TH-CD-207B-04: Is TTF a True Representation of the Sharpness Property of a Non-Linear CT System?

Authors


Abstract

Purpose:

To investigate if the task-transfer-function (TTF) accurately models the transfer properties of a CT system for lung nodule imaging.

Methods:

An anthropomorphic lung phantom was imaged using a standard chest protocol on a clinical CT scanner with and without 24 physically inserted synthetic lesions (nominal diameter: 8 – 10 mm). Images were reconstructed using FBP and iterative algorithm (SAFIRE, Siemens Healthcare). 3D TTF was measured using an established technique. Corresponding idealized virtual lesions were blurred with the TTF and superimposed onto lesion-less phantom images. Images of the physically and virtually inserted lesions were compared in terms of rendition of spatial features (blurriness of the edges), lesion morphology, and lesion volume. Feature rendition was measured in terms root-mean-square (RMS) of the frequency power of the native and TTF-transferred lesion edge transition. Morphology was assessed with the Regional Hausdorff Distance (RHD) using custom written code (MATLAB v2015b). Volumes were measured using a clinical segmentation tool (iNtution, TeraRecon).

Results:

The RMS was less than 0.02 and 0.05 for FBP and IR respectively. Using the nonlinear mixed effect (nlme) package (R, www.r-project.org), the difference in RHD between virtual and physical lesions was 5% on average. There was less than 1 ± 5% (R2 > 0.97) and 3 ± 4% (R2 > 0.97) difference between the volumes of the physical lesions and the corresponding virtual lesions for FBP and IR respectively. Additionally, was closer concordance for images reconstructed with FBP than iterative reconstruction.

Conclusion:

The TTF was found to accurately model the transfer properties of the CT imaging system on lung lesions for both FBP and iterative reconstruction algorithms. TTF was found to offer slightly better lesion renditions when modeling images reconstructed with FBP versus iterative reconstruction. This methodology will be used in future investigation of more complex imaging tasks such as low contrast detectability of known lesions.

Ancillary