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

MIDAS eNewsletter


JUNE, 2017


New and Improved


SID:  The SID display program dates back over 20 years. It was created in a time when computers only came with 16-bit CPUs, had RAM measured in Mb, grey scale display, and MRSI only came in single slice. The programming used IDL code that has since become obsolete, and although it’s been tinkered with over the years, the underlying program design didn’t change much. Well finally, SID has been upgraded with features such as 24-bit color and several new features. MRI and metabolite maps can now be displayed in axial, coronal and sagittal orientations, and each can have independent color. The image value readout function now works with 3D volume data, and ROI options include the ability to delete a previously selected voxels (use the Ctrl key) and saving ROIs to a file. Major new functions are image editing, which is primarily meant to modify Quality_Maps but can be applied to any map, and displaying blended MRI and metabolite maps. Interactivity is also improved with the use of the mouse wheel to scroll through slices, window level and width using the right mouse button, and zooming in on a part of the spectral display by swiping the left mouse across the plot window. Other options include a readout of image values in each image window, color scales, and copying images to the clipboard, and blended images (see example). SID 3.0 is now available in the latest distribution.

 MAPPER: The MAPPER program imports Analyze format images created from other sources. Example uses of this program include tissue segmentations or user-defined masks done outside of MIDAS. This is a useful support function for doing ROI analyses in the MINT program.

The example to the right shows the import of a tumor segmentation, indicating regions of necrosis, contrast-enhancement, and FLAIR hyperintensity. These three ROIs were saved in an indexed Analyze format file, which was then opened in the MAPPER program where the orientation can be checked, the three tissue type labels assigned, and the data then imported into MIDAS. The visualization allows the user to confirm orientation (a flip in Y is usually needed when importing Analyze). The program can also be used to visualize the atlas files.

SINorm: This program applies the metabolite image normalization and creates metabolite ratio maps. Options have now been added to normalize images using the mean values in normal-appearing white matter (NAWM) of either the same parameter or the tCreatine signal. For example, the first option could create a map of tChoTumor/tChoNAWM, while the second would result in a map of tChoTumor/tCreNAWM. The program automatically locates the largest contiguous white-matter volume from which the mean NAWM values are calculated. For studies of brain lesions, if a tissue segmentation map marking the lesion location (the “Other” map) is available then this will be used to help locate the normal-appearing brain region, but even without this tissue map the program appears to be quite robust for finding the correct hemisphere. See the updated SINorm help document for a description of this option and the naming convention for the resultant maps.

SPM Segmentation: The default segmentation program used in MIDAS is the FSL/FAST program version 4.1. This implementation does not support the option to use a prior probability map and as a result some deep grey matter structures tend to get incorrectly classified. To provide an alternative segmentation method we have now added support for SPM. This requires that SPM is already installed on your computer, though this can also be run in standalone mode without having Matlab installed.

MSREG: This program applies rigid registrations to ensure all image series are aligned. It corrects for subject motion and differences in the angulation of any acquisitions. With sagittal MPRAGE acquisitions we have found the registrations to be less reliable and the TRUNCATENECK function was introduced to improve this, as discussed in the last newsletter. We have now found that running the MSREG program twice is more robust. The could previously be done from the GUI, but we have now modified the program to allow this to be done in BATCH mode. It requires the keyword REPROCESS to be present on the command line, so the additional BATCH instruction would be:

MSREG           0          1          none    1          none    reprocess        -1,0,0

An additional problem we have encountered with sagittal acquisitions is that the MSREG program is wrapping the SI data in the z direction, so as to cut off the top of the head. This is discussed in more detail below.


MIDAS Tips, Questions, and Answers


DICOM Support: There are three ways to export images to DICOM. The first is in the MIDAS Viewer program, from “Tools->Export as Dicom” or “Send to Dicom Receiver”. The second is using a utility function called DICOMExport, which is accessed from the tools section of the toolbar. This utility requires a separate IDL DICOM Network Services license, so it is not included with the main distribution, but is available on request. The third is in the IMCALC program, under the DICOM Export Tab.

