(a) Flow chart for the LTI deconvolution algorithm and (b) flow chart for the NLCSC algorithm. An important modification made for the NLCSC algorithm is changing the state variable to the estimated stored charge.
Flowchart of the NLCSC calibration algorithm.
Stored charge estimates Q n for (a) the longest time constant and (b) the second longest time constant charge traps. Data points were fit with a fourth-order polynomial.
Exposure-dependent lag rates for (a) the longest time constant and (b) the second longest time constant charge traps. Data points were fit with an exponential function. The form of the function ensures that a 2,n (0) = 0.
The uncorrected and corrected RSRF found during the global search for the optimal a 2,n (x) at x = 27%.
Uncorrected and corrected FSRFs for step-response data at several different exposures, (a) LTI corrections derived from a 27% step-response, (b) NLCSC correction which takes into account the exposure dependence, and (c) the intensity weighted only non-LTI correction.
1 × 1 × 5 mm3 reconstructions of a pelvic phantom (42 × 26 cm2) with (a) no lag correction, (b) best LTI correction, (c) worst LTI correction, (d) intensity weighted lag correction, and (e) NLCSC correction. Window, level = 200, −25 HU.
0.5 × 0.5 × 5 mm3 reconstructions of a pelvic phantom (20 × 16 cm2) with (a) no lag correction, (b) best LTI correction, (c) worst LTI correction, and (d) NLCSC correction. Window, level = 100, −25 HU.
Percent of a-Si FP saturation signal for different exposure intensities at 125 kVp with 0.5 mm Ti.
LTI IRF parameters from calibration FSRF data at an exposure level of 27%. The base lag rates a 1, n have units of frames −1 and the lag coefficients b n have units of (detector counts with lag/detector counts without lag).
Summary of 1st and 50th frame residual lag for uncorrected, LTI, and NLCSC algorithms.
Summary of ROI reconstruction errors for different lag correction algorithms.
Article metrics loading...
Full text loading...