Viewing the Analysis Results
Abstract. MRI spin relaxometry is the process of recovering the spin density spectrum from the time samples of the spin signal for each pixel of a magnetic resonance image. Since healthy tissue exhibits different spin relaxation rates from diseased tissue, MRI spin relaxometry potentially has utility for diagnosing disease. However, recovering the spin relaxation rates involves solving an inverse problem which requires substantial computation. The computation's running time can be reduced by processing the pixels in parallel on a parallel computer. A parallel program for solving the MRI spin relaxometry problem, SRSolve, was implemented in Java with MPI, its running time was measured on a 32-processor cluster parallel computer, and its performance was compared to the CONTIN program. CONTIN required about 44 sec on the average to solve one pixel and about 3600 sec to solve an entire 64x64-pixel test image (with 2,597 unmasked pixels) on the parallel computer. SRSolve required 3.04 sec on the average to solve one pixel and 263 sec to solve the entire image on the parallel computer.
Abstract. 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.
You may download a Java Archive (JAR) file containing documentation, Java source code, and Java class files for the MRI Spin Relaxometry Library.
Version 27-May-2005 (478,024 bytes): sr20050627.jar
You may download a GZIP TAR archive file containing a test MRI input data set as well as an output file with the results of the spin relaxometry analysis on the test data set.
GZIP TAR archive (5,556,495 bytes): data3.tgz
Instructions for using the SRView program are in the technical report (above).
The MRI Spin Relaxometry Library was developed using Java 2 Platform Standard Edition (J2SE) 5.0, also known as Java Development Kit (JDK) 1.5.0.
Program SRSolve uses mpiJava, a Java implementation of the Message Passing Interface (MPI) standard for parallel programming on cluster computers. To run this program, the mpiJava distribution must be installed. (The mpiJava distribution is not part of the Library.) For further information about MPI, see the MPI Standard (http://www-unix.mcs.anl.gov/mpi/). For further information about mpiJava and to obtain the mpiJava distribution, see the mpiJava Home Page (http://aspen.ucs.indiana.edu/pss/HPJava/mpiJava.html). For further information about running the SRSolve program, see the package summary documentation and the SRSolve documentation.
Note: mpiJava is not needed to run the SRView program.
The Library is distributed as a Java Archive (JAR) file including Java class files, Java source files, and Javadoc.
To run the programs, simply store the JAR file somewhere and specify the JAR file in the Java class path.
To examine the Java source files and Javadoc, unpack the JAR file. The Javadoc home page is in the file "doc/index.html" under the directory where the JAR file was unpacked.
The files in the MRI Spin Relaxometry Library ("The Library") are copyright © 2005 by Alan Kaminsky. All rights reserved. For further information, contact the author, Alan Kaminsky, at firstname.lastname@example.org.
The Library is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
The Library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
A copy of the GNU General Public License is provided in the file gpl.txt. You may also obtain a copy of the GNU General Public License on the World Wide Web at http://www.gnu.org/licenses/gpl.html or by writing to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.