annotation.JPG

Collaborative Research Project 3

Principal investigators: Adam Flanders (Jefferson University), Carol Wu (MD Anderson Cancer Center) and George Shih (Cornell University)

Helper AI for annotation (HAIFA).

Updated January 20, 2023

Research collaboration between Jefferson University, MD Anderson Cancer Center and Cornell University to create standardized methods for labeling and annotation of COVID-19 related clinical imaging studies and to develop crowdsourcing best practices to generate appropriate labels/markups to support internal and external MIDRC research as well as MIDRC Data Science Grand Challenges. Activities include recruitment and training of experts, image label adjudication, methods to assess accuracy of experts, and data pre-processing.

Current plans include

  • Use the annotation infrastructure created in TDP 1c to accelerate the annotation process of multi-institutional data hosted in the MIDRC repository. 

  • Annotate 6,000 chest x-ray exams from the MIDRC data repository for COVID-19 pneumonia using a HAIFA approach

  • Development of an optimal HAIFA process for future semi-supervised annotation efforts on other MIDRC related research projects and data

  • Open source of all Helper AI For Annotation (HAIFA) models 

Previous

CRP 2: Algorithmic approaches for improving health equity
Next

CRP 4: Deep Learning of Echocardiography for Detection of Myocarditis Cardiac Injury ...