Denoising particle trajectories containing measurement uncertainties from magnetic particle tracking using neural networks

Magnetic particle tracking is a recently developed technology that can measure the translation and rotation of a particle in an opaque environment like turbidity flow or fluidized-bed flow. Existing analytical trajectory reconstruction algorithms usually require magnetic field gradients, which are challenging to measure accurately. A novel analytical method is developed for an arbitrary sensor arrangement to resolve this. Additionally, to reduce measurement uncertainty in practical applications, deep neural network (DNN)-based models are developed to denoise the reconstructed trajectory. Compared to traditional approaches such as wavelet-based filtering, DNN denoisers are more accurate in position reconstruction. However, they often over-smooth the velocity signal, and so a hybrid method is developed that combines the wavelet and DNN models, for a more accurate velocity reconstruction. DNN algorithm is trained using processed sensor and image data collected from a series of experiments. Processed image and sensor data serve as the label and input to DNN respectively. 300000 points in space are to be processed and denoised. We expect DNN to show vast reduction in noise and fluctuation levels of the position, orientation, and velocity signals.