Computational tools for automatic ROI extraction in pathology
Principal Investigator
Name
Joëlle Barral
Degrees
PhD
Institution
Google Life Sciences
Position Title
Senior Hardware Engineer
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-171
Initial CDAS Request Approval
Oct 23, 2015
Title
Computational tools for automatic ROI extraction in pathology
Summary
Digital scanning is expensive and time consuming because the whole slide has to be scanned at the highest resolution that might be later needed. Pathologists, however, typically rely on a few regions of interest (ROIs) to perform their diagnoses, even if they spend most of their time reviewing slides (or portions of slides) with benign tissue. If tumor-specific ROIs can be extracted from slides scanned at low resolution, then high-resolution scanning can be limited to these regions and the pathologist’s time can be spent most effectively. In this project we propose to develop and validate computational tools for automated, reliable extraction of ROIs in H&E slides from the NLST data.
Aims
1. Develop algorithms to automatically extract tumor-specific vs. benign ROIs from digital slides
2. Manually annotate NLST data
3. Evaluate algorithms on NLST data