Phantom-Based Automated Quality Control of Proton Density Fat Fraction Imaging Using Embedded Pocket Phantoms In Clinical MRI
Abstract
Purpose
Proton density fat fraction (PDFF) is widely used as a quantitative MRI biomarker in clinical care. Accurate PDFF measurement requires robust quality control, commonly performed using phantom-based assessment. This work evaluates an automated algorithm for quantitative analysis of a commercial pocket phantom within patient MRI exams to facilitate embedded phantom-based quality control.
Methods
Pocket phantoms (Phantom Packs, Calimetrix, Madison, WI) with nominal PDFF values of 0, 10, 20, 30, and 75 percent were positioned within the imaging field of view beneath patients during routine MRI examinations (IDEAL-protocol), with data from three studies included. A Python-based automated analysis pipeline was used to detect phantom vials on water-only DICOM images, and map circular regions of interest to the corresponding PDFF DICOM images. Mean, median, standard deviation, minimum, and maximum PDFF values were automatically computed for the slice identified by the algorithm as optimal for phantom analysis. Automated PDFF measurements were compared to known nominal and manually calculated phantom values to assess quantitative consistency and identify out-of-range measurements.
Results
The algorithm successfully detected phantom vials and generated PDFF measurements, with values closely matching known nominal PDFF values. As a representative case, the mean(standard deviation) PDFF values measured in each of the vials were 3.6(±0.7), 9.2(±0.5), 22.1(±0.6), 31.4(±0.6) and 74.4(±1.0) reflecting consistency with nominal values. Further, the automated framework selected this slice for quantitative analysis, yielding PDFF measurements consistent with manual measurements, with near-unity agreement (slope ≈ 0.99, R ≈ 1.0).
Conclusion
Automated analysis of PDFF pocket phantoms in patient MRI exams enables effective embedded quality control. Using known phantom values as a reference, this approach reduces reliance on manual review and enhances the efficient of quality assurance processes. It has the potential to improve data consistency in clinical trials and support routine clinical MRI quality assurance.