Program Information
Efficient and Quantitative Measurement of Task-Based Image Quality Metrics in CT
S Dirks*, L Yu , T Vrieze , C Favazza , A Ferrero , S Leng , J Weaver , C McCollough , Mayo Clinic, Rochester, MN
Presentations
TH-EF-601-5 (Thursday, August 3, 2017) 1:00 PM - 3:00 PM Room: 601
Purpose: Task-based image quality metrics, including channelized Hotelling observers (CHO), have been validated in detection and localization tasks for non-linear iterative reconstruction (IR) algorithms. However, its practical use for quality control or protocol optimization has been limited by the lack of efficient implementations. The purpose of this work was to develop and implement an efficient and quantitative approach to measuring task-based image quality metrics in CT for low-contrast detection tasks.
Methods: A phantom containing small spherical and cylindrical objects at various contrast levels and sizes was 3D-printed. Task-based image quality metrics, including CHO models that were validated previously for both detection and localization tasks, and contrast- and noise-dependent spatial resolution, were implemented on an integrated software platform using this phantom. The number of channels in the Gabor filter was optimized to improve the statistical properties of CHO and the number of repeated scans was reduced. The area under the ROC curves (AUC) and the index of detectability (d’) for all low-contrast objects were calculated as the output of this software platform. Traditional metrics such as SSP, MTF, and 3D NPS were also incorporated.
Results: CHO-based AUC and d’ were efficiently calculated as functions of tube current settings for different reconstruction algorithms and for varying phantom sizes. Without sacrificing accuracy, the number of Gabor filter channels in the CHO calculation were determined to be 12 (4 passbands x 3 orientations), which required only up to 40-50 images that can be acquired with approximately 15-20 repeated scans at each dose level.
Conclusion: An efficient and quantitative approach was developed to measure CHO-based image quality metrics and contrast- and noise-dependent spatial resolution. This approach can be directly translated into clinical practice for CT protocol optimization.
Funding Support, Disclosures, and Conflict of Interest: CHM received research support from Siemens Healthcare.
Contact Email: