Practical dose point-based methods to characterize dose distribution in a stationary elliptical body phantom for a cone-beam C-arm CT system

Abstract

Purpose: To propose new dose point measurement-based metrics to characterize the dose distributions and the mean dose from a single partial rotation of an automatic exposure control-enabled, C-arm-based, wide cone angle computed tomography system over a stationary, large, body-shaped phantom.

Methods: A small 0.6 cm(3) ion chamber (IC) was used to measure the radiation dose in an elliptical body-shaped phantom made of tissue-equivalent material. The IC was placed at 23 well-distributed holes in the central and peripheral regions of the phantom and dose was recorded for six acquisition protocols with different combinations of minimum kVp (109 and 125 kVp) and z-collimator aperture (full: 22.2 cm; medium: 14.0 cm; small: 8.4 cm). Monte Carlo (MC) simulations were carried out to generate complete 2D dose distributions in the central plane (z = 0). The MC model was validated at the 23 dose points against IC experimental data. The planar dose distributions were then estimated using subsets of the point dose measurements using two proposed methods: (1) the proximity-based weighting method (method 1) and (2) the dose point surface fitting method (method 2). Twenty-eight different dose point distributions with six different point number cases (4, 5, 6, 7, 14, and 23 dose points) were evaluated to determine the optimal number of dose points and their placement in the phantom. The performances of the methods were determined by comparing their results with those of the validated MC simulations. The performances of the methods in the presence of measurement uncertainties were evaluated.

Results: The 5-, 6-, and 7-point cases had differences below 2%, ranging from 1.0% to 1.7% for both methods, which is a performance comparable to that of the methods with a relatively large number of points, i.e., the 14- and 23-point cases. However, with the 4-point case, the performances of the two methods decreased sharply. Among the 4-, 5-, 6-, and 7-point cases, the 7-point case (1.0% [±0.6%] difference) and the 6-point case (0.7% [±0.6%] difference) performed best for method 1 and method 2, respectively. Moreover, method 2 demonstrated high-fidelity surface reconstruction with as few as 5 points, showing pixelwise absolute differences of 3.80 mGy (±0.32 mGy). Although the performance was shown to be sensitive to the phantom displacement from the isocenter, the performance changed by less than 2% for shifts up to 2 cm in the x- and y-axes in the central phantom plane.

Conclusions: With as few as five points, method 1 and method 2 were able to compute the mean dose with reasonable accuracy, demonstrating differences of 1.7% (±1.2%) and 1.3% (±1.0%), respectively. A larger number of points do not necessarily guarantee better performance of the methods; optimal choice of point placement is necessary. The performance of the methods is sensitive to the alignment of the center of the body phantom relative to the isocenter. In body applications where dose distributions are important, method 2 is a better choice than method 1, as it reconstructs the dose surface with high fidelity, using as few as five points.

FIG. 1.

Design of an elliptical-shaped body phantom with tissue-equivalent material. Dose measurement inserts were placed at the 23 distributed points as shown in the axial view. Rods of equivalent material were placed in all holes that were not being used for dose measurement.

FIG. 2.

Steps for region-weighted sum method. For the proposed mean dose metrics, the 23 dose points were grouped into four regions (C₁, C₂, P₁, and P₂), as shown in (a). ${\overline{D}}_{C_1}$ represents the average dose of the dose points in region C₁ and (b) shows each dose point’s representative voxels based on each point’s closeness to the voxels. Then, the voxels of the dose points in the same region were grouped as shown in (c).

FIG. 3.

Steps for dose point surface fitting method. Six dose points from four regions were well distributed, as shown in (a). (b) shows the dose points and their mirrored dose points across the y-axis. Using the dose points, including the mirrored dose points as a control point, surface fitting over the body phantom was conducted, as shown in (c).

FIG. 4.

Tested dose point distributions. Six different cases of dose point numbers were tested, and some of the cases had different dose point distributions. The last distribution, “CTDI case,” shows the dose point measurement locations for CTDI.

FIG. 5.

Built-in AEC system responses to the requested dose settings (acquisitions 1–6). The labels, e.g., “109 kVp, full,” refer to (1) a minimum value of the kVp allowed and (2) the size of the collimator used for the scan.

FIG. 6.

Dose distributions delivered to the central plane of the body phantom. The delivered dose was simulated for acquisitions 1–6, shown in [(a1)–(b3)].

FIG. 7.

Comparison of the performances of the proposed methods as a function of the number of dose points used (4, 5, 6, 7, 14, and 23). The differences of CTDI_w(1/3, 2/3) and CTDI_w(1/2, 1/2) were 8.7% (±2.8%) and 6.8% (±1.0%), respectively.

FIG. 8.

Horizontal and vertical line profiles of the two methods extracted along the x-axis and the y-axis of the 2D dose distribution, respectively. “MC” is the Monte Carlo simulation, and “dose points” show the location of the control points for method 1 and method 2.

FIG. 9.

Analysis of the sensitivity of the performances of the methods as a function of the phantom displacement from the isocenter. The performances of the methods were computed using the difference metric in Eq. (6) after shifting the body phantom by 1, 2, and 3 cm along the x-axis.