Automated MPI segmentation in X-space and calibration to quantify iron concentration in a tracer distribution

Abstract

Magnetic particle imaging (MPI) is a new, radiation-free medical imaging modality that relies on the non-linear magnetization response of superparamagnetic iron oxide nanoparticles (SPIONs) to reconstruct their concentration distribution with high sensitivity and medical safety. Current quantification methods for region of interest (ROI) are inadequate and are usually outlined manually or using deep learning methods. We propose two new models for ROI selection based on machine learning, one is the K-means++-based threshold-inflated image segmentation model and the other is the image segmentation model based on MPI simulation and SVM. We have developed an accurate quantification of 2D MPI images and established the calibration curve to predict the corresponding iron content based on the MPI image.