MEASUREMENT OF DIFFERENTIAL IMAGE QUALITY

点击翻译按钮获取中文摘要

DESCRIPTION (Adapted from Applicant's Abstract): The objective of this research is to develop a method for determining differences in diagnostic task performance between any two modalities A and B. The new technique is called the paired image (1) observer performance method. While the research will be done with x-ray images, the 1 method is expected to have much wider applicability (e.g., CT verses MRI). In the 1 method, the reader is shown a pair of images of the same subject, from each modality. The observer selects the image that is superior for the specified diagnostic task and assigns a rating (A>>B,A>B,A=B,A<B,A<<B). Analysis of this experiment yields the area under the Differential ROC curve (!ROC). After development of the method, it will be applied to phantom and clinical images under a variety of processings. The clinical tasks are detection of microcalcifications, detection of masses, classification of microcalcifications,and classification of masses. The applicants proposed to extend the method to other observer performance experiments that include localization information, namely, FROC and LROC,and assess the feasibility of applying it to CAD evaluation. In each case, they would measure the improvement in statistical power of the 1 method over the traditional ROC method. The applicants proposed to apply the method to evaluating the effect of several compression algorithms in digital mammography. The 1 method has the potential for detecting very subtle differences in image quality, much smaller than is detectable by the present ROC method. It will allow rapid optimization of imaging systems, without the need for expensive and often inclusive ROC studies. The compression algorithm evaluation is expected to be useful to imaging scientists and engineers who need to design more cost-effective digital mammography systems. This will greatly facilitate the adoption of digital methods in mammography and thereby lead to improved health care for women with breast cancer.