a Computational Observer for Objective Evaluation of Ultrasound Image Quality.

We present the design, development, and implementation of a totally objective, computational method for the determination of various image quality parameters for ultrasound B-scan images. This method evolved from the development of an image quality measurement tool, the Contrast-Detail (C/D) Phantom, developed by the author and others, and from Human Observer experiments, also conducted by the author, initially designed to use the C/D phantom for the evaluation of ultrasound images. It was found that human observer performance is not sufficiently consistent or accurate for objective evaluation of images or imaging systems, thus, the phantom evaluation method was further refined by the development of a computer -based, computational method for analysis of digitized ultrasound images of the C/D phantom. Our computer-based method, the Computational Observer, analyzes many regions of interest in the digitized images of the C/D phantom and computes detectability indices, based on the signal-to-noise characteristics of the regions of interest, from which Contrast-Detail plots are derived. We compare the results of the Computational Observer to those of human observers for the same set of digitized images, and find that the Computational Observer method meets the criteria of being accurate, objective, transportable, and relevant to human observer evaluations.