Diffusion Tensor Imaging - Science method

The Camino toolbox is a good place to start at UCL. Camino | UCL Camino Diffusion MRI Toolkit. http://camino.cs.ucl.ac.uk/

It has a easy syntax to get started and runs on a wide range of platforms.

@Linda Hoffman You can read the User Guide on the website of TBSS to get other scalars (MD, AD RD):

Using non-FA Images in TBSS

It is straightforward to apply TBSS to other diffusion-derived data than FA images. For example, you may be interested in how MD (mean diffusivity) or the first diffusion tensor eigenvalue varies between different subjects.

To achieve this we recommend using the FA images to achieve the nonlinear registration and skeletonisation stages, and also to estimate the projection vectors from each individual subject onto the mean FA skeleton. The nonlinear warps and skeleton projection can then also be applied to other images such as the second eigenvalue. The following instructions assume that you want to run TBSS on the second eigenvalue, named L2 by dtifit:

Hi Bassam,

so there is 1 b0 volume in each phase encoding direction - hopefully 180° opposing? Then just calculate the topup distortion field with these 2 b0s and feed the result into eddy. Here you would use all of the data (b0_PED1+bXXX_PED1+b0_PED2+bXXX_PED2) of both PEDs concatenated into one file…

What you are looking for is the standard case for topup&eddy and well described on their user guide website.

Best,

Martin

FDR is more robust.

Here's the answer from the FSL list.

hello Ziba Arghiani

i think that the MRIconvert is the best tool

Have a look at my Handbook (http://jeromemallershandbookofstructuralbrainmrianalysis.yolasite.com/?_sm_au_=iWHrSsBDn5JWbQVVML8tvK34L00HF ), Part 1, from page 61 (fslmaths scripts to extract the ROIs to make them masks then fslmeants command to extract the values within each ROI). :-)

Jerome

I suggest that you first denoise your data: this is very important. Easy to do with MRtrix3 with the 'dwidenoise' command.

Then get rid of field inhomogeneities with the MRtrix 'dwibiascorrect' command.

Then get rid of Gibbs ringing with the MRtrix 'mrdegibbs' command.

Then you can correct for eddy and movement artefacts with FSL's 'eddy' module.

Then, you will be ready to conduct analyses such as tractography.

After whole brain tractography, then you can filter the tracts to get more of the 'true' tracts using the MRtrix 'sift' module.

-Jerome

Hi

This article may help you:

https://www.frontiersin.org/articles/10.3389/fnins.2013.00004/full

There is a huge gap between the paper and landing of AI on medical. The fact is that some paper just publish toy model with off-the-shelf tool (all kinds of CNN model and opencv) to run the accuracy and metrics without understanding of the details. These models are hard to extend to real scenarios of images with local complex ROI. Resort to some top conference like CVPR and MICCAI as well as some top journals in medical diagnosis. There is also standard open dataset online with order of 10GB and 100GB. If your model can handle that, then your research make sense.

Dear Vincent Adegoke :

Building on Shafagat Mahmudova answer, shallow neural networks have at most 3 layers (input, hidden layer, and output layer). For complex problems, the hidden layer was "fat", having many neurons. This was because more layers were not computationally feasible. Deep neural networks appeared when computers became more powerful.They added more layers instead of making the one hidden layer fatter.

It is important to note that there is not one size fit all approach with DNN to maximize prediction accuracy. By trial and error, or by the use of optimization techniques (like genetic algorithms (GA)), you can identify the correct meta-parameters of your DNN model.

Hi. Are you using the Windows version of dcm2niigui? If yes, then this may be your issue.

I suggest you use the Linux (or Mac) version of dcm2nii, or even better is dcm2niix. If you still have problems, then I suggest you use MRtrix3 (specifically the command called mrconvert).

How many dicoms are in the PPMI DTI files? If you know how many volumes there are supposed to be (number of diffusion directions plus number of bzero volumes), and multiply that by the number of slices per volume, then that will equal the total number of slices (dicoms).

