MIDAS eNewsletter




This issue of the MIDAS Newsletter comes after a long break that included the end of the previous grant period and a transition to a new phase of the project. We now welcome new partnerships with Peter Barker (Johns Hopkins), Dan Spielman (Stanford), and Brian Soher (Duke), and their associates. These new collaborator projects will be extending the volumetric MRSI methods to MR instrumentation from multiple-vendors, as well as continued improvement of the software methods.

MRMcover.jpgA figure from this project made the cover of MRM in the March 2009 issue and this article generated some interest and, we hope, some new users. The article described the whole-brain metabolite mapping, which provides the normative information for comparison against individual subject data.

A part of the figure in the article is shown here:  BrainMappingFigure.jpg

New and Improved

MIDAS Viewer and Importer: Many functional and performance improvements have been made to these two modules, thanks to JC and Jeff Alger. There is a substantial performance improvement for the importer. Users are asked to check out the functionality of the Viewer and provide feedback that may help us continue to improve it, with the aim of this providing the interface to the clinical researcher. The Viewer now includes remote display capability, allowing for example, large datasets in the central database to be viewed from a remote networked location.

EPSI3D: This is the program that takes care of the data resampling to account for sampling during the gradient ramp period, as well as doing the interlaced-Fourier transform reconstruction that accounts for the fact that the ky-t trajectory is not acquired on a Cartesian grid. When the program was initially written we only had single-channel receivers and performance was of no concern, but with the move to 32-channel data this has become more important. Several changes have been implemented and the next release will include a version that is much faster.

PRANA: This is new module that provides functions for reviewing images from multiple studies and some analysis functions. It is integrated with the brain atlas and includes functions for analysis by metabolite values by brain region as well as voxel-based analyses. Below is an example of a multi-subject display. The image shown above of the average metabolite distributions from a group of subjects was also obtained using this program.


Release Packaging: The MIDAS package consist of multiple components, including program modules written in IDL, Java, and C, documentation, data files, and many different files used in the processing, and putting all this together can be quite a challenge. We have had an automated build that supports the multiple IDL versions for some time, but over time the computer systems supported have changed and distributing the package has become more involved. Rajesh and JC have done great work on updating our release procedures, to include a greater level of automation and reliability.

MIDAS Web: Our trusty web server has had to go, forced out by our IT support team that doesn’t like even good computers once they pass their warranty period. This has prompted us to update our source code control and bug reporting systems, and at the same time starting a new version of the web site. Rajesh has done a great job in getting a new web site up. The new link is: http://mrir.med.miami.edu:8000/midas

You can register for a login directly on the site, but also need to send a request to Rajesh to get access to the file download section of system. It’s still under development and it will take a while to get everything in there. This is a Wiki based system and everyone is encouraged to help us develop the site.

Bug reporting is done by submitting a “New Ticket”, and if you want to get access to the source code section you need to get a separate account for that from Rajesh.


A report on the inter-subject reproducibility of the EPSI+MIDAS processing has hit the Early View section of NMR in Biomed. Bottom line is that the reproducibility compares well against many previous single voxel studies, which were mainly done at 1.5T. The figure below summarizes the distributions of COVs taken over all voxels covering approximately 70% of the cerebrum. 




Why is the Choline/Creatine value from MIDAS different from the result obtained with my other spectral analysis program?


The FITT program returns values based on concentration, not signal integral. Since the choline singlet at 3.18 ppm has 9 protons, (CH3)3, and creatine at 3.02 ppm has 3 protons, -(CH3), the signal integral for choline is three times larger for the same concentration as creatine.



The MIDAS package looks complex, what’s the best way to start using it?


There are two documents included in the distribution and also available on the web site, the “MIDAS_Processing_Overview” and “MIDAS_Tutorial”. The tutorial takes you through processing of a small 2D SI dataset, which must first be downloaded from the web site (see under “Downloads”). Getting a sample data set to go through the processing for a 3D dataset is more challenging, with data sizes of 10.7 Gb for an 8-channel acquisition, but we can arrange for you to get one if necessary.



With one of my datasets, when I select the study in the browser, the program hangs and displays a message saying “Refreshing Browser”. What’s going on?


It is likely that a processing step was interrupted while updating the parameter information in the subject.xml file. This can be fixed by deleting a file named “subject.xml.lock” in the top-level directory of that subject. This file is used to synchronize access when multiple processes are trying to update the parameter file at the same time.

EPSI Sequence

EPSI on GE and Philips: Preliminary versions of the EPSI sequence are now running on the GE and Philips 3T instruments, and thanks go to Meng Gu and Juan Wei for those efforts. Development is still ongoing, but we should soon be able to examine consistency across instruments very soon.

EPSI on Siemens: We now have a version that supports 32-channel acquisition and removes the requirement for extra memory in the acquisition computer for even smaller numbers of receiver channels. The challenge here was handling raw data sizes of 43Gb!  Sulaiman has implemented a method that transfers the data to the host computer during acquisition.

Developer’s Corner

The new source-code control system uses TortoiseSVN, with the cute logo: SVNlogo.png. Go to http://tortoisesvn.net/downloads to check it out.

More Midas



Andrew Maudsley

October 2009

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).