2023 AIChE Annual Meeting

Impact of Sample Volume on Magnetic Particle Imaging Quantification Accuracy

Magnetic Particle Imaging (MPI) is a noninvasive and radiation-free imaging technique with the ability to directly quantify superparamagnetic iron oxide nanoparticle (SPION) tracers. MPI uses nonlinear re-magnetization behavior found in SPIONs to generate a signal directly related to SPION mass. This allows for negligible background signal and tissue attenuation making MPI an increasingly attractive modality for medical applications such as cancer diagnosis and tumor sizing. Despite this, the quantification accuracy of MPI may be hindered by its finite resolution limits which are seen in other medical imaging devices, such as MRI, PET, and CT. With resolution-based errors being a normality in novel imaging modalities, this study aims to detect the presence of these errors arising from changing volume sizes in MPI.

To assess quantification error, 3D-printed disc shaped phantom models were designed to have volumetric cavity sizes ranging from 16-512µL while maintaining a constant depth. Models were then loaded with Feruocarbotran, a commercially available SPION tracer, using a constant mass dilution series, generating a set of models with constant mass, constant depth relative to scan projection, and varying volumes. Disc models were then imaged to produce 2D projections of tracer distributions in different MPI built-in scan modes: High Sensitivity (HS), High Resolution (HR), and High Sensitivity/High Resolution (HSHR). A set of spherical 3D-printed models were loaded using the same procedure but imaged to produce 3D projections of tracer distributions in HR scan mode. The mass of each model was estimated by using three constant volume calibration curves which constructed a scalar relationship between the total signal and tracer mass. The difference between estimated mass and ground truth defines the error.

Results show significant quantification error in tracer mass estimation. As the sample volume size deviated from that of the calibration curve, the percent error increased indicating a relationship between calibration curve size, model size, and quantification accuracy. Quantification error showed a scan mode dependency with the High Sensitivity scan mode having the least amount of error. 3D scans showed more accurate quantification likely due to spatial resolution improvement from numerous scan angles which allowed for overlaid projections, possibly reducing the effects of resolution-based errors in one direction. Although some scan parameters showed more accurate results, there were still consistent quantification errors present throughout the trials. Because of this, addressing the quantification inaccuracies associated with volume variation is the next pivotal step to increase MPI’s suitability for clinical applications.