Purpose: Volumetric breast density (VBD) measurement using dual-energy (DE) material decomposition is prone to patient motion. This can be mitigated by using a dual-layer detector which simultaneously acquires the DE images in a single exposure. However, the decomposed VBD map is still suffered from high image noise which introduces large variation in the VBD estimate. We aim to quantify the VBD using a dual-layer detector and reduce the uncertainty using the breast thickness as constraint for decomposition.
Methods: An in-silico experiment was performed using a virtual clinical trial software (VICTRE). The imaging system was simulated with a rhodium x-ray filter, a direct detector (a-Se) as the top layer (TL), and an indirect detector (CsI) as the bottom layer (BL). The breast attenuations measured from the TL and BL were calibrated for a wide range of VBD and breast thicknesses and were used as a look-up table for decomposition. A digital breast phantom was simulated for which the ground-truth thickness map of the breast and the fibroglandular tissue were determined using ray tracing and were used to derive the VBD. The projection images for TL and BL were simulated without scatter radiation. The decomposed breast thickness map was corrected to maintain uniform thickness in the center and allow smooth decrease in the periphery, which was applied as constraint to determine the VBD map.
Results: The decomposed breast thickness and VBD map were compared to the ground truth on a pixel-by-pixel basis. The average error in breast thickness is 1.01 mm (standard deviation = 0.70 mm), and the error in VBD is -0.89% (sd = 10.7%).
Conclusion: We developed a single-shot spectral mammography technique to quantify VBD using a dual-layer detector and a k-edge filter. Preliminary results using VICTRE simulation showed its potential to provide accurate estimate of breast thickness and consistent measurement for VBD.