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

MIDAS eNewsletter


MAY, 2012

New and Improved

MIDAS v2: Version 2, operating with IDL 8.1 and 64-bit, has been undergoing testing for a number of months and is considered sufficiently stable for use. Two known issues remain: 1) The inter-process communication for FITT2 will very occasionally hang. A fix has been implemented and is under evaluation.  2) The registration program fails on some specific image sizes, which has only been found for DTI images. A fix has been found and is currently being implemented.

VIEWER: This program now includes “Layouts” that define which image types are displayed and the position and size of all windows. These can be set as the default layout for a project, or saved and recalled as needed. Other changes include password protection for the remote viewing function; improvements to the DICOM write function; and a “Zoom Tool” function, which allows the same zoom and pan to be applied to all displayed images.

PRANA: Now includes some between-group comparisons, including difference between mean values from each group of subjects or t-test. The False discovery Rate calculation and minimum cluster size can be applied to the t-test result to indicate only those voxels with significant differences. Here’s an example:



Brain Atlas Labels: The brain atlas supplied with the MIDAS package is mapped to the simulated MRI obtained from the Montreal Neurological institute (MNI). This was initially created using the neurological convention, meaning the “Left” region labels correspond to the left side of the image. Since the developers and the majority of users are in Radiology departments, in MIDAS v2 we have decided to make the MIDAS atlas correspond to the radiological convention, i.e. the Left region labels correspond to the right side of the image and the left side of the subject when using the standard SI reconstruction files supplied with MIDAS.

For comparison with other packages we note that SPM uses neurological convention and DTI Studio and FSL use radiological convention.


MIDAS Tips, Questions, and Answers

Inter-Series Registration:

The MSREG program aligns all image Series to the MRI_T1. This not only accounts for possible motion between the studies, but also aligns all image series from a second or later study to the T1 of the first study.  The SI-to-MRI registration generally performs very well, but on occasions small shifts (~half a SI voxel) have been observed. Here are a couple of pointers for checking this.

To check the SI-to-MRI registration, the water-reference SI image generally provides the best comparison. Either SID or the Viewer can be used, but SID offers two useful features. First, under the MRI/Options can be found an option to “Sum MRI Slices”. For the typical EPSI MR protocol, setting this to 5 or 7 mm will make the displayed MRI match the slice thickness of the SI. The second option is the “Shift MRI X, Y, Z Position”. By changing the relative position any mismatch between the MRI and SI data can be corrected. This can then be applied to the data, which means that the position vector for the SI data is corrected to match the settings.

When inter-Study registrations are done, all MRIs of the later studies will be aligned to the T1 MRI of the first study.  In this case it is recommended that any operations on the MRI data, namely image segmentation and T1 mapping (TIMO), be done on the acquired data, since any registration involves some (small) image smoothing. An additional reason to run segmentation before MSREG is that the program will save a bias-corrected version of the T1, which will help with the registration.

Remote Viewing of MIDAS Data: 

Several complete MRSI studies from the University of Miami, including 173 normal subject datasets can be seen from any networked computer, using the Java based MIDAS Viewer.  This is a good way to show some examples to your colleagues. The Miami server is password protected, so please contact the MIDAS developers to get access. To enable the remote viewing select the menu option FilesàServer. This shows by default three options, with the “Local” option being the normal operation. The “” option to access the data at the University of Miami site.  You can also set up the MIDAS server on your own computers, and this is described in the file \Midas\Documents\Help_files\MIDASSERVER_help.pdf.


Reduced k-space EPSI:

A version of the EPSI sequence with reduced k-space acquisition and GRAPPA reconstruction is now available, and operating on 3T instruments from GE, Siemens, and Philips. This obtains data with the same image matrix as before in 15 minutes. The raw data size is also reduced. With the 8-channel coil and TE=70 ms the loss in SNR is acceptable, but 32-channel acquisition is recommended. This is also being used with our short-TE acquisition.


We now have a short-TE version of the EPSI sequence that has TE=17.6 ms.  Here’s an example:

The data processing for this acquisition is still under development.

Shimming Tips:

Our implementation of the EPSI sequence almost always uses only the standard automatic shimming procedure. Even with this simple setup good quality spectra are obtained over 70% of the cerebrum, which is in part aided by the high-resolution sampling scheme used in the EPSI sequence (See Ebel et al. Magn. Reson. Imag. 21: 113, 2003). There are a couple of tips that have been found to help shimming for human subjects. Firstly, place the center of the EPSI FOV (landmark) at the eyebrows level. This places the center of the FOV at the magnet iso-center for most subjects. A backward tilt of the head improves the overall B0 field homogeneity (See Tyszka et al, JMR, 159, 2002). The EPSI FOV is then positioned at a 15-deg angle to the AC-PC line, which has been observed to give better quality data. On the Siemens Trio an overall line width of about 20 Hz can be achieved using this setup.

