Purpose: To validate an automated MTF-measurement procedure for achieving accurate and objective assessment of digital x-ray system spatial resolution in routine quality control surveys.
Methods: The automated procedure (autoMTF) was evaluated for a full-field digital mammography system, a Senographe Pristina (GE Healthcare, IL), wherein the manufacturer prescribes sub-system MTF sampling at defined object frequencies. The autoMTF procedure utilizes a detection algorithm to (i) locate the line-pair phantom (ii) place elliptical ROIs in prescribed locations, and (iii) output tabulated ROIs statistics from which MTF samples can be calculated. Phantom images were acquired for both targets (Mo, Rh), both focal spots (small SFS, large LFS), two bar patterns, and phantom directions (width, length). Manual and autoMTF measurements were performed by placing four ROIs in each image: at the “space”, “bar”, and two line-pair sections (2.09 and 3.93 lp/mm for LFS, 5 and 8 lp/mm for SFS). Images were copied to external storage media (CD or USB drive) for autoMTF processing. Performance in terms of procedure time and accuracy was evaluated by three participating medical physicists.
Results: The autoMTF completed the resolution testing, including storage media manipulation, in an average of 4 min. compared to manual measurements in 15-30 min. Test outcome (pass/fail) was equivalent for both procedures. Physicist measurements exhibited larger percent deviations (2-24%) compared to autoMTF, as well as more inter-variability, for the most challenging ROIs, i.e., higher spatial frequencies.
Conclusion: This work produced a tool to facilitate oftentimes tedious and lengthy spatial resolution evaluations that are prone to error and inter-observer variability. The autoMTF algorithm is versatile and can be ported to similar spatial resolution testing in other modalities such as digital radiography and fluoroscopy. Future investigations will utilize the autoMTF tool to quantitatively assess physicist subjectivity in performing line-pair phantom tasks.