Inverse Problem Analysis in Magnetic Nanoparticle Tomography with Minimum Variance Spatial Filter

Abstract

INTRODUCTIONIn recent years, imaging of magnetic nanoparticles (MNPs) has attracted attention as a new diagnostic imaging technique for the imaging of certain targets, such as cancer cells [1]. We previously proposed an MNP imaging method, “magnetic nanoparticle tomography” (MNT), that uses multiple magnetic sensors [2], [3]. In the previous study, an inverse problem analysis was performed to estimate the position of MNPs using the non-negative least squares method (NNLS), which is often used for magnetic particle imaging (MPI). The NNLS method achieves high sensitivity and spatial resolution. However, due to the existence of measurement noise and high sensitivity of NNLS, some MNPs are incorrectly estimated, that is, some artefacts appear. In this study, we apply the minimum variance spatial filter (MV-SF), which is often used in magnetoencephalography (MEG) inverse problem analysis [4], to reduce the artefacts.METHODSThe experimental setup is presented in Fig. 1. The Resovist MNP sample was arranged in a sinusoidal magnetic field, which was generated using an excitation coil. The third harmonic magnetic field from MNPs was detected using 16 detection coils. To improve the sensitivity, a cancelation circuit for the fundamental magnetic field was employed. The two-dimensional concentration map of the MNP sample was subsequently obtained by the application of the MV-SF. The distance between the MNP sample and the imaging system (excitation coil and detection coils) was set from 25 mm to 40 mm at intervals of 5 mm. We used two samples with different concentrations of 100 µg-Fe and 500 µg-Fe and estimated the positions for each of them.RESULTSThe result of the reconstructed map when the MNP sample was set at are shown in Fig. 2. The white circle in the figure represents the actual sample position. Fig. 2 (a) depicts the result for a sample containing 500 µg-Fe. This result reveals that a sharp signal peak position is observed in the vicinity of the sample position. Fig. 2 (b) exhibits the result for a sample containing 100 µg-Fe. This result indicates that the error between a sharp signal peak position and the actual sample position is within 10 mm. No artefacts appear in both results, contrary to the results using NNLS [3]. From these results, we can conclude that MV-SF is useful for reducing the artefact and estimating the positions of MNPs in magnetic nanoparticle tomography.ACKNOWLEDGMENTSThis work was supported by the Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for Young Scientists [grant number JP19K14996], and by a research grant from the Mazda Foundation. **