Development and Evaluation of Sterographic Display for Lung Cancer Screening
Abstract
The main purpose of this project is to investigate the feasibility and efficacy of using a stereo display workstation for lung cancer screening on CT images. The tasks included in this project are development and evaluation of stereo image projection and display for chest CT images, observer performance evaluation for the stereo display, and stereo feature analysis and comparison to the conventionally used display methods for lung cancer detection. During the funding period, we have made progress in following tasks. 1. We have built stereo display workstation for chest CT images and investigated effects of several commonly used compositing methods for nodule representation and detection in stereo CT images. Among these methods, conventional maximum intensity projection (MIP) produced the highest image contrast, but gave ambiguities in local geometric detail and texture, whereas averaging compositing resulted in the lowest contrast, but preserved geometric details. Distance-weighted MIP partially recovered geometric information, which was lost in images composited by conventional MIP, therefore is the best compositing method for stereo display. 2. Consensus truth of the cases collected for this project has been done by three experienced radiologists. 3. A pilot observer performance study was conducted. Six radiologists participated the pilot observer performance study. The study has three display modes, conventional slice-by-slice mode, conventional MIP display mode and stereo display mode. The performance of lung nodule detection are examined and compared for the three modes with Free-response Receiver Operating Characteristic (FROC) statistic method. The results indicate that the stereo display achieved the best performance followed by the slice by-slice display, and the conventional MIP display gave the worst performance, although there is no statistically significant difference between the three display modes.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2008
- Accession Number
- ADA495768
Entities
People
- Carl R. Fuhrman
- David Gur
- Howard E. Rockett
- Walter F. Good
- Xiao H. Wang
Organizations
- University of Pittsburgh