Most PACS systems require that the data be identified, i.e. with subject name etc. However, MIDAS processing is commonly performed for research studies where the data has been anonymized. In this situation, to create the DICOM files with full subject identifiers the DICOMExport function provides an option for using an existing non-anonymized DICOM file, from which the necessary header information can be copied.

Remote Viewing: A little-used, though useful, feature of the MIDAS Viewer is an option to view data from a remote location using the web. This requires that a MIDAS Server be set up that uses Apache Tomcat, and that a MIDAS user account be set up. The web address for that account is then defined in the file \MIDAS\, and the account login information is defined in the file This is described in the document MIDASSERVER.pdf.

Handling Sagittal T1 MRI Acquisitions: When the T1-weighted MRI image is acquired in sagittal orientation, as commonly done with the MPRAGE sequence, the FOV along the Head->Foot direction is typically ~256 mm, which is much larger than he brain, which is typically ~140-160 mm. The FOV in Z may then span from the top of the skull all the way down to the base of the neck and sometimes to the shoulders. This extra coverage below the brain creates a couple of different issues inside MIDAS registration:

1)         During tissue segmentation, the BET program incorrectly includes sections below the brain (neck, etc) as part of its brain mask, and tissue segmentation is subsequently performed on this non-brain part of the image.

2)         In MSREG, this can lead to cutting off of the top of the brain in SI and SI_Ref data. This is because MSREG both aligns and co-centers the floating images (SI, SI_Ref) to the reference image (MRI_T1). In this case, the center of the MRI_T1 is close to the bottom the brain, and the SI data is then centered around that location. As MSREG does not change the resolution or FOV of the floating images, the registered SI data will still be 180 mm FOV in F>H, but now centered around the bottom of the brain. Therefore, the top of the brain will be cut off in SI and SI_Ref data after MSREG.

There are two solutions to this issue:

1)         Use the TRUNCATENECK program as part of the processing pipeline in Batch. Place this program right after the MRI_T1 volumizer step. This will, as the name indicates, reduce the slices of the MRI_T1, so that the neck slices are no longer part of the MRI_T1 data. The MRI_T1 is then centered in the middle of the brain, which solves the MSREG issue, and as there is no neck data included, BET will find the correct brain mask.

2)         For future acquisitions, for the sagitally acquired MRI_T1, reduce the FOV in the F>>H direction to ~160-180mm, and center the slab at the AC-PC line. This should leave several axial slices above the head (see example at the top of this newsletter). It is preferable to co-center and co-align all MRIs and the EPSI slab. If you reduce the FOV, then you do not need to use the TRUNCATENECK program, though this should have no effect if left in.


Developer’s Corner


Brain Atlases: The MIDAS processing pipeline includes non-linear registration to a standard spatial reference, which can be applied to metabolite maps as well as any MRI data. The default target image for the registration that is provided in the MIDAS distribution is the Montreal Neurological Institute (MNI) BrainWeb distribution, which is mapped to a lobar atlas (developed by Colin Studholme) that marks the temporal, parietal, occipital, and frontal lobes and the cerebellum; however, any other target image and atlas can be used. To use a different target and atlas the following steps are needed: First, an XML file must be created that describes the target image and the corresponding atlas. An example of this can be found with the default atlas, which is in the file \MNI-Atlas\LobarAtlas.xml. The target image can be of any spatial resolution and this XML file can contain entries for multiple resolutions, but the default that has been found to be useful for the volumetric EPSI data is 2 mm isotropic resolution. Once this file is created the path to this file must be set up the Project that will be using it. This is done using the “Edit” function in the MIDAS Importer, where you can then edit the Atlas Path:

Running the Registration process will then use the new target image. Analyses done in this atlas space can then be done using the PRANA program.

