Hide/Show Apps

Temperature estimation using magnetic nanoparticles: a simulation study.

Onuker, Gamz
Lately, in biomedical imaging systems, algorithms for providing the temperature data became more important both for diagnosis and for treatment. Aiming to serve this purpose, in this thesis study, it is planned to estimate the temperature data using magnetic nanoparticles; which are actually imaging tracers used in a new medical imaging modality, Magnetic Particle Imaging (MPI). Since the introduction of MPI, the effect of temperature changes on magnetic nanoparticles (MNPs) and their magnetic behavior, is being investigated and the results show that there is a non-linear relation between the temperature and the particle magnetization. For this reason, as the starting point of this study, a detailed MPI system was modelled using MATLAB with its main components, which are the Selection Field, Drive Field and Receiver coil pairs. All coils were modelled to have a circular shape, the Selection and Drive field coils have 25cm radius whereas the receive coils have 10cm radius. Scanning sequence was chosen to be ’Lissajous Trajectory’ with sufficient density in order to provide a good spatial resolution. After determining the geometrical setup, a system matrix was obtained, which describes the magnetization signal, induced by the MNPs, as a function of the imaging space. In order to reconstruct images by the use of the system matrix, Algebraic Reconstruction Technique (ART), Selective Singular Value Decomposition (SSVD) Method and Truncated Singular Value Decomposition (TSVD) Method were applied to solve the matrix equation. Reconstructed images have resolution ranged in sub-millimeter level. In order to examine the effect of temperature changes in the model, a thermal solver was implemented using Pennes’ Bioheat Equations. Energy of a focused ultrasound beam was used as the heat source and the equation was solved both in time and space, using the Finite Difference Time Domain Method (FDTD) in 2-Dimensional space. As a result, a realistic time-varying temperature data was included in the MPI model. Previous studies show that it is possible to estimate the temperature by utilizing a calibration curve that is acquired from the ratio of the fifth and the third harmonics of the magnetization signal, generated by MNPs, under different strength of applied sinusoidal field. But this was showed only for spectroscopic (0-D) data acquisition systems. In this thesis study two different methods were proposed for temperature distribution and applied for noise free case. For the first method, the algorithm is based on pixel-wise scanning and using a calibration curve obtained from 1D line scanning. The resolution and the maximum relative error was found to be 0.01°K and 0.003 per cent, respectively. Secondly, a linearization method for temperature estimation was proposed and the resolution and the maximum relative error was found to be 0.001°K and 0.063 per cent, respectively.