抄録
We have been developing an automated image quality evaluation method (AIQEM) for quantitative evaluations of specks on the American College of Radiology (ACR) 156 mammographic phantom images. The AIQEM consists of the following two main steps:(a) Extracting speck shadows using a density gradient vector analysis and (b) analyzing a speck signal to noise ratio (SNR) and a size of a speck shadow to detect it as visible or not. In this paper, we describe experimental results of subjective evaluations by nine Japanese experienced radiologic technologists and the AIQEM on forty-two phantom images, and the AIQEM performance and reproductively. As initial results, the AIQEM sensitivities were 98% for 1 facility database and 100% for 21 facilities database for the 3rd group of specks, and 91% for 1 facility database and 95% for 21 facilities database for the 4th group of specks. The AIQEM reproductively for the 3rd group of specks was sufficient for the image quality evaluation. It was concluded that the AIQEM was an effective method for a quantitively image quality evaluation on the ACR 156 mammographic phantom.
