Click here to


Are you sure ?

Yes, do it No, cancel

Task-Based Evaluation of a Novel High-Matrix Size Reconstruction Employed to Best Utilize Spatial Resolution in State-Of-The-Art Computed Tomography

T Schuermann1*, J Solomon2 , E Samei3 , J Ramirez-Giraldo1 , (1) Siemens Healthineers, Durham, NC, (2) Duke University, Durham, NC, (3) Duke University Medical Center, Durham, NC, (1) Siemens Healthineers, Durham, North Carolina


(Monday, 7/15/2019) 1:15 PM - 1:45 PM

Room: Exhibit Hall | Forum 9

Purpose: The maximum achievable spatial resolution of CT is often not realized in clinical practice due to choices such as the display field of view (FOV). The purpose of this study was to perform a task-based evaluation of the impact and tradeoffs of high matrix size reconstructions.

Methods: An ACR phantom was scanned with a third generation dual-source CT system using standard body protocol: 120 kVp, 200 reference mAs, 192x0.6mm collimation, 0.5 seconds rotation time and pitch of 0.6. Offline reconstructions were performed with matrix sizes of 512, 1024 and 2048 using weighted-filtered back projection (WFBP) as well as advanced modeled iterative reconstruction (ADMIRE), and using a medium (Qr40) and sharp (Qr69) non-edge enhancing kernels. Task-based modulation transfer function (TTF), noise-power spectrum (NPS) and qualitative bar patterns in ACR phantom were evaluated. Tradeoffs were calculated by recording reconstruction times, storage requirements and image noise.

Results: Larger FOV reconstructions (50cm) and sharp kernel correlated with improved spatial resolution as measured by TTF (0.72, 0.98, 1.00 mm-1), bar pattern (6, 10, and 12 lp/cm), noise (52, 141, 194 HU) for the 512, 1024 and 2048 matrix sizes, respectively. NPS of 512 matrix reconstruction exhibited aliasing which less or not present with 1024 and 2048 matrices. FOV of 25cm and medium kernel correlated with only modest differences for TTF, lp/cm, and noise. Compared to traditional WFBP at 512 matrix size: reconstruction times increased by 1.5- and 4-fold for the 1024 and 2048 WFBP reconstructions, and by 4- and 25-fold for the 1024 and 2048 ADMIRE reconstruction. Storage size requirements increased by a factor of 4 and 16 for the 1024 and 2048 matrix sizes, respectively.

Conclusion: Larger matrix size improves spatial resolution particularly for sharper kernels and larger FOV reconstructions. Tradeoffs such as increased reconstruction times and storage need to be considered.


Reconstruction, Resolution, MTF


IM- CT: Image Reconstruction

Contact Email