Tumors across multiple documents
- visualize results at patient and cohort levels using our DeepPhe Viz tool.
The system has been tested using documents from five cancer domains:
- Breast Cancer
- Ovarian Cancer
- Malignant Melanoma
- Prostate Cancer
- Childhood brain cancer
There are two different products associated with DeepPhe:
- DeepPhe-XN: a full suite of tools, designed to support cohort discovery for cancer clinical research. Instructions for our installation tool are available on our Wiki for Windows, MacOS, or Linux
- DeepPhe-CR: separate web-service based software similar in some ways to DeepPhe, but designed to support cancer registries. It uses a different ontology, extracts only top-confidence attributes pertinent to Cancer-Registries, and processes single documents instead of a patient's entire record. Installation instructions are available in the DeepPhe-CR release repository.
-
DeepPhe is provided under an Apache 2 software license.
-
DeepPhe includes portions of the HemOnc.org ontology.
-
DeepPhe includes portions of the NCI Thesaurus.
-
DeepPhe includes portions of the NLM UMLS.
-
DeepPhe uses the Apache cTAKES clinical NLP framework.
-
DeepPhe uses the Neo4j graph database platform.
If you'd like to drop us a note, or have technical questions, please post on the issue tracker.
Metrics on downloads and usage could help us with funding future enhancements.