Effect of the Number of Images on the Normalized Noise Power Spectrum (NNPS) in Digital Mammography Detectors

Abstract

The noise power spectrum (NPS) characterizes the noise properties of imaging systems in the frequency domain. In literature, a normalized NPS (NNPS) is given regardless of the linearization method. NNPS is typically measured from images acquired from the digital detector without the use of any test apparatus. However, there is no clear information about the number of images in the current standards. Furthermore, the number of images utilized in such assessments is either unspecified or varies across studies. This study aims to evaluate the influence of the number of images on NNPS calculations. Average NNPS values were computed using one, three, five, and six detector images acquired from two Fujifilm Amulet Innovality full-field digital mammography systems at selected spatial frequencies. The reference air kerma for the measurements ranged between 90 and 95µGy, and all results were computed in the radial direction. The x-ray spectrum utilized was 29kV with 2 mm of additional aluminum filtration. Statistical evaluation of the NNPS results was estimated using standard deviation and coefficient of variation (cov). For the first system, NNPS values at spatial frequencies of 0.5, 2.0, and 5.0mm-1 were 2.55x10-06 ± 7.33x10-08 (cov=2.87%), 2.43x10-06±4.99 x10-08 (cov=2.06 %) and 2.26x10-06± 2.63x10-08 (cov= 1.16 %), respectively. Corresponding values for the second system were 2.63 x10-06 ± 3.74x10-08 (cov = 1.42 %), 2.3 x10-06 ± 4.76x10-08 (cov = 2.07 %) and 2.17x10-06 ± 1.73x10-08 (cov=0.80 %), respectively. The findings indicate that the number of images used does not significantly impact the NNPS results particularly at higher spatial frequencies