FAQ - The EPSI sequence and MIDAS software

What is EPSI and MIDAS?

MIDAS provides data processing and analysis functions for the volumetric proton EPSI acquisition developed at the University of Miami. These two packages are generally distributed together.

Is EPSI+MIDAS useful for my application?

The volumetric EPSI acquisition aims to sample a large fraction of the brain with relatively good spatial resolution and the MIDAS processing is optimized for this type of data. It samples a larger brain volume than vendor-provided implementations of SI; however, there are some trade-offs to this approach which means it may not be the best for all applications.
While MIDAS has many novel processing features and could be adapted for other MRSI variants, it is not anticipated that it would provide any advantage over the vendor-provided reconstruction for the vendor-provided SI acquisitions. In addition, adapting it to other MRSI acquisitions may require additional support from the developers. It is therefore only recommended for processing the EPSI data.

Can EPSI/MIDAS be used for clinical studies?

Only with appropriate institutional approvals. These are provided for investigational studies only.

How do I get the EPSI sequence?

This sequence is available for 3T instruments from Siemens. For VE11C software version (Skyra and Prisma) the EPSI sequence is distributed as WIP 814. Contact your local Siemens collaboration scientist for distribution. For other software versions the sequence is distributed as a C2P, which requires that the user obtains a letter from Siemens authorizing distribution of the EPSI sequence from the University of Miami. The C2P distribution also requires a materials transfer agreement with the University of Miami, which can be downloaded from the link below. Not all software variants are supported but source code can be provided for recompilation at your site (with approval from Siemens for distribution of source code).

What is included in the latest version of the EPSI sequence?

The EPSI sequence for VE11C includes real-time frequency adjustment and spatial reconstruction on the scanner. The on-line reconstruction removes the problem of dealing with the large raw data sizes; however, there are limitations depending on the available scanner memory and data sizes. Contact the developers for more details. Spectral reconstruction and fitting is still done off-line and a different processing pipeline is made available for this.

This implementation of EPSI samples the whole brain. Isn't this a problem for MRS?

  • The standard setup uses automatic shimming, as for any MRI sequence, although there can be a benefit to using some additional manual shimming adjustment.
  • The sampled volume includes a large fraction of the brain, including the cortical surface, but there will always be regions where MRS data cannot be obtained. Depending on the scanner, between 60 to 70% of the brain can generally be sampled.
  • There are several trade-offs to the volumetric SI approach, however, the method has been found to have a reliability and data quality equivalent to or better than other SI implementations. The methods used to sample a large fraction of the brain involve a trade-off with sensitivity so, for example, it may not be the best method for measuring glutamate in small volumes. The use of reduced phase encoding and the long acquisition time (10 to 20 minutes depending on resolution) also increases sensitivity to subject motion.

How do I get the MIDAS software?

MIDAS is available to academic research sites at no cost and is distributed under the GNU General Public License. To obtain the software you must read our license agreement and send an email to the MIDAS team (amaudsley at Miami dot edu) that includes a statement that you agree with the terms of the agreement and are requesting a login. You must provide an institutional email address. An account will then be created for you that enables access to the download page. It would also be useful if you could tell us which MR instrument you will be acquiring data on.
The package is provided in a self-extracting zip file that installs easily on Windows PCs. See the installation manual in the documents page for additional information. It is no longer being actively developed, though bug fixing is ongoing and it is recommended that you update your installation periodically to get the latest version. Example processing files and sample data used in the tutorials and also available for download.

What is needed to run MIDAS?

MIDAS requires IDL running under Windows OS on 64-bit CPUs. It can also be run under Linux using a Virtual Machine with Windows OS. IDL can be purchased from Harris Geospatial Solutions. Prices are very reasonable for academic sites. There is also a run-time RT license that offers a lower cost for non-academic sites.
It is also possible to use the free IDL Virtual (VM) license, but there are two limitations: 1) multicore processing is not supported. The speed improvement with a full IDL license on a multi-core CPU is substantial for the spectral fitting. 2) The IDL log window is not shown (also the case with the RT license), meaning that informational messages and the occasional unhandled error messages cannot be seen. You can download the Virtual Machine version from Harris. This is free, but requires that you set up an account with them.
With the full IDL license the spectral fitting will go much faster with multi-core CPUs. A minimum of 6 Gb RAM is required and up to ~32Gb will help with multi-core CPUs and processing multichannel data. A high resolution display is essential and a large monitor is highly recommended. A high speed disk I/O has also been found to be useful to handle the large datasets.