Uses for the Brain Atlas in Subject Space: The PRANA program includes an option for getting a brain atlas into subject space, and was described in our September 2015 newsletter. Here we mention a couple of examples where this can be used.

The atlas is an indexed format; so for example, the left frontal lobe in the example shown on the left has a value 2. In the “Read Image Values” function of the SID program you can use the “Read Image Values/ROI Averaging” function to integrate all spectra over this region by looking for voxels that meet the criterion of having value 2.0. This spectral integration can also be done in the MINT program, which includes additional options and enables multiple individual regions to be processed and analyzed in one operation.  Similarly, in the IMCALC program you could either select or exclude voxels based on an atlas region value. One example of this is to obtain a mean value over the cerebrum, while excluding the cerebellum.


Example Results


A New Subject Exclusion Criterion?: A study carried out in a healthy control subject revealed an unusual signal that could be seen throughout the brain and ventricles. Our initial concern was that this EthanolSpectrumsubject was hypoxic from sleep apnea (the subject slept during the scan) and that this was from the doublet resonance of lactate, which at TE=70 ms appears as an out-of-phase pseudo-triplet. However, the pattern is inverted relative to that of lactate and the ppm offset was incorrect, at ~1.17 ppm. 


After some puzzling, we realized this was from the triplet from the CH3 group of ethanol. The subject had enjoyed a drink or two during lunch, which was just before the scan. The fitted maps show that the ethanol is distributed throughout the brain:

Using a three-compartment tissue regression analysis, the relative concentrations in CSF, gray matter, and white matter, as a mean value over the cerebrum, were 1.0, 1.3, and 0.9. This differs from an earlier report of this measurement (Spielman et al., Alchol Clin Exper Res, 17:1072 (1993)) that reported 1.0, 0.72, and 0.37. Taking a ratio to creatine provides mean concentration estimates of 8.9 mM, 8.1 mM, and 6.7 mM (based on relative Cr concentrations reported in Malucelli et al, NMR Biomed, 2009) in CSF, gray, and white-matter.


More Midas


Nike MIDAS Touch Kicks




Congratulations and many thanks to our collaborators for these recent reports:

·        Lopez CJ, Nagornaya N, Parra NA, Kwon D, Ishkanian F, Markoe AM, Maudsley A, Stoyanova R. Association of Radiomics and Metabolic Tumor Volumes in Radiation Treatment of Glioblastoma Multiforme, Int J Radiat Oncol Biol Phys. Mar 1;97(3):586-595 (2017).

·        Ding X-Q, Maudsley AA, Schweiger U, Lichtinghagen R, Bleich S, Lanfermann H, Kahl KG. Effects of a 72 hours fasting on brain metabolism in healthy women studied in vivo with magnetic resonance spectroscopic imaging. J Cereb. Blood Flow Metab. Epub Jan 1. (2017).

·        EL. Dennis, JR. Alger, T. Babikian, F. Rashid, JE. Villalon-Reina, R. Mink, C. Babbitt, JL. Johnson, CC. Giza, RF. Asarnow, PM. Thompson. Tract-based spectroscopy to investigate pediatric brain trauma. Proc. SPIE 10160, 12th International Symposium on Medical Information Processing and Analysis, 1016019 (January 27, 2017).

·        Maudsley AA, Goryawala MZ, Sheriff S. Effects of tissue susceptibility on brain temperature mapping. Neuroimage. 146:1093-1101 (2017).

·        Cordova JS, Kandula S, Gurbani S, Zhong J, Tejani M, Kayode O, Patel K, Prabhu R, Schreibmann E, Crocker I, Holder CA, Shim H, Shu HK. Simulating the Effect of Spectroscopic MRI as a Metric for Radiation Therapy Planning in Patients with Glioblastoma. Tomography. 2(4):366-373 (2016).

Please let me know of additional publications that make use of MIDAS.


Andrew Maudsley, June 2017

This newsletter provides information to the developers and users of the MIDAS software package. Please let me know If you would prefer not to receive notification of these reports (