Date of this Version

11-25-1994

Keywords

computer simulation models, imaging, ROC, training

Abstract

Research in the processing, compression, transmission, and interpretation of digital radiographic images require the testing and evaluation of a wide variety of images, varying both in format and in spatial resolution. If receiver operating characteristic (ROC) analysis or a related method is used to evaluate the performance of observers using novel vs. conventional displays, large numbers of test images containing known abnormalities are required. This report describes a convenient, inexpensive, and reproducible source of test images, having any desired resolution and containing precisely defined abnormalities of unlimited subtlety. The images are generated by computing x-ray transmission through mathematically defined, three dimensional masses according to Beer's Law. A procedure is presented for generating computer simulated chest radiographs and mammograms, which can contain various classes of abnormalities, including tumors, infiltrates, cavities, pleural effusions, cardiac chamber enlargement, and soft tissue calcifications. Test images can be created from simple computational models of superimposed spherical densities. The approach provides a flexible, inexpensive, easy-to-use research tool for investigators exploring digital techniques in diagnostic radiology. Such simulation software may also be of benefit as a training tool, when employed to generate numerous test images containing subtle abnormalities for programmed instruction and testing.

Share

COinS