For this project I aided in smoothing ultrasound data from a PhD student's research. The motivation for this project is that heart defects in newborns develop in embryonic stages. If the development of the heart in these stages can be studied using CFD and muscle strains simulations, more can be learned about these defects. Stacked ultrasounds over the period which the heart beats are used to make a 3D model for simulations. This data creates a model with a lot of high frequency noise in the movement over time, and some artifacts of the discrete data in space. The data is smoothed over time using a Fourier transform, which approximates a function using superposition of sine and cos waves. Then, the high noise frequencies are removed from the function. I did this using Matlab Fast Fourier Transform (fft) and its inverse (ifft). The data is smoothed in space using a spherical harmonic transform at a relatively low value of of order l, which approximates the surface with less spherical basis functions to make a smoother shape. For this step I used the uja_shfd code from this article where researchers were doing something similar with a model of a brain for their own research. This smoothed model can be used in simulations for more accurate results.
If you’d like to look at the time smoothing code for this process, check out my GitHub, and if you’d like to learn more you can read my final report below.
