An ALTERNATIVE APPROACH TO CDMAM ANALYSIS USING A NEW PERCEPTUAL ALGORITHM: MS-SSIM^⋆

Abstract

Contrast detail phantoms for image Quality Control in diagnostic radiology has been proved as a powerful tool assuming a correct phantom design and a clear scoring methodology. For mammography, CDMAM test object is widely accepted, however, results are highly dependent from observer variability and also a great time effort is required for the evaluation task. Automatic readout software is available, named CDCOM, but still some limitations are reported in scientific literature. A new alternative approach based on the use of the perceptual visual algorithm MS-SSIM^⋆, oriented to the recognition threshold of different image representations is now presented. Our software tool has been implemented on the Java platform. Some initial results presented here show a promising alternative to CDCOM software. The correlation obtained between our program and the human observers is better than 0.998 and the range of useful correlation comprises diameters from 0.10 mm to 2.0 mm of the gold disks inside the CDMAM. These results remarkably improve some restrictions of CDCOM, even with the smoothing and psychometric curve fitting technique suggested by CDCOM authors to establish correlations between its contrast detail detection algorithm and human observers.