Introduction to PyNUFFT (Python non-uniform fast Fourier transform)
Jyh-miin Lin (~jyh-miin) |
PyNUFFT provides a pure Python implementation of the NUFFT algorithm (Fessler and Sutton, IEEE TSP, 2003), which enables users to freely access the core NUFFT function in Python. PyCUDA/PyOpenCL also offers a significant acceleration with the newest graphic processing units (GPUs). To date, PyNUFFT has been used by research groups in Swissland, France, Germany, Netherland, USA, and China.
In this talk, Dr. Jyh-Miin Lin will cover the following points:
Introduction to the basic concepts of non-uniform fast Fourier transform (NUFFT), and why it is important in data science/industry. (5 min)
A brief introduction to the advantage of Python, and how PyCUDA/PyOpenCL code snippets are integrated into PyNUFFT using the Reikna package. (5 min)
Installation and testing of PyNUFFT (5 min)
Benchmarks of PyNUFFT on the GPU. (5 min)
Some applications of PyNUFFT in compressed sensing MRI and other areas. (5 min)
Q&A (5 min)
- Engineers who plan to work in the field of iterative medical imaging reconstruction (MRI or tomography) using Python
- Engineers who are familiar with Numpy/Scipy, and fundamentals of Fourier transform
- Engineers interested in GPU computing using PyCUDA/PyOpenCL
Dr. Jyh-Miin Lin has an MD and MSC (EE) from National Taiwan University and a Ph.D. (Radiology) from the University of Cambridge. He is interested in image processing, modeling, and cloud/edge computing in medical imaging.