Author: Ross I. Berbeco, Vera Birrer, Raphael Bruegger, Pablo Corral Arroyo, Roshanak Etemadpour, Dianne M. Ferguson, Rony Fueglistaller, Thomas C. Harris, Yue-Houng Hu, Matthew W. Jacobson, Mathias Lehmann, Marios Myronakis 👨🔬
Affiliation: Medical Physics Department, Medical School, University of Thessaly, Brigham and Women's Hospital, Harvard Medial School, Dana-Farber Cancer Institute, Department of Radiation Oncology, Dana Farber/Brigham and Women's Cancer Center, Department of Radiation Oncology, Brigham and Women’s Hospital, Dana-Farber Cancer Institute, Harvard Medical School, Brigham and Women's Hospital, Varian Imaging Laboratory 🌍
Purpose:
Dual-layer kV imager (DLI) architecture enables the capability for single-shot dual-energy (DE) intrafraction imaging. As noise from individual DLI layers are primarily uncorrelated, simple SNR and CNR analysis of DE imaging suggest poorer performance than single-energy (SE) alternatives, while not fully capturing the outsize benefit. We quantify the effect of thoracic anatomy as a source of structural-noise (SN), which may impede detection and tracking of targets-of-interest, as an analytical framework for quantifying the benefit of employing DE to improve lesion conspicuity.
Methods: Clinical, anonymized lung images were gathered from an experimental DLI installed on a TrueBeam linac. Images acquired at gantry angle 0o (i.e. DLI at 90o) were analyzed. Log-subtraction between the top layer (TL) and bottom layer (BL) was performed, with several weighting factors (estimated based on modeled x-ray spectra and attenuation coefficients). A 512x512 region-of-interest (ROI) was selected within the lung tissue and separated into half-overlapping sub-ROIs. 2D detrending was applied to each sub-ROI to eliminate low-frequency fluctuations and the effect of scatter and the Fourier transform was calculated. The mean over all sub-ROIs were calculated to yield the noise-power spectrum.
Results: Theoretical calculation of weighting factors yields approximately 0.79 and 0.93 to yield bone-only and soft-tissue only images. Qualitative review of images agrees well with theoretical calculations. Further, low-frequency noise (i.e. noise arising from anatomical features) is minimized with bone-only images within the lung tissue.
Conclusion: Theoretical calculation of dual-energy weighting factors agreed well with empirical measurements. Within the lung, structural noise is reduced by equalizing variations due to soft tissue or bony anatomy. Qualitatively, lesion conspicuity is maximized by implementing soft-tissue only images, implying that in-spite of reduced structural noise in bone-only images, reduction in soft-tissue contrast is severe enough to limit visualization.