The Department of Biomedical Informatics and Data Science (DBIDS) has received set-aside funds to collaborate with Boston Children’s Hospital of Harvard Medical School through August 2023 for cancer research, as part of the U24 Project “Cancer Deep Phenotype Extraction from Electronic Medical Records.”  

John Osborne, Ph.D., Assistant Professor at DBIDS and the O’Neal Cancer Control and Population Science Program, will oversee the installation of a Deep Cancer Phenotyping (DeepPhe) system, in collaboration with Dr. Gaurav Goyal, MD, Assistant Professor and oncologist at the O’Neal Cancer Center.

DeepPhe is designed to extract cancer information from clinical text. As part of this project, DeepPhe will be extended to detect cancer recurrence and reportability status for all reportable cancer types and integrated with UAB’s Phenotype Detection and Registry System (PheDRS).

Dr. Gaurav Goyal, MD, Oncologist, O'Neal Cancer Center  

PheDRS identifies clinical phenotypes of interest using machine learning, including Natural Language Processing (NLP) of UAB clinical text and will replace the existing Cancer Registry Control Panel (CRCP) case-finding tool.

Currently, the existing CRCP presents candidate reportable cancer cases to UAB’s Certified Tumor Registrar (CTR) through a web interface, where it validates or rejects cases. This project aims to exploit DeepPhe’s extraction functionality, to reduce CTR abstraction burden, and to extend DeepPhe to identify cancer recurrence which is poorly represented in Electronic Health Records.

It will expand prior Natural Language Processing (NLP) and machine learning work on reportable case detection to codify cancer and cancer recurrence vocabularies that will be incorporated into the DeepPhe Ontology.

Dr. Osborne envisions that the successful implementation of this project will allow O’Neal Cancer researchers to reduce chart review and abstraction of cancer phenotypes (including recurrence) to facilitate clinical trial patient recruitment and data abstraction.

“The need for this project was brought about due to the continual increase in the number and complexity of cancer cases, due to an aging population and expansion of complex treatment types which has not been met by a concomitant expansion in the number of CTRs,” Dr. Osborne said. “This situation has placed greater pressure on cancer registries to scale up their manual abstraction processes using more advanced software.”