Jerome

Is DTI definitely the model you are trying to fit to the diffusion images? I ask because typically DTI is fit to b-values closer to 1000. Many models/methods will be able to handle a difference of 50 in bvals without much extra consideration.

Have you tried with DWI sequence.

If the damage is in Gray matter, then you can go for FDG-PET.

Thanks

I did partial correlation analysis controlling for age, gender, and education level. And git some results. For example we found attention is correlates with anterior thalamic radiation. But do I need to do anything else after that? Do I need correction because I did manu correlation or regression analysis?

Dear Guevara,

Thank you for your reply, unfortunately there is no MRI multi modal images with DTI with tumors ground truth in NITRC Image Repository .

Thanks for your answers everyone.

Chen, how was your experience with the fiber tracking module on DKE? I didn't get a good result off of it.

Hi,

could/should add different sites as (dummy-coded) covariates in any regression analysis you perform and of course include/discuss as limitations.

You could further assess whole brain similarity of FA values e.g. compared to an FA template (relative values across brain regions) or across sites (absolute voxel-wise values).

Hope this helps & good luck!

Well I'll be ... thank you sir. There is a lot to like there. Cheers, -Morgan

Sahil,

I'm not sure if this may be helpful for you, but if you have tested that correlations on Qdec you should check your a mc-z.abs.th??sig.ocn.annot file on your "corr" folder and open it on tksurfer.

once your results are visible, you can save and register each cluster as a label on each of your subjects, and then run the mri_anatomical_stats command to have the mean volume/thickness/whatever value of your interest

While TRACULA has some interesting and unique abilities, it is quite limited since it is confined to the tracts that have been pre-defined. This is largely related to the method in which tracts are automatically identified, which is also based on a "normal" brain; therefore, it is also not ideal for tracking in structurally abnormal brains.

You can certainly use freesurfer ROIs for seeds in many diffusion packages. It somewhat depends on what you are actually looking for in your research. Keep in mind that TBSS is very different from many other methods of tractography! 

For probablistic tractography, the FSL diffusion package is quite easy to use and allows use of freesurfer ROIs as seeds. If you are more interested in deterministic tractography, I have found DSI Studio to be great (also offers other methods of ODF modeling other than tensor, such as generalized q-sampling, that can give better results than DTI). MRTrix and DTI Studio are other popular diffusion packages.

You will need to determine which package best fits your needs and invest some time in learning the software from the available manuals/tutorials.

I don't know your complete set up (and what the subjects did inbetween). But if you used the same scanner and get significant results in part of the GM I would consider the data meaningful.

Noise could be a problem, as in general the signal intensity with higher b-values in the GM is lower than in the WM, but if you have enough signal in your raw data in the GM than it is probably neglectable.

Even in the GM you have fibers with a myelin sheet (just have a look at publications regarding MS plaques in GM --> demylination). Just the percentage of myelin is much less, therefore a change in FA could be meaningful.

A good place to start might be the Human Connectome Project (HCP) which releases very high quality structural connectivity data estimated from diffusion MRI ( http://www.humanconnectome.com)

I'd recommend going through the HCP workbench tutorial (https://www.humanconnectome.org/software/connectome-workbench.html) to get a feel for the data and the various cortical/subcortical parcellation schemes that can be used. I'm not aware of support for AAL in HCP, but it should be relatively straightforward to generate a CIFTI-format dlabel image from an AAL label image using workbench command-line tools.

Hi

I have tried in both ways. During pre-processing, I have deleted 2 volumes at one time and at 2nd time, I have taken all the volumes in the same subject. There is not a very big difference with or without deleting volumes.  

Hi Theo,

Interesting, but quite difficult question. Current Fibertracking algorithms only give you "the most probable" fiberBundle-prientation. In addition, the thalamus is a highly interconnected region, which makes it hard to isolate fibertracts and to see, whether they end in a different cortex location.

One approach would be to create ROIs in the cortex and thalamus and use SPMs StreamlineTracking with Global FiberTracking data (Reisert et al.). You can use Freiburgs DTI toolbox to extract the interconnecting fibertracks (between your ROIs). All these toolboxes are matlab based.

