Overview of study methods.
Simulated lesion examples. This includes a malignant mass (left), benign mass (center), and a subtle microcalcification distribution (right).
(a) Normalized noise power spectrum of the detector noise for an average mammogram in our database. The relative inherent quantum noise magnitude, or the integral of the NNPS, was , , and for , , and , respectively. (b) Example mammographic regions at each of the noise levels with on the left, in the center, and on the right.
(a). Three resolution levels evaluated in this study. These correspond to the resolution of an indirect digital detector convolved with three different display devices. The geometric sharpness, or integral of the , was 2.14, 0.92, and 0.90 for , , and , respectively. (b) Example mammographic regions at each of the three resolution levels with on the left, in the center, and on the right.
Distribution of images in each resolution and noise category. The backgrounds were paired, as noted by the split lines with gray circles, to increase statistical power for certain tasks. The category of was not evaluated in this experiment to reduce the duration of the human observer experiment.
Overall classification accuracy (a) and lesion detection accuracy (b) for the average observer at different resolution and noise levels.
Task performance for average observer at different resolution and noise levels.
(a) Effect of resolution on task performance using VDM observer. (b) Effect of noise on task performance using VDM observer.
(a) Effect of resolution on task performance using NPWE model. (b) Effect of noise on task performance using NPWE model.
Microcalcification detection accuracy as a function of dose. The solid line illustrates the data fit by a square root function.
Overview of display properties. The CRT display was further modified by defocusing its electron gun in order to produce both the standard resolution and degraded resolution CRT evaluated in this experiment. The first six rows reflect manufacturer specifications, while the luminance values in the last two rows were measured in our laboratories (Refs. 2 and 27). All monitors were calibrated to the DICOM and AAPM TG18 standards.
Article metrics loading...
Full text loading...