Description

The high-resolution research tomograph (HRRT) is a dedicated human brain PET scanner. Unfortunately, clear, high-resolution images are almost never obtained from this scanner. During the acquisition time of 20 - 30 minutes for a typical brain scan, patients frequently move, thus corrupting the image. When imaging a rigid object, such as the brain, these movements may be tracked and recorded with fairly high precision. Here we assume that the motion information was recorded by Polaris VICRA from NDI. This motion information may be then used in the deconvolution process to reconstruct a better, sharper image. Parallel HRRT Deconvolution is an ImageJ plugin for motion correction of PET brain images.

References

• K. Wienhard et all The ECAT HRRT: performance and first clinical application of the new high resolution research tomograph. IEEE Trans. Nucl. Sci., vol. 49, no. 1, pp. 104–110, Feb. 2002.
  
• T. L. Faber, N. Raghunath, D. Tudorascu and J. R. Votaw Motion correction of PET brain images through deconvolution: I. Theoretical development and analysis in software simulations. Phys. Med. Biol. vol. 54, no. 3, pp. 797-811, Feb. 2009.
• N. Raghunath, T. L. Faber, S. Suryanarayanan and J. R. Votaw Motion correction of PET brain images through deconvolution: II. Practical implementation and algorithm optimization. Phys. Med. Biol. vol. 54, no. 3, pp. 813-829, Feb. 2009.

Features

• Multithreading (user can choose the number of computational threads)
• Four iterative methods:
• MRNSD - Modified Residual Norm Steepest Descent
  
• OSEM - Ordered Subsets Expectation Maximization
  
• CGLS - Conjugate Gradient for Least Squares
  
• HyBR - Hybrid Bidiagonalization Regularization
  
• Automatic segmentation of the motion information
• Visual segmentation editor
  
• Solve button - use different method and/or max number of iterations for the same input data 
  
• Two interpolation schemes (Nearest Neighbor and Trilinear)
  
• Different output types (Same as source, Byte, Short or Float)  
  
• Single and double precision
• Show iterations option
• Non-modal GUI (user can use ImageJ while the plugin's window is showing)
• The plugin can be called from ImageJ macro
  

Benchmark

Testbed

• 2 x Quad-Core Intel Xeon E5472 (3GHz, 12MB L2 Cache)
• 32 GB RAM
• Ubuntu 8.10 (64-bit)
  
• Sun Java 1.6.0_10 (64-Bit Server VM)
• ImageJ 1.42l
  

Settings

• Blurred size: 147 x 180 x 180 (real patient data)
  
• Automatic segmentation
• Data series: AVG
• Quaternions weight: 50
• Sampling rate: 20
• Scan duration: 5700
• Time offset: 150
  
• Auto segment size
• Interpolation: Nearest Neighbor
• Output: Same as source
• Precision: Single
  

Method	1 thread	2 threads	4 threads	8 threads
MRNSD (10 iterations)	21.57	17.17	15.56	14.86
MRNSD (50 iterations)	47.55	35.59	30.96	30.06
HyBR (10 iterations)	23.86	19.96	17.61	17.13
HyBR (50 iterations)*	29.95	24.70	23.18	21.25

Example

Input data	After 15 iterations of MRNSD

Documentation

To install the plugin, just download the binary distribution, unpack the archive into the plugin directory and restart ImageJ. The plugin will appear under Plugins » Parallel HRRT Deconvolution menu.

License

Parallel HRRT Deconvolution is released under the GNU General Public License, except for Parallel Colt.

Download

The source code distribution, besides Ant build file, contains also Eclipse project files. To build the source code you need to download Parallel Colt or use the jar files from binary distribution of Parallel HRRT Deconvolution.

version 1.5 (August 27, 2009) Changelog

binary:

• parallel_hrrt_deconvolution-1.5-bin.zip [4,890,070 bytes]
• parallel_hrrt_deconvolution-1.5-bin.tar.bz2 [4,889,221 bytes]

doc:

• parallel_hrrt_deconvolution-1.5-doc.zip [225,190 bytes]
• parallel_hrrt_deconvolution-1.5-doc.tar.bz2 [44,064 bytes]

source:

• parallel_hrrt_deconvolution-1.5-src.zip [105,765 bytes]
• parallel_hrrt_deconvolution-1.5-src.tar.bz2 [63,421 bytes]

Donations

If you have found this software useful and would like to thank the author, you can make a donation to show your appreciation.