Increasing the Accuracy of Mammogram Interpretation.
Abstract
The project's aim is to refine and evaluate clinically a computer-based system to provide decision support to the radiologist in interpreting mammograms, and generate automatically a standardized report of the mammogram findings to the referring physician. In Year 3, representative case sets of mammograms and clinical data, with pathology or follow-up, were finally assembled at the Brigham and Women's Hospital (referral setting) and at Harvard Pilgrim Health Care (a screening setting). Five BWH radiologists read their 200 cases to provide scale values for each of a checklist of perceptual features, in order to train a set of statistical prediction rules. Five HPHC readers have performed a baseline reading of their 150 case, as a comparison for the aided readings to be made in Year 4. The automated report writer was brought to a stage suitable for evaluation by a focus group of referring physicians in Year 4.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1997
- Accession Number
- ADA346062
Entities
People
- John A. Swets
Organizations
- BBN Technologies