Evaluation of a Digital Telemammography System: A Model for a Regional System
Abstract
The research hypothesis being tested is that a telemammography system provides a mechanism for digitizing, transmitting, archiving, and displaying mammograms so that a trained radiologist utilizing grayscale monitors (2k x 2k x 8/12 bits) can interpret the images with an accuracy level sufficient for primary diagnosis. A series of retrospectively collected 200 normal mammograms and 200 abnormal mammograms containing a single lesion including masses, calcifications, and asymmetric densities were collected. The ROC study, which used 12 readers compares analog screen-film mammograms and the mammograms which are digitized at 50 microns spot size. The digitized mammograms were displayed on a 2k x 2k x 8/12 bit grayscale workstation using two monitors. An evaluation performance is being conducted using the metrics of throughput and throughput-to-cost-ratio. For all readers the results of the ROC analog images varied 10% from the 0.83 area under the ROC curve, which reflects proper case selection. The throughput analysis of the analog images showed that the technologist was the bottleneck (0.041) jobs/minute) and the throughput to cost ratio was 0.001203. While the ROC analysis results on the workstations are not very different the users found the display workstation graphics cumbersome. To be clinically useful, telemammography must provide user-friendly, rapid-throughput soft-copy interpretation of digitized screen-film mammograms.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1998
- Accession Number
- ADA367583
Entities
People
- Ellen S. De Paredes
Organizations
- University of Virginia