Magnetic resonance imaging (MRI) is an imaging technique used primarily in the medical profession to produce high quality images of the inside of the human body, without using x-rays. The process of obtaining the spin density spectrum from the time samples of the spin signal for each pixel of a magnetic resonance image is known as MRI spin relaxometry. The spin density spectrum is obtained by performing linear regularization on samples of magnetic resonance images obtained at different times. The spin density spectrum obtained from the body is used by radiologists to diagnose disease in its early stage. However, the computation involved is substantial, and processing the pixels of the images in parallel reduces the time it takes to obtain the spin density spectrum. Parallel computing is suitable for the computations because the value of each pixel is independently calculated. This project involved developing a parallel algorithm implementation using the Parallel Java Library (developed by Prof. Alan Kaminsky) and an MPI implementation in C. The algorithm implementation was done using both linear regularization and the non-linear least squares approach along with proper load balancing. Also, a performance comparison was done between both the implementations for C as well as Java. Appropriate user interfaces were also developed for both the implementations.
Department, Program, or Center
Computer Science (GCCIS)
Parikh, Hardik, "Solving an MRI spin relaxometry problem using Parallel Java" (2008). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus