Keywords:
Artificial Intelligence, Radiation physics, CT, CT-Quantitative, Neural networks, Perception image, Physics, Screening, Cancer, Image verification, Quality assurance
Authors:
K. Boedeker, D. Shin, N. Akino, T. goto, K. Haioun, Y. Hamada
DOI:
10.26044/ecr2023/C-15391
Results
Low Contrast Detectability The LCD for the Silver filter and DLR was equivalent to that of the baseline LCS protocol using a standard filter with HIR at the same dose (1.2mGy for standard beam filter with HIR/1.1mGy for Silver beam filter with DLR). The d’ at 1.2mGy/1.1mGy ranged from 0.5-2.5, depending on object size. For the other dose levels examined for Silver filter with DLR, LCD performance trended with dose. Results are summarized in Figure 1.
Noise and Noise Texture
The noise magnitude was reduced for all dose levels with the Silver filter and DLR vs the baseline standard filter protocol with HIR, as shown in Table 4. At same dose, the Silver Filter with DLR reduced noise by 55.4% for Body Sharp reconstruction and 67.8% for high resolution lung reconstruction.
The peak frequency of the noise power spectrum was shifted by the DLR to the lower frequencies, as shown in Figures 2-3.
Contrast to Noise Ratio
The CNR improved at same dose by 133% for Body Sharp and 220% for Lung reconstruction, relative to baseline, as shown in Table 5. Note the CT number of the test object (Acrylic) increased by 7HU with the Silver filter, due to the energy difference of the beam relative to the standard filter.
A clinical example of the CNR properties is provided in Figure 4, which shows a small lung nodule well-demonstrated in the upper right lung using a low dose, 0.9mGy, chest CT combining a beam-shaping Silver filter with DLR.
Spatial Resolution
The High Contrast MTF showed similar or improved resolution for the Silver filter with DLR compared to baseline. The Low Contrast MTF was similar for the Body reconstruction for the Silver filter with DLR compared to baseline; for the Lung reconstruction the limiting resolution was similar, but had lower mid-frequencies values for the Lung reconstruction. The MTFs are given in Figures 5-8.