An official website of the United States government
The Framework for Exploring Scalable Computational Oncology (FrESCO) is a modular deep learning natural language processing (NLP) library for extracting structured information from clinical text documents and classifying information to a given data standards.
The FrESCO framework is a computational science tool that enables the automatic extraction of information from dense clinical reports. FrESCO’s modular deep learning natural language processing (NLP) library and associated tools provide the foundation for downstream research tasks and prediction algorithms.
Automatic extraction of information from electronic health records and conversion to a common data model is an important task for precision public health and precision medicine. The Framework for Exploring Scalable Computational Oncology (FrESCO) is a modular deep learning natural language processing (NLP) library for extracting structured information from clinical text documents and classifying information to a given data standards.
The application of automating this information extraction process from clinical text documents has the potential to improve the quality of the data by consistently extracting information and to improve the quality of patient outcomes by reducing the time to assimilate new data and enabling time-sensitive applications such as precision medicine and precision public health.
FrESCO, provides a modular deep-learning natural processing library for extracting structured information from clinical electronic records. While the usage to date for FrESCO has been extraction of structured information from pathology reports, FrESCO can in theory be trained with any clinical text document for a desired endpoint.
FrESCO and its components are available on GitHub.
Users can train report level and case level models across a range of information extraction tasks. This training can evoke the deep abstaining classifier (DAC), which incorporates abstention at the task level or the report level.