Characterisation of noise and sharpness of images from four digital breast tomosynthesis systems for simulation of images for virtual clinical trials

Abstract

In-depth evaluation of the noise and sharpness characteristics of FujiFilm Innovality, GE SenoClaire, Hologic Selenia Dimensions and Siemens Inspiration digital breast tomosynthesis (DBT) systems was performed with the intention of improving image simulation for virtual clinical trials. Noise power spectra (NPS) and modulation transfer function curves (MTF) were measured for planar modes and for the first and central projections for DBT modes. In DBT mode, the x-ray beam was blocked for the projections before the central projection in order to remove the influence of lag and ghosting from the previous images. A quadratic fit between the NPS and linearised pixel value gave the noise coefficients for planar and DBT imaging modes. The spatial frequencies corresponding to an MTF of 0.5 (MTF) were calculated from the MTF measurements made on the breast support and at 40 mm above the breast support. This was done for the first and the central projections. The percentage of signal carried over from the first projection to subsequent images (lag) was measured using a slit. The noise associated with lag was also evaluated. The DBT modes typically had lower electronic noise coefficients but higher structural noise coefficients compared to the respective planar mode MTF measured 40 mm above the table was between 6% and 47% lower for continuous scanning systems compared to 1% lower for step and shoot systems. For wide angle DBT, the MTF of the first projection was 18% (FujiFilm) and 28% (Siemens) lower than for the central projection. Lag in the second projection was 2.2%, 0.3%, 0.8% for the FujiFilm, GE and Hologic systems respectively. In all cases, the noise associated with lag was negligible. Current modelling frameworks for virtual clinical trials of breast DBT systems need to be adapted to account for signals from lag and variations in the MTF at wide angles.

Can Breast Compression Be Reduced in Digital Mammography and Breast Tomosynthesis?

“…The second is an effect called “super resolution” that is observed in tomosynthesis images as a consequence of the image of the object being shifted by subpixel increments of the detector element, which leads to improved visibility of microcalcifications in the reconstructed images [14]. We also know that an effect of blurring due to focal spot motion occurs in tomosynthesis projections that is not seen in full-field digital mammography [15]. However, although this finding is interesting, for this study we were concerned with the comparison in blurring due to motion of breast tissue between standard and reduced compression images during DBT, be they projections or reconstructions, not in comparing the observed blurring between projections and reconstructed images.…”

Technical Note: Noise models for virtual clinical trials of digital breast tomosynthesis

“…A single was estimated for each system model. Correlation kernel of the electronic noise : Throughout this paper, we assume that the electronic noise is not correlated, that is, . This assumption is based on measurements performed by other authors on both a‐Se and CsI(Tl) detectors. Gain map of the quantum noise λ : It is estimated using the ratio between the variance of the quantum noise and the linearized MPV, as described in Ref. .…”

“…• Correlation kernel of the electronic noise K e : Throughout this paper, we assume that the electronic noise is not correlated, that is, K e ¼ dð0Þ. This assumption is based on measurements performed by other authors on both a-Se 17 and CsI(Tl) 15 detectors.…”

Multi‐domain features for reducing false positives in automated detection of clustered microcalcifications in digital breast tomosynthesis

“…In DBT volume, a MC cluster may be split into different slices. Also, increased noise and blurring effect are associated with multiple low‐dose projection view acquisition and limited‐angle reconstruction . The large amount of DBT data also pose stress in image reading for MC detection.…”

“…Also, increased noise and blurring effect are associated with multiple lowdose projection view acquisition and limited-angle reconstruction. 9 The large amount of DBT data also pose stress in image reading for MC detection. As a second reader, computer-aided detection (CAD) of clustered MC has been shown helpful to assist radiologists in reading DBT images.…”