Developer’s Corner

Programming Using the MIDAS Library:

The XML scheme used to describe all data parameters is a hierarchical scheme, based on Project; Subject; Study; Series; DataSet; Data; Frame. To access this information a set of library functions are provided that are written in Java, which can be called from most other programming languages, including IDL. These use object-oriented programming conventions. There is also an equivalent library written in Python. To navigate through the system a set of Unique Identifiers (UIDs) are defined at each level of the hierarchy, and additionally the image series and data frames are identified with a label, e.g. “MRI_T1”, or “NAcetylaspartate”. When writing a program to read the data parameters it is a simple matter to navigate through the UIDs and labels, starting with information provided from the MIDAS browser or on the command line, which always includes the SubjectID, and optionally the StudyID, and SeriesID.  Here’s an example:

; Get the UID of all Studies for this Subject. We will only use the first, i.e. StudyID[0].

; The “MIDASLib” is the MIDAS library object, which has already been obtained.

StudyID = MIDASLib->GetStudiesGivenSubjectID(SubjectID[0])


; Some parameters are located in the Study node

B0 = LibRef->GetParameterGivenID(StudyID[0], 'Magnetic_Field_Strength', 1)


; Now get the UID of the SI/Spectral data by stepping down the nodes in order.

SeriesID = MIDASLib->GetSeriesGivenStudyID(StudyID[0], 'SI')

DataSetID = LibRef->GetProcessIDGivenProcessLabels(SeriesID[0], 'Spectral')

DataID = LibRef->GetDataGivenID(DatasetID[0])


; We can now read all parameters for this data into a 2xN ASCII array [name, value]. Note the transpose

; is needed to move data between Java and IDL.

Params = transpose(LibRef->ReturnAllParameters(DataID[0]))

There are of course many additional steps needed, including handling errors, and more information is provided in the MIDAS documentation.  A full example of the programming can be found in the MIDAS\Documents\Development directory, and the library functions are listed in MidasLibrary_MethodsByCategory.pdf.  Detailed information on the XML scheme can be found in the file …\Help_files\MIDAS Project Description.pdf. Perhaps the best way to write a new procedure is to take an existing IDL program and modify it as needed. Contact the MIDAS developers for pointers to selecting a program.

Spectral Fitting:

The prior information used in the FITT program is obtained by spectral simulation, based upon in vitro measurement of chemical shift and J couplings1,2. For this purpose we have previously developed and used the GAVA program3, and distributed this to accompany the MIDAS software; however, we now have a better option, VESPA Simulation, that has been developed by Dr. Brian Soher. It is simpler to install, offers additional functions, and is written in Python so doesn’t require a software be purchased. We recommend you check it out, at

1. Young et al. Magn. Reson. Med. 40, 812, 1998.   2) Govindaraju et al. Magn. Reson. Med. 39, 1011, 1998.  3) Soher et al. J. Magn. Reson. 185: 291, 2007.


Other News


The NIH bioengineering partnership grant that has supported the MIDAS development (R01EB000822) is coming to an end after ten productive years, in March 2013.  There are currently no plans to continue the partnership grant, although a proposal to continue just the MIDAS software development portion of the project has been submitted. I would like to thank all of those that provided their support for this application. There are also a couple of grant applications from other sites that plan to include the EPSI acquisition and MIDAS processing for clinical studies, and of course we hope they all get funded.  We are also continuing to look for additional projects, worldwide. The EPSI/MIDAS development now has users as far afield as India, Indonesia, France, Germany, U.K., Poland, and of course the USA.


In the paper by Govind et al. we demonstrate atlas-based analysis of the cortical spinal tracts: V. Govind, K. Sharma, A. A. Maudsley, K. L. Arheart, G. Saigal, and S. Sheriff. Comprehensive evaluation of 1H MR-observed brain metabolites of the corticospinal tract in amyotrophic lateral sclerosis. PLoS ONE,7(4):e35607 (2012).

Another paper came out this year that demonstrated the potential for the PRANA data analysis program to easily map parameters that would otherwise take considerable effort. In this case we show that MRS results should also take into account body weight:  A.A. Maudsley, V. Govind, and K. L. Arheart. Associations of age, gender, and body mass with MR-observed brain metabolites and tissue distributions. NMR in Biomed. 25(4):580-93 (2012). 


More Midas

Description: Description: Description: Description: Description: cid:A43F3752-3D29-4191-808D-48BF285FF54E

From: Cox D. et al. Integrins as therapeutic targets: lessons and opportunities. Nature Reviews Drug Discovery 9804-820 (2010).


Andrew Maudsley, May 2012

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 (