Description: Description: Description: Description: Description: Description: Description: 75x75

MIDAS eNewsletter

 

SEPTEMBER, 2014

 

We are pleased to report continued support for the MIDAS support from NIH. Although at a reduced level from the original request the project will continue to be funded until 2017. Since the last newsletter in February 2013, when we announced a major release for Version 2.0, we have not had any major changes to the MIDAS package, but bug fixes and new features continue to be added at regular intervals and we therefore strongly recommend that you update your software installation regularly. For the latest release (version 2.35) several new features have been added, and are described in the following section.

New and

Improved

PRANA and DTI: MIDAS has supported import and display of other parametric image types such as perfusion or diffusion MRI for some time, and this has now been extended to include some DTI-specific statistical analyses methods in the PRANA program, specifically including voxel selection based on FA or MD thresholds. PRANA also supports the white-matter-parcellation atlas from the MRI Studio distribution (available on request). An example is in the following image, which shows z-score maps for FA and MD in a subject that had a severe TBI:

This image shows as an overlay on a reference group mean FA image (top) or MRI atlas reference image (bottom) those voxels that were significantly different from the mean value of the normal control group (for p<0.05 with application of the false discovery rate correction). The mean FA map and the z-score maps were created using the PRANA program, saved as Analyze format files, and displayed using the MRICro program.

PET-MRI-MRSI: In collaboration with Drs. Mauler, Langen, and Shah (Forschungszentrum Jülich, Germany), MIDAS now includes import of PET data in ECAT7 format and spatial registration of PET images to the MRI and MRSI data. The following example shows data acquired using a hybrid 3T MRI-PET system in a subject with a brain tumor:

QMAPS: An ongoing topic of investigation is dealing with voxels of inadequate quality in metabolite images. The QMAPS program provides a way of creating a “Quality Map”, based on a threshold value for the linewidth value returned from the FITT program. Changes to this program include the option to use the Cramer-Rao Bound value and to use the linewidth or CRB value from a spectral fitting of the water reference SI. While the Quality Map can be displayed along with the metabolite maps, this can still be difficult to interpret. An alternative approach is to remove the regions of questionable quality from the image, which has now been implemented in new option in the QMAPS program. The following figure shows NAA images with a) all fitted voxels, and b) with voxels not meeting a linewidth threshold of 13 Hz being set to zero.

In this example the bright signals in the frontal region, which could be misinterpreted, have been removed.

ZipStudy Utility: A new utility has been added that facilitates sending data between sites or archiving. From the “Tools” section of the MIDAS toolbar can now be found an option for “Copy Study Data and Zip”. This brings up a widget that allows selecting multiple studies and then different options for copying either the raw or processed data. For example, to send reconstructed data to another site the “Final” data can be selected and the option set to “Zip all files” options can be selected, which then packages all reconstructed images in a single zip file. By not including the raw data files a smaller file is produced that is easier to send to another site. This utility uses an excellent data compression program, “7z”, that is not distributed with MIDAS but must be downloaded from the developer (http://www.7-zip.org).

To view a dataset after it has been compressed it first needs to be uncompressed, which can also be done from the widget, and then the subject needs to be added to a MIDAS project using the corresponding function in the Importer.

MIDAS Tips, Questions,

and Answers

Multiple SI Series: Some studies may acquire more than one EPSI dataset in a single session, e.g. SI data could be acquired before and after some intervention, or data taken twice for different TE values. The MIDAS data management allows for multiple SI Series by using different Study labels, with the additional SI series being labeled SI2, SI3, etc. Similarly, additional SI_Ref data will be labeled SI2_Ref, SI3_Ref, etc.

The following example shows how 2 EPSI acquisitions, each with SI (Series 4 and 5) and SI_Ref (Series 4b and 5b) data, get displayed in the Importer:

After import the label assignment is then made as shown here:

note that the Series number is used to identify which is the first and second series, which is not necessarily the order in the list. This is how it then appears in the MIDAS Browser:

For automated processing of both SI series the BATCH processing file must be modified to include the second SI series in every step, i.e. as:

volumizer                     SI_Ref

volumizer                     SI

volumizer                     SI2_Ref

volumizer                     SI2

epsi2   EPSI2.xml      SI_Ref

epsi2   EPSI2.xml      SI

epsi2   EPSI2.xml      SI2_Ref

epsi2   EPSI2.xml      SI2

... etc.

Functions that are not for a specific image series, like MSREG, will automatically process both SI series. The MRMask program, which creates the SI-resolution brain and lipid masks from the MRI, will only save one result under the SI_Ref series, although in principle different masks could be generated for each SI series if needed.

Integrated Volumes:  The SID program provides an option for integrating spectra over a region of interest. Because the default EPSI acquisition and processing uses spatial oversampling and the resultant spectral datasets are heavily interpolated, the volume of tissue that corresponds to a ROI requires some explanation. First, some numbers:

·        The NOMINAL voxel volume, which is that defined by the k-space acquisition, for 50x50x18 k-space points over a FOV of 280x280x180 is 0.313 mL.

·        The INTERPOLATED voxel volume, after zero-filling to 64x64x32 points, is 0.107 mL.

·        The EFFECTIVE voxel volume, which includes the effect of apodization applied in the reconstruction, is 1.55 mL with elliptical k-space (applied in X and Z only) and 1.46 mL for square sampling (this value will vary with different spatial smoothing values).

