Phantom-Based Optimization of CT Gout Identification: Impact of Acquisition, Reconstruction, and Processing Settings
Abstract
Purpose
To assess the impact of parameters on CT gout identification
Methods
A phantom containing 28 insert mimics for gout, tendon, bone, tissue, and pseudogout was created. A baseline 70keV image with gout identification overlay was created from our clinical protocol. Further images were acquired with varying tube current (mA), kernel, reconstruction, thickness, uric acid (UA) and hydroxyapatite (HAP) thresholds, noise reduction strength, and software voxel size. Images were assessed on a Likert scale with 0 no gout identified, 1- 80% volume identified as gout. Non-gout inserts were identified on a similar scale with 0 no false positives and 4- >80% of insert incorrectly identified as gout.
Results
Changing reconstruction kernel was most impactful (15 inserts changing with 1.5-point mean rating shift) along with large UA threshold shifts (14 inserts changing with 2.4-point mean rating shift). Noise reduction strength (5 inserts changing; 1.6-point mean shift), small uric acid threshold shifts (3 inserts changing; 1.6-point mean shift), and slice thickness (3 inserts changing; 2.0 mean shift) also had a substantial impact. Larger and higher concentration uric acid inserts were less impacted. Pseudogout only produced false positives when the reconstruction kernel was changed, though larger changes to the HAP threshold than those tested would likely induce false positives in the pseudogout. Typically, changes improving gout identification sensitivity also increased false positives with tendon. Tendon was the most likely material to exhibit false positives.
Conclusion
Gout identification is surprisingly insensitive to individual settings, but care must be exercised when changing UA detection thresholds, slice thickness, and reconstruction kernel. A trade-off occurs between positive gout identification and volume of false positives, especially in tendon.