2-069
Machine Learning-Based Noise Analysis of Magnetization Response Signals for Magnetic Particle Imaging
◎Naboua Donan Jose Agbehounkpan・Bagus Trisnanto Suko・Yasushi Takemura(Yokohama National University)・Asahi Tomitaka(University of Houston-Victoria/Kennesaw State University)・Satoshi Ota(Shizuoka University)
Magnetic particle imaging (MPI) is a new tomographic imaging technique that uses the non-linear response of magnetic nanoparticles (MNPs). Applying an oscillating magnetic field with frequency f0 and sufficiently high amplitude A to MNPs cause them to exhibit magnetization M(t), where t is time. M(t) consists not only the applied frequency f0 but also the harmonic frequencies. Odd-harmonic frequencies can be isolated from an incoming signal using various filtering methods. However, noise is widespread during measurement, which reduces the signal quality and makes the retrieval of the desired frequencies challenging. The goal of this study was to demonstrate that machine learning approaches can be used to target specific types of noise, such as Gaussian or 1/f noise, and remove them from the measured signal, thereby increasing signal quality.