Pipeline Description - DeepPhe/DeepPhe-Release GitHub Wiki

DeepPhe-XN Pipeline Description

The DeepPhe Pipeline is configured using Apache cTAKES Piper Files. Piper Files contain simple commands that assemble an NLP pipeline, but knowledge of existing DeepPhe and cTAKES components is required. Piper Files also allow the user to add their own custom components to the DeepPhe pipeline.

At this time, the per-document NLP pipeline performs the following main tasks:

• Medical Section Identification
• Text Paragraph and Sentence Splitting
• Text Term Identification
• Term Code Normalization
• Term Negation, Uncertainty
• Term Medical Relations Identification

The Patient Summarization step performs the following main tasks:

• Term Coreference Resolution
• Neoplasm Concept Identification
• Collation of Neoplasm Concepts into one or more Cancer Containers
• Collation of Neoplasm Concepts into one or more Tumor Containers
• Cancer and Tumor Connection
• Cancer and Tumor Attribute Assignment