Can I add my processing/display/analysis module to MIDAS?

The processing pipeline is highly configurable and can run IDL procedures or any other program from a command line call. The parameter information is maintained using a XML format so some software development may be needed, but examples are available to access this using Java, Python, or IDL.

Can I get the source code?

Yes, for MIDAS, all IDL source code is provided as a separate download from the web site to registered users. It is recommended that you contact the developers to be sure you have the latest version. Access to the Java code can also be provided on request. As per the license agreement, modifications to the source code should be made available to include in the distribution. For the EPSI sequence the source code is available following an additional agreement from Siemens.

What are some of the limitations of MIDAS?

  • This is a large package that requires time and effort to fully understand. However, if you stay with the standard methods, for which an automated processing pipeline is available to download, then most users are up and running quickly.
  • The MRSI reconstruction is done off-line from the scanner. If the spatial reconstruction is not done on the scanner then the raw SI data has to be transferred to a local computer for processing. These are large datasets, e.g. 42 Gb for 32-channel detection.
  • For multi-channel acquisitions the data processing can take up to two hours, and longer if you run the VM version of IDL.
  • Integration with PACS systems is not included in the standard distribution, but can be provided using a module that requires the IDL DICOM license. Contact the MIDAS team for details. It is also possible to export metabolite images in DICOM format.

How do I get started using MIDAS?

  • Extensive documentation is available with the distribution (look under Documents subdirectory) and on this web site.
  • Start with the tutorial, which uses a small 2D SI dataset that can be downloaded from the web site. When you have gone through that there is another example using a volumetric SI dataset that can be downloaded from the FTP site. Contact the MIDAS team for the FTP login information.
  • You should start with the automated processing pipelines that are provided on the web site, but as you gain experience you may then want to modify these processing steps to suit your protocols.
  • Instruction can be arranged. An informal course is available that takes roughly two-days. This is outlined in this document .

What about long-term support?

The NIH support under which the MIDAS and EPSI developments were done is now finished. Under other funding the EPSI sequence will be maintained for newer software versions of Siemens scanners until approximately 2026. Options for long-term support include obtaining all source code for your own in-house efforts, or contractual arrangements with the group at the University of Miami.

How do I cite MIDAS?

There are two main reports that describe the MIDAS software:

  • AA Maudsley, C Domenig, V Govindaraju, A Darkazanli, C Studholme, K Arheart, and C Bloomer. Mapping of brain metabolite distributions by volumetric proton MRSI. Magn. Reson. Med. 61: 548-559 (2009).
  • M Sabati, S Sheriff, M Gu, J Wei, H Zhu, PB Barker, DM Spielman, JR Alger, AA Maudsley. Multi-Vendor Implementation and Comparison of Volumetric Whole-Brain Echo-Planar MR Spectroscopic Imaging. Magn. Reson. Med., 74(5):1209-20 (2014).

Can I get some existing MRSI data?

Yes, all data taken at this site is made available. We have over 190 normal control datasets for TE=70 ms and additional data is being acquired for TE=17 ms. Studies from several clinical projects can also be made available, including TBI, brain cancer, and epilepsy. If you want to have access to the patient studies or to do more than just take a quick look at an example dataset a data use agreement is available.

EPSI Software Agreement

The following agreement is needed for distributions of the EPSI sequence using the C2P method. This must be signed by an authorized institutional official (not the researcher).

We will not consider any modifications to the agreement. It must be accepted as is. The signed agreements can be sent electronically to the University of Miami Technology Transfer Office, at mta.ott@… and cc'd to CDomenig@…