Automated Tools for MR Imaging QA
Abstract
Purpose
To develop a unified and fully automated desktop-based MRI coil QA application designed to minimize operator dependence and enhance workflow efficiency
Methods
Python-based desktop applications were developed with specific processing pipelines for routine quality assurance (QA) procedures recommended by NEMA for body, torso, and head and neck radiofrequency coils. These applications analyzed DICOM images and DAT files using an automated algorithm employing Otsu thresholding and morphological refinement to delineate the phantom boundary. Circular signal regions of interest (ROIs) were placed on the identified phantom, while noise ROIs were defined as constant-area circular shapes (340 cm²) to ensure inter-dataset consistency. The QA processes generated statistical metrics of the acquired images, including derived measures such as signal-to-noise ratio (SNR) and percent image uniformity (PIU). For torso coil assessments, SNR and uniformity were additionally calculated for combined views of paired norm-off, normalized and unnormalized datasets. Furthermore, for each coil element, the SNR of single coil channels was calculated in accordance with established QA protocols. To assess magnetic field homogeneity in the frequency domain, raw free-induction-decay (FID) data were analyzed. Tests were performed on the ViewRay’s MRIdian®, and the corresponding output data files were processed using the twix package. From these, the center frequency offset, full width half maximum (FWHM), and magnetic field homogeneity in parts per million (ppm) were derived. All outputs were automatically exported into structured excel reports and plots.
Results
The developed applications automatically performed QA analyses and generated reproducible metrics for body, torso, and weekly QA datasets. The structured outputs enabled comprehensive assessment of all coils across configurations. Routine clinical QA was automated through standardized image processing and fixed-geometry ROIs, thereby enhancing reproducibility and minimizing user-dependent variability.
Conclusion
This work automated and standardized MRI coil performance evaluation, ensuring reproducible and efficient analysis. This automation enables long-term independent QA monitoring.