The spatial-response function (SRF) is illustrated in the figure below (left). The integral of this is corresponds to the effective voxel volume, which is much larger than the interpolated voxel volume. Therefore, if 2 neighboring voxels are averaged the effective volume of tissue from which the signal came is not doubled, because there is considerable overlap of their SRFs. If spectra from 10x10 voxels in-plane are averaged, then the spatial distribution for this result is shown in the figure on the right, and corresponds to a volume of 32 mL. Note that 100 voxels times the interpolated voxel size is 10.7 mL, indicating that there is still a large signal contribution coming from outside of the 10x10 region used for the voxel selection.

For integration over a cube the relative contribution of signal from outside the cube is smaller than for the 2D example. The effective voxel volumes must therefore be calculated for the specific ROI. To get an estimate for different cases, the effective ROI volume has been calculated using a simulation of the PSF for voxels selected along a line (1D, green line), from a square N´N region in-plane (2D, blue line), or from a cubic N´N´N volume, where N is the number of voxels in the interpolated image in each dimension (3D, red line). From this we get the following plot:

Here, the effective voxel volume is plotted against the simple calculation of the number of voxels times the interpolated voxel volume (0.1076mL). With the exception of small numbers of points, these relationships are linear, so for example, if you integrate over 7x7 voxels in plane, the simple calculation of the volume is 7´7´0.1076 mL=5.27, and the effective volume of the ROI is:

2.83´5.27 + 0.84 = 15.7 mL

This is clearly an approximation for smaller regions and will not hold if the regions are not a straight line, square, or cubic, for the 1D, 2D, and 3D examples, but it gives a reasonable estimate.

EPSI

Text Box:  Zipper Artifact: With the short TE EPSI acquisition we encountered a “zipper artifact”, which occurs along the Y direction for the center of the phase-encoded dimensions (which is at x=33 and z=17), and would be most strongly seen in the choline map. An example is shown to the right. This is believed to result from transverse magnetization from the water in out-of-slice regions created by the refocussing pulse, and shows up in the center of the phase-encoded dimensions as it remains unchanged with every acquisition, making it look like a DC baseline artifact.

This artifact has been significantly reduced by shifting the phase encoding to be after the 180 pulse. This effect can also be minimized by ensuring good placement of the saturation slab to reduce signal from the lower edge of the slab selection profile.

Developer’s Corner

METAFIT: The METAFIT program is a wrapper for the FITT program that provides some improvement in image quality for short TE studies. This was introduced in the February 2013 newsletter and a few more details of the program are given here.

This program runs the spectral fitting three times. The first just fits the singlet peaks and is used to apply the phase and frequency correction to the data. This result is then spatially smoothed and this result is then fit using the full spectral model, which may, for example, include contributions from glutamate (or Glx), myo-inositol, and lactate. Since the SNR is better for the smoothed result the program can get a better result for the low SNR metabolite signals. For the final step the original data is restored, with the option of subtracting the baseline determined on the previous fit, and third fit performed that uses the previous fit results as starting values. See Zhang X. et al., MRM 43:331 (2000) for a similar approach to “multiscale” fitting.

The METAFIT processing file contains the links to the three different FITT processing files used. There is no GUI for this program, so changes have to be done manually. The parameters used are described in the METAFIT.pdf document that is included in the MIDAS distribution, and some examples of the resultant images shown.

Publications

Many thanks to our collaborators for contributions to these recent reports:

M. Sabati, A.A. Maudsley. Fast and high-resolution quantitative mapping of tissue water content with full brain coverage for clinically-driven studies magnetic resonance imaging. Magn. Reson. Imag., 31(10):1752-9 (2013).

A.A. Maudsley, R.K. Gupta, R Stoyanova, N.A. Parra, B. Roy, S. Sheriff, N. Hussain, S. Behari, Mapping of glycine distributions in gliomas, AJNR, 35(6):S31-36 (2013).

X-Q. Ding, A.A. Maudsley, S. Sheriff, M. Sabati, P.R. Dellani, H. Lanfermann. Reproducibility and reliability of short-TE whole brain MR spectroscopic imaging of human brain at 3T. Magn. Reson. Med. Published online Mar 26 (2014).

A. Parra, A. A. Maudsley, R. K. Gupta, M.D., F. Ishkanian, K. Huang, G. Walker, K. Padgett, B. Roy, J. Panoff, A. Markoe, R. Stoyanova. Volumetric spectroscopic imaging of GBM radiation treatment volumes. Int. J.Rad. Onc., Biol., Phys. Published online, July (2014).

AA Maudsley, V Govind, B Levin, G Saigal, L Harris, and S Sheriff. Distributions of MR diffusion and spectroscopy measures with traumatic brain injury. J. Neurotrauma, In Press, 2014.

A. Lecocq, Y. Le Fur, A.A Maudsley, A. Le Troter, S. Sheriff, M. Sabati, P.J. Cozzone, M. Guye, J.P. Ranjeva. Whole-brain quantitative mapping of metabolites using short echo 3D-proton- MRSI. J. Magn. Reson. Imag. Pending final revision. 2014.

M. Sabati, S. Sheriff, M. Gu, J. Wei, H. Zhu, P. B. Barker, D. M. Spielman, J. R. Alger, A. A. Maudsley .Multi-Vendor Implementation and Comparison of Volumetric Whole-Brain Echo-Planar MR Spectroscopic Imaging. Magn. Reson. Med., In Press. 2014.

 

More Midas

http://www.midastouchinstitute.com/mt/wp-content/themes/yoo_noble_wp/images/midastouch_logo_top_horizontal.png

A cancer treatment facility in south Florida http://www.midastouchinstitute.com/mt/

 

Andrew Maudsley, September 2014

Funding for the MIDAS project is provided by NIH grant R01 EB016064 and the initial development was done under R01 EB000822.

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).