(Sunday, 7/29/2018) 1:00 PM - 1:55 PM
Room: Room 205
Purpose: An upcoming AAPM task-group report, TG-233, recommends using task-based metrics to properly characterize image quality in CT. However, there are few tools available to clinical physicists to estimate such metrics. The purpose of this work was to design, implement, and utilize a software package, imQuest, to facilitate task-based image quality assessment and matching CT protocols across clinical practice.
Methods: The software is comprised of a graphical interface running on top of a computational library of Matlab functions allowing the user to quickly display a CT series and measure image resolution and noise properties in terms of the task transfer function (TTF) and the noise power spectrum (NPS), respectively. The TTF and NPS are further deployed to ascertain detectability indices based on a non-prewhitening matched filter observer model. To demonstrate the software’s utility, CT images of the ACR image quality phantom from six scanner models (GE Lightspeed 16, GE VCT, GE 750 HD, GE Revolution CT, Siemens Flash, and Siemens Force) were acquired using a consistent technique. Images were reconstructed with all available kernels and iterative algorithms for a total of about 1600 image series. For each image series, the imQuest software automatically measured the TTF and NPS. The results were used to determine the reconstruction settings needed to homogenize image quality across our fleet of scanners for our standard chest protocol.
Results: The software automatically analyzed all 1600 image series without any failures. It was found that the Standard kernel (with ASiR/ASiR-V 30% when available) on GE scanners produced images with very similar noise and resolution properties as the I41f\2 and Bf40d\2 kernels on the Siemens Flash and Force, respectively.
Conclusion: The imQuest software package provides the necessary tools for an imaging physicist to utilize the task-based image quality assessment framework recommended by TG-233.
Keywords
CT, Image Analysis, Lesion Detectability
Taxonomy
Contact Email
