Abstract: We describe the DeepPhe system for extracting detailed phenotype information from Electronic Medical Records for oncology patients. Our goal is to fully automate the extraction of key clinical variables (e.g. TNM, stage, size, location, receptor status, procedures, medications) from clinical text and relate them to cancers or individual tumors, as well as specific events (e.g. recurrence, metastasis). Such phenotype information enables understanding the effects of genetic, epigenetic, and other factors on tumor behavior and responsiveness.

Learning Objective 1: Describe the DeepPhe system for deep phenotype extraction for cancer

Learning Objective 2: Present challenges associated with the task of automatic deep phenotype extraction for cancer


Guergana Savova (Presenter)
Boston Childrens Hospital

Eugene Tseytlin, University of Pittsburgh
Sean Finan, Boston Childrens Hospital
Melissa Castine, University of Pittsburgh
Timothy Miller, Boston Childrens Hospital
Olga Medvedeva, University of Pittsburgh
David Harris, Boston Childrens Hospital
Harry Hochheiser, University of Pittsburgh
Chen Lin, Boston Childrens Hospital
Girish Chavan, University of Pittsburgh
Rebecca Jacobson, University of Pittsburgh

Presentation Materials: