Optimization of x-ray spectra in digital mammography through Monte Carlo simulations

Low-dose whole-spine imaging using slot-scan digital radiography: a phantom study

BACKGROUND

Slot-scan digital radiography (SSDR) is equipped with detachable scatter grids and a variable copper filter. In this study, this function was used to obtain parameters for low-dose imaging for whole-spine imaging.

METHODS

With the scatter grid removed and the beam-hardening (BH) filters (0.0, 0.1, 0.2, or 0.3 mm) inserted, the tube voltage (80, 90, 100, 110, or 120 kV) and the exposure time were adjusted to 20 different parameters that produce equivalent image quality. Slot-scan radiographs of an acrylic phantom were acquired with the set parameters, and the optimal parameters (four types) for each filter were determined using the figure of merit. For the four types of parameters obtained in the previous section, SSDR was performed on whole-spine phantoms by varying the tube current, and the parameter with the lowest radiation dose was determined by visual evaluation.

RESULTS

The parameters for each filter according to the FOM results were 90 kV, 400 mA, and 2.8 ms for 0.0 mm thickness; 100 kV, 400 mA, and 2.0 ms for 0.1 mm thickness; 100 kV, 400 mA, and 2.8 ms for 0.2 mm thickness; and 110 kV, 400 mA, and 2.2 ms for 0.3 mm thickness. Visual evaluation of the varying tube currents was performed using these four parameters when the BH filter thicknesses were 0.0, 0.1, 0.2, and 0.3 mm. The entrance surface dose was 59.44 µGy at 90 kV, 125 mA, and 2.8 ms; 57.39 µGy at 100 kV, 250 mA, and 2.0 ms; 46.89 µGy at 100 kV, 250 mA, and 2.8 ms; and 39.48 µGy at 110 kV, 250 mA, and 2.2 ms, indicating that the 0.3-mm BH filter was associated with the minimum dose.

CONCLUSION

Whole-spine SSDR could reduce the dose by 79% while maintaining the image quality.

CHARACTERISATION OF TLDS-200 AND OSLDS AT LOW X-RAY ENERGIES AND DETERMINATION OF EYE LENS DOSE, THYROID DOSE AND MEAN GLANDULAR DOSE DURING STANDARD MAMMOGRAPHY AND TOMOSYNTHESIS

The objective of this study was to characterise thermoluminescent (TLDs) and optically stimulated luminescent dosimeters (OSLDs) at low X-ray energies and estimate the eye lens (DL), thyroid (DT) and mean glandular (DG) doses received during Full-Field Digital Mammography (FFDM) and Digital Breast Tomosynthesis (DBT). The dosimeters were characterised in mammography energies. DL, DT and DG were estimated in FFDM and DBT mode taping dosimeters on the skin of the thyroid gland and on the left eye lens of an Alderson phantom. Dosimeters were also placed on the top of a NORMI PAS phantom simulating a compressed breast. The accuracy, precision and lower limit of detection (LLD) for TLDs and OSLDs were 5 and 8%, 6 and 3%, and 38 and 11 μSv, respectively. The linearity of the kerma response had an R2 > 0.99 and energy dependence was lower than 40%. DT ranged from 0.40 to 2.87 μGy for FFDM and 1.27 to 5.99 μGy for DBT. DG was between 0.50 and 1.27 mGy for FFDM and 1.07 and 1.60 mGy for DBT. DL was below the LLD. Dosimeters showed good performance. DG values were lower than those found in the literature, whereas DT value agreed with references. Differences between DG and DT determined with OSLDs and TLDs were lower than 10% and 200%.

Cascaded systems analysis of anatomic noise in digital mammography and dual-energy digital mammography

In x-ray based imaging of the breast, contrast between fibroglandular (Fg) tissue and adipose (Ad) tissue is a source of anatomic noise. The goal of this work was to validate by simulation and experiment a mathematical framework for modelling the Fg component of anatomic noise in digital mammograpy (DM) and dual-energy (DE) DM. Our mathematical framework unifies and generalizes existing approaches. We compared mathematical predictions directly with empirical measurements of the anatomic noise power spectrum of the CIRS BR3D structured breast phantom using two clinical mammography systems and four beam qualities. Our simulation and experimental results showed agreement with mathematical predictions. As a demonstration of utility, we used our mathematical framework in a theoretical spectral optimization of DM for the task of detecting breast masses. Our theoretical optimization showed that the optimal tube voltage for DM may be higher than that based on predictions that do not account for anatomic noise, in agreement with recent theoretical findings. Additionally, our theoretical optimization predicts that filtering tungsten-anode x-ray spectra with rhodium has little influence on lesion detectability, in contrast with previous findings. The mathematical methods validated in this work can be incorporated easily into cascaded systems analysis of breast imaging systems and will be useful when optimizating novel techniques for x-ray-based imaging of the breast.

Towards standardization of x-ray beam filters in digital mammography and digital breast tomosynthesis: Monte Carlo simulations and analytical modelling

In digital breast tomosynthesis and digital mammography, the x-ray beam filter material and thickness vary between systems. Replacing K-edge filters with Al was investigated with the intent to reduce exposure duration and to simplify system design. Tungsten target x-ray spectra were simulated with K-edge filters (50 µm Rh; 50 µm Ag) and Al filters of varying thickness. Monte Carlo simulations were conducted to quantify the x-ray scatter from various filters alone, scatter-to-primary ratio (SPR) with compressed breasts, and to determine the radiation dose to the breast. These data were used to analytically compute the signal-difference-to-noise ratio (SDNR) at unit (1 mGy) mean glandular dose (MGD) for W/Rh and W/Ag spectra. At SDNR matched between K-edge and Al filtered spectra, the reductions in exposure duration and MGD were quantified for three strategies: (i) fixed Al thickness and matched tube potential in kilovolts (kV); (ii) fixed Al thickness and varying the kV to match the half-value layer (HVL) between Al and K-edge filtered spectra; and, (iii) matched kV and varying the Al thickness to match the HVL between Al and K-edge filtered spectra. Monte Carlo simulations indicate that the SPR with and without the breast were not different between Al and K-edge filters. Modelling for fixed Al thickness (700 µm) and kV matched to K-edge filtered spectra, identical SDNR was achieved with 37-57% reduction in exposure duration and with 2-20% reduction in MGD, depending on breast thickness. Modelling for fixed Al thickness (700 µm) and HVL matched by increasing the kV over (0,4) range, identical SDNR was achieved with 62-65% decrease in exposure duration and with 2-24% reduction in MGD, depending on breast thickness. For kV and HVL matched to K-edge filtered spectra by varying Al filter thickness over (700, 880) µm range, identical SDNR was achieved with 23-56% reduction in exposure duration and 2-20% reduction in MGD, depending on breast thickness. These simulations indicate that increased fluence with Al filter of fixed or variable thickness substantially decreases exposure duration while providing for similar image quality with moderate reduction in MGD.