Description: 75x75

MIDAS eNewsletter

 

MAY, 2007

New and Improved

 

MSREG:  This is a new application for Rigid Registration. This serves several functions: 1) to correct for subject motion between MRI and MRSI image series in a single study; 2) to align all images of the first study taken over multiple studies; 3) to make all DICOM position vectors consistent; and 4) to shift the SI data to place the image in the center of the image. This latter function benefits MRSI data taken using the 8-channel phased array data, where the head is placed higher in the magnet but the FOV remains centered, sometimes causing wrap-around of the nose region.

The following figure illustrates its use. In this case an arbitrary, but significant, spatial transformation (rotation and translation) was applied to each image series (MRI_T2, and SI_Ref in this case) of a study that was already aligned. The MSREG program was then applied to correct for the misalignment.

Description: MSReg_Example

Initial evaluation of this spatial transform indicates that the effect on the quality of the reconstructed metabolite images is negligible. Further evaluation studies are currently underway.

The MSREG program should be applied after spatial reconstruction of the SI data and before the brain or lipid masks are created.

BATCH: This program now runs from the command line. The applications for this will be: 1) to automatically kick off the processing pipeline following import of the data; and b) to run a series of automated test procedures.

VIEWER: This program has been extensively rewritten and is now built upon the open source Java ImageJ software (http://rsb.info.nih.gov/ij/). Changes include:

o   Viewing images after spatial transformation into the Atlas space: This includes display of results of the statistical tests generated using the MSTAT program. Metabolite data from a single subject can now be compared to average metabolite data generated from a group of subjects. This is done by selecting the "Atlas Space" tab from the Midas Viewer tool bar.

o   Comparisons of spectra from multiple voxels: Voxels can now be "marked" and a copy of the spectrum at that location is made, thereby enabling comparisons to other Voxels.

o   ROI Support: The Midas Viewer now has ROI analysis capability.  ROIs can be drawn in any volume and the ROI is applied to all the open volumes. A number of different measurements can be taken, including for example: Area, Standard Deviation, Mean, Median, Mode and Perimeter.  ROIs can also be saved and loaded to compare measurements across multiple subjects in standard space.

o   Performance Enhancements:  Loading MRI data is now significantly faster. Memory issues are handled much more gracefully. When the Midas Viewer runs out of memory a dialog is displayed prompting the user to close any number of open volumes to free memory.

o   Window Level Adjustment:  The Window and Level of an Image is now easily adjusted by using the conventional method of dragging the mouse over the target image.

SEGMENTATION: A new fuzzy clustering algorithm has been implemented and is included in the new MIDAS release. The system works in an incremental fashion, making it is faster than FSL (by about 17 times for the 512x512 1.5T MRI data and 10 times for the 3T 256x256 data), as well as our previous approach. The algorithm clusters part of the volume, creates weighted exemplars and uses the exemplars in clustering the succeeding parts of the volume.  New weighted exemplars are created each time a portion of the volume is clustered. The clustering is done on T1-weighted images by default; however, it can also be applied to multispectral images.  Using just T1-weighted images it results in segmentations that are much closer to FSL than our slice by slice approach. These segmentations have more CSF and are robust. There is also a partial volume feature available. Figure 1 shows a segmented slice without partial volume and Figure 2 shows a slice with a partial volume segmentation.

Description: figure1-0407

Description: figure2-0407

Publications are in preparation: [1] Prodip Hore, Lawrence Hall and Dmitry Goldgof, Single Pass Fuzzy C Means, IEEE International Conference on Fuzzy Systems, July, 2007, To Appear. [2] Prodip Hore, Lawrence O. Hall and Dmitry B. Goldgof, Creating Streaming Iterative Soft Clustering Algorithms, NAFIPS 2007, June, 2007, To Appear.

MIDAS Tips, Questions, and Answers

Importer Performance: A known issue with the Importer is the length of time it takes to process data from a phased-array EPSI study, especially when transferring data over a network. We are looking at improving performance, but we should also keep in mind that volumetric, phased-array, EPSI data is huge. For a 12-channel acquisition, we have 16 Gb for the raw SI and maybe another 100Mb for the MRI. Each DICOM file needs to be read into memory, and the binary data saved out, so our ability to speed up this step is limited.

Plans include the option to combine the import with the processing in a batched pipeline mode, so at least you won’t have to wait for the import to start the processing.

JavaBridge Memory: The IDL JavaBridge includes several options in the file C:\RSI\IDL62\external\objbridge\java\idljavabrc

One of these controls how much memory is assigned at initialization, and is memory assignment problems are encountered using the Java programs then it may be worthwhile adjusting this to match your system. Look at the command:

JVM Option2 = -Xmx512m

EPSI WIP

VB13 is now available. A timeout problem with the data transfer has been temporarily fixed by inserting a delay before each transfer. Full fix will require change to the computer by Siemens. This version includes full support for data export in DICOM format.

Developer’s Corner

“Register Viewer” Parameter:  This parameter is included with every image to indicate its relevance for the viewing programs. Normally, most investigators are interested in viewing resultant metabolite and MRI images, and not many of the intermediate image files generated as part of the processing, as for example the complex weighting functions used for phased-array combination or residual values from the spectral fitting. Images considered of primary interest are labeled ‘TOP’, while the others are ‘Other’.

The SID program allows changing which frames are displayed in the reference image window (lower left) from the Options/Toggle list of images.

More Midas

Description: Haliotis_midae Midas's ear

Haliotis midae : Midas Ear Abalone.

Original link: http://www.gastropods.com/4/Shell_964.html

 

 

Andrew Maudsley

May 2007

 

This newsletter is aimed at providing information to the developers and users of the MIDAS software package. If you would prefer not to receive notification of these reports just let me know (amaudsley@med.miami.edu).