Optimizing 4DCT and BHCT Imaging for Ventilation Calculations

In this work, image noise was calculated as the standard deviation of HU in a centrally-placed circular region of interest (ROI) in axial phantom images acquired with each mAs value. All mAs and corresponding image noise values were fit to an exponential using Microsoft Excel (Microsoft, Inc., Redmond, WA) power-law curve-fitting to identify the relationship between image noise and mAs values with constant scan parameters listed previously for both static and 4DCT acquisitions. Equation 3 shows the general relationship between image noise and 4DCT mAs (mAs_4D) with constants A_4D and B_4D. Similarly, the relationship between BHCT image noise and mAs (mAs_BH) is given in Eq. 4 (with constants A_BH and B_BH). Based on the theoretical relationship between image noise and mAs (Eq. 1), exponential constants B_4D and B_BH are unitless and expected to have numerical values near 0.5. Since noise is expressed in units of HU, A_4D and A_BH have units of HU $\times \sqrt{mAs}$ . $Nois{e}_{4DCT}=A_{4D}\times {\left(mA{s}_{4D}\right)}^{-B_{4D}}\left[units=HU\right]$
$Nois{e}_{\mathit{BHCT}}=A_{BH}\times {\left(mA{s}_{BH}\right)}^{-B_{BH}}\left[units=HU\right]$
At UW-Madison, 4DCTs are currently used for CT-simulation and CT-ventilation calculation. With a low pitch of 0.09, 4DCT scans are time-consuming to acquire and have high associated dose; these aspects limited feasibility of acquiring several consecutive 4DCTs with unique image noise levels. To address this challenge, 4DCTs were acquired at two noise levels and BHCT scans were acquired at multiple intermediate noise levels. With a pitch of 1, BHCT scans require less dose and time than 4DCT acquisitions. Including BHCT imaging has the added benefit of expanding applicability of the study, since BHCT scans are commonly used for CT-ventilation calculation and the further reduced dose broadens potential clinical uses.
To ensure clinically relevant image noise levels were selected for BHCT imaging, mAs_BH values with image noise at and above current practice (mAs_4D = 100 mAs/rotation) were chosen. Appropriate mAs_BH values were determined by setting Eqs 3, 4 equal to discern mAs_BH with an equivalent noise level to mAs_4D, as listed in Eq. 5. For WMS imaging, 4DCTs were acquired with reduced (10 mAs) and standard of care (100 mAs) mAs values. BHCT images were acquired at intermediate noise levels equivalent to 4DCTs with 15, 20, 25, 30, 35, 40, 60, 70, 80 and 100 mAs according to Eq. 5. Tighter sampling was used at lower equivalent mAs values (15–40 mAs) since the low noise region is the steepest portion of the exponential curve relating image noise and mAs. $mA{s}_{BH}={\left(\frac{A_{BH}}{A_{4D}}\times {\left(mA{s}_{4D}\right)}^{B_{4D}}\right)}^{\frac{1}{B_{BH}}}$
Beyond varying mAs values, application of iterative reconstruction (IR) was used to evaluate the impact of image noise; IR is commonly used for noise reduction. The commercially available Siemens IR algorithm, SAFIRE, was applied with strength three (out of possible strengths 1–5). All 4DCT and BHCT phantom scans used for noise calculation were reconstructed with 512 mm extended field of view (FOV), 1 mm slice thickness, a medium smooth kernel (Br51f) and both with and without SAFIRE3 IR.