Hi Max,

Fixels from FODs have nothing to do with the diffusion tensor model.  The eigenvectors from a tensor are also not "main directions" of anything biophysical nor anatomical...  What you can get (per fixel) from an FOD, is the apparent fibre density (AFD); which does come with a reasonable piophysical interpretation.  It would be best to Google the latter term and do some reading first though.  If you've got any questions on how to work with MRtrix, we can help you out better over at our forums on http://community.mrtrix.org !

Cheers,

Thijs

I would use the participants own bvecs, particularly if you are doing tractography.

The bvecs file has the information about which way the diffusion gradients were pointing. The calculated directions of the eigenvectors (and hence the direction of the calculated tracts) depends on the direction of the gradients relative to the patient.

If there are only minor differences in the bvecs, then there will only be minor differences in the results. But you don't know that there are ALWAYS only minor differences. So to be on the safe side I would go with the patients bvecs.

I guess it would be interesting to find out why the differences exist. It might depend on the position/ angulation of the subjects head

Although, assuming it is the same MR sequence for everyone, the calculated MD and FA should be the same, since they are not telling you about specific directions. 

Thank you for your input!

Diffusion is a pretty noisy technique anyway (low signal to noise ratio).  From the acquisition standpoint you have noise coming from the pulse sequence used to collect it (typically a single-shot EPI), which can have spikes or distortions caused by sinuses and phase encoding direction.  Also the b-value itself can introduce noise, as the higher the b-value the lower the SNR for both GM and WM structures.  We try to compensate for some of these by collecting multiple averages of each direction...  but the noise is substantial, and that's just on the collection side of things.  

When it comes to actually estimating the tensors, the metric you are using (FA, ADC) can also have issues depending on a variety of factors impacting your data.  

Some useful papers: 

Soares, J. M., Marques, P., Alves, V., & Sousa, N. (2013).  A hitchhiker's guide to diffusion tensor imaging. Front Neurosci, http://dx.doi.org/10.3389/fnins.2013.00031

Jones, D. K., & Basser, P. J. (2004).  Squashing peanuts and smashing pumpkins”: How noise distorts diffusion-weighted MR data.  Magnetic Resonance in Medicine, 52(5), 979-933.  

Diffusivity of water in CSF is high (no diffusion barriers).

If both tissue and CSF are present in a voxel, this means that diffusivity will likely be increased compared to a voxel in which only tissue is present. 

Hi! Interesting question indeed. Maybe NODDI, an advanced diffusion technique developed at UCL, can be something to look at? 

In comparison to regular DTI measures such as FA and RD, you get measures of orientation dispersion (ODI) as well as isotropic diffusion from NODDI, this might be something to look into. Otherwise, maybe some quantitative T2 relaxometry techniques such as mcDESPOT or Myelin Water Imaging (MWI) could be something to look into. 

I don't have a specific answer regarding the symptoms but I am just thinking if axonal loss is the only reasonable interpretation. "Reductions in radial diffusivity (RD) have been interpreted as improved myelin but reductions in axial diffusivity (AD) involve other mechanisms, such as axonal density." Axonal loss would cause reduction in AD but there are quite a few confounding factors in DTI that convolute the interpretation from my perspective. For instance, wouldn't increased isotropic diffusion cause the same effect? 

I would really like to see the same type of study done with more advanced DWI such as NODDI, Diffusion Basis Spectrum Imaging or similar. That would probably shed more light on the underlying changes in the tissue micro structure. 

There are several procedures. One is to build a device to secure the front and hind legs and simultaneously tie the upper teeth with surgery thread. Also is necessary to added two lateral pillows  in order to reduce oscillatory and  involuntary movements

 A quite late answer, if you want to just compute a mean ADC value, you could use the method in this paper: http://www.ncbi.nlm.nih.gov/pubmed/25242901

The method uses the statistics of the mean magnitude MR data to compute the ADC so it is more accurate than a simple regression-based technique.  

