PURPOSE: To quantitatively demonstrate inconsistencies in microcalcification specks in RMI 156phantoms approved by the American College of Radiology (ACR) for use with their mammography accreditation program.
MATERIALS AND METHODS: Signal intensities of the largest three specks (Cl,540, μm; C2,400pm; C3,320 μm) were measured on images of 14 RMI 156 (version 4) phantoms. Three images were analyzed per phantom: a packaged image of the wax plate and two phantom images obtained at an optical density of 1.2 to 1.3. Phantom images were digitized using a VXR 12 scanner at a resolution of 85 μm and 8 bit/pixel. Speck signals were measured as the average pixel values using the NIH Image program (version 1.61).
RESULTS: Signal intensities of specks measured on phantom images were consistent with those measured on wax plate images, with an average correlation coefficient of 0.898 for the Cl groups and of 0.794 for the C3 groups. Overlap of speck sizes was found between different groups. Speck signals of some inserts showed considerable inconsistencies within a group. In some phantoms, signal intensities of specks in different groups overlapped to each other, and Cl and C2 groups were not clearly separated. In all 14 phantoms, speck signal intensities of the C3 group had a between phantom variance which was larger than the within phantom variance.
CONCLUSION: Microcalcification specks in RMI 156 phantoms of the same version were found to be substantially inconsistent in terms of size and intrinsic signal intensity. As lack of uniformity in specks within a phantom and between phantoms is not appropriate as a standard of image quality control, further quality control is apparently needed in manufacturing RMI 156 phantoms and also in approving specific phantoms.
抄録全体を表示