Also the energy used in the process of breaking such hard bones is not available to go through to harm the underlying tissue (that's what a safety helmet does or a football helmet, for example - helmet is toast so the tissues underneath are not).  It is also why you have to replace a bike helmet every 3 years or after any accident - the lining materials are no longer useful as an energy trap and the energies of a blow go right through.  But the majority of the brain parenchyma is only loosely associated with the pia mater so it is the blood vessels (which run through the arachnoid and have to deal with entrances and exits through bones) and the white matter tracts (which are ultimately attached through the brain stem to the spinal cord and which run through longer spaces within the brain, connecting grey areas) that are vulnerable to the energies of an acceleration/deceleration injury.  The exception of course is the olfactory bulb and tract which run through bone and get sheared off leaving folks unable to smell in frontal injuries and accel/decel ones all the time.  Have you thought about doing a simple olfactory screen with your patients to document that kind of injury?

Hi Marco,

de Reus published a paper on this which I have used as a guideline, he suggests 60% as a suitable guideline. See also here http://www.sciencedirect.com/science/article/pii/S1053811913000050

Good luck!

To keep it simple, let's say that neurons exhibit notable plasticity when still viable whereas axons and axon tracts, to much lesser extent when damaged.  Depending on the cortical areas and activities involved, there are vastly different levels of redundancy and neuroplasticity.  Often, cortical neuronal damage may be treated by activating parallel pathways (neural network approach as evidenced by Carrick, recently on a study of 200 soldiers, victims of blast injuries).  And that, dear sir, brings us to pathways....  Axonal damage means that the corridors of communication are disrupted and therefore cannot pass information from one area of the brain to another...depending on function and redundancy, naturally.  Neuroplasticity is compromised more by axonal damage, and the risk of diaschisis is superior in axonal damage due to a domino effect that direct cortical damage, especially in the higher centers, manifests less.  However, if the axonal damage is significant in the more medial and caudal pathways (cord to brainstem/cerebellum, brainstem/cerebellum to mesencephalon, mesencephalon to thalamus) the manifestations will be much more pronounced than in supratentorial axonal lesions.  That make sense, right?  There are more connections from the more primitive areas of the brain when compared to the more recently developed cortical areas.  We can liken this to damage to a tree trunk that has larger implications to the tree's functions than damage to an area of tiny branches and leaves.  Unfortunately, in mTBI, the major damage is to central structures where axonal damage has more pronounced domino-effects that dramatically effect the basal ganglia, with it's consequent motor, cognitive, and limbic manifestations.

Dear Timothy Shea,

The "miserable minority" are those patients with mild traumatic brain injury who continue to experience prolonged cognitive deficits even after months and years, way beyond the average cutoffs (which is about 3-6 months post trauma).

You were absolutely right that one should consider preexisting comorbidities/(maybe even age) factors and as well as neurobehavioral/ neuropsychiatric condition arising from the trauma that could adversely influence the cognitive performance of these patients. Having said that, there are many studies (Ref 1-5) that has proven that even when all these factors are accounted for, and removed from the analysis, these individuals continue to exhibit deficits in their neurocognitive functions. So what then would best explain these manifesting derangements/ impairments ?

One possible explanation may lie in the structural changes. Patients with visible injuries -complicated mild TBI (radiologically visible lesions  i.e hematoma, contusion, edema, etc) and the corresponding functional and cognitive deficits are findings that are widely accepted by scientific community. How about the wider group of mild TBI patients whose radiological findings are usually negative and discharged from the ED without even being seen by neurosurgeons and without any subsequent follow-ups? Does normal CT/ MRI in the acute stage means there isn't any injury to start with with?

Many advances imaging protocols have proven otherwise. Diffusion Tensor Imaging (DTI) studies (reference to which I will give at the end of this reply) for example are able to pick up changes to the microstructure like the axons or the white matter tracts which are usually missed by the CT/ conventional MRI sequences. Could these structural anomalies/injuries then  explain the neurocognitive impairments experienced by the "miserable minority"? I guess the answer would be affirmative (see ref 5-8). With all these new evidences of microstructural abnormalities (even at the earliest hours post trauma), we can no longer reject the patient experiences as purely psychogenic in origin, ensuring we do our due diligence. 

1.  Stapert et al 2006 http://arnop.unimaas.nl/show.cgi?fid=4933

2. Silver et al, 2009 http://ajp.psychiatryonline.org/doi/pdf/10.1176/appi.ajp.2009.08111676

3. Bigler, 2013a http://journal.frontiersin.org/article/10.3389/fnhum.2013.00395/abstract

4. Le et al 2008 http://online.liebertpub.com/doi/abs/10.1089/neu.2008.0566

5.Spencer et al 2010 http://online.liebertpub.com/doi/abs/10.1089/neu.2008.0566

6. Shenton et al 2012 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3803157/

7. Bigler , 2013b http://www.mccauslandcenter.sc.edu/CRNL/sw/tbi/7/2013_BiglerA.pdf

8. Mayor et al, 2010 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2830922/

9. Irimia et al 2012 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3757727/

10. Croall et al, 2014 http://www.neurology.org/content/early/2014/07/16/WNL.0000000000000666

11. Lo et al, 2009 http://www.ncbi.nlm.nih.gov/pubmed/19346863

MRTrix is good for HARDI data and user-friendly! 

There is a large number of variety of software for DTI analysis. There are some general packages and some that do better in certain aspects of DTI analysis. It also depends on the operating system you want to install and whether you are looking for special more advanced types of DTI-related techniques which are a bit different, such as HARDI, DSI, or diffusion kurtosis imaging. Or whether you are thinking of doing tractography with it or not. Most MRI manufacturers also have some basic software you can use. Here are some:

MedINRIA

DTI Studio -- One of the first packages by Susumu Mori's lab at Hopkins and widely used for ROI analysis

DSI Studio -- Includes DTI, DSI, Q-ball and fiber tracking by Isaac Tseng's lab in Taiwan

ExploreDTI -- Comprehensive and include nice motion correction method.

Camino -- comprehensive toolsbut not very easy to use

TrackVis -- cross-platform and does DSI

Here's also a nice list (nonexhaustive) of various software with the advanatages of each: http://www.diffusion-imaging.com/2013/03/list-of-software-tools-for-dti.html

I also would suggest looking at this paper: A hitchhiker's guide to diffusion tensor imaging http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3594764/pdf/fnins-07-00031.pdf which has a nice list of the software and their features

Thank you for your kind suggestions, Dr Altay. Will surely check the paper out.

I am currently working on how certain genetic predispositions actually affects the events that occurs in between the primary insult and the secondary trauma in brain injury. For example, rs6265 of BDNF has been implicated to negatively affect certain psychiatric conditions. However, not much work has been done in mTBI, especially with regards to its influence on microstructural changes across different time intervals. While working on the DTI results, we realised that some quarter of our patients with vasogenic edema and astrogliosis (evinced in the preliminary results) performing poorly in neurocognitive assessments across the time points in comparison to others from the same mTBI group. We would like to understand why..

(I posted the question in order to encourage people to share their opinions and engage one another in the effort of advancing the limited knowledge on the subject matter.)

The DTFIT tool in the FDT toolbox of FSL is what we use, then feed this back into MATLAB to work with the diffusion tensor.

If I understand correctly, you want to build a gel phantom that reflects the T1/T2/ADC characteristics of the myocardium ? If so, look into the following articles:

You will have to adjust the formulas to match T1/T2/ADC values of the myocardium. You can find T1/T2 in this paper:

And depending on the purpose of your study, you should look into papers reporting ADC values for in vivo or ex vivo. They can differ quite a lot! Finally you should validate your phantom with a gold standard technique for each characteristic value (T1/T2/ADC).

Good luck on all that

I know of FSL and ExploreDTI. The former has been proposedabove and needs a Linux environment. The latter is a toolbox for matlab or standalone .exe file for Windows. In ExploreDTI, you can go to Plugins >> "Shuffle/select 3D volumes in 4D .nii files". Then specify the volumes to retain like 1:3 5:32 (to remove volume 4 and retain all until 32nd). You need to know in advance which number is your bad volume, but the software has tools for that if I remember correctly.

YES. HOWEVER, THE QUESTION MIGHT NOT BE SO STRAIGHTFORWARD AS IT SOUNDS.  FOR EXAMPLE, IN APATHY, A FREQUENT AREA OF INVOLVMENT IN TBI/MTBI, GENERAL APATHY, IN WHICH LOSS OF SEXUAL INTEREST IS OFTEN INCLUDED IN A SYNDROME OF SYMPTOMS, THE DORSOLATERAL AND MEDIAL PREFRONTAL/FRONTAL REGIONS ARE TYPICALLY INVOLVED, AND THE ASSOCIATED CIRCUITS THROUGH DOPAMINERGIC PATHWAYS. MORE DORSOLATERAL INVOLVEMENT IS USUALLY CHARACTERIZED BY PASSIVITY, REDUCED VERBAL OUTPUT, SLOWNESS IN INITIATING RESPONSES TO QUESTIONS, AND FLATTENED AFFECT. APATHY IMPLIES MORE MEDIAL INVOLVEMENT, AND ITS SUBCOTRICAL PROJECTIONS TO THE VENTRAL STRIATUM/NUCLEUS ACCUMBENS REGIONS, GENERALLY A "MOTIVATIONAL DRIVER," AND THIS CAN OCCUR INDEPENDENTLY OF DEPRESSION, ALTHOUGH APATHY IS OFTEN MISTAKEN FOR DEPRESSION. HOWEVER, THE KEY HERE IS THE LINKAGE TO THE Nac, THE "HEART" OF THE DOPAMINERGIC REWARD SYSTEM. HOWEVER, IN MTBI, THERE CAN BE LACK OF DIRECT "CORTICAL" INVOLVMENT, BUT INSTEAD, INVOLVMENT OF THE CEREBELLUM, WHICH HAS BEEN DEMONSTRATED TO SERVE AS A REGULATOR OF DOPAMINERGIC ACTIVITY IN MEDIAL ANTERIOR CORTICAL/VENTRAL STRIATAL REGIONS, AKIN TO A PHENOMENON OF DIASCHIZIS. BUT, THE CEREBELLUM IS ALSO INVOLVED IN MANY OF THE FUNCTIONS ASSOCIATED WITH THE PHYSIOLOGICAL RESPONSES OF ORGASIM; SO, IN GENERAL, THE MEDIAL/ACC REGIONS OF ANTERIOR CORTEX, THE VENTRAL STRIATUM, AND THE CEREBELLUM ARE ALL INVOLVED IN SEXUAL ACTIVITY. BUT "SEXUAL FUNCTIONG " IS A VERY GENERAL TERM, AND UNDER THIS OVER-ARCHING CATEGORY WE HAVE ATTRACTION, INTEREST, CONSUMPTION, ETC., AN EXTREMELY WIDE-ARRAY OF FACTORS THAT CAN BEST BE UNDERSTOOD WITHIN THE CONTEXT OF LARGE SCALE BRAIN SYSTEMS, WHICH ALWAYS IMPLIES INVOLVMENT OF WHITE MATTER TRACTS. I AM ATTACHING A PAPER WHICH I THINK WILL BE EXTREMELY RELEVANT TO YOUR QUESTION, "IF" I UNDERSTAND IT PROPERLY. --LK (LARGE FONT FOR ME - NOT YOU)

If I remember correctly the graph of ExploreDTI shows deviations of single volumes (or slices) from the remaining set. It should be an intuitive and quick graph indicator. However, rather than relying only on graphs produced by software, I would suggest visually inspecting all the volumes. All you need to do  is  find a software that can open 4D files and loop through them quickly. MRIcron cannot do this. FSL have a couple of +/- buttons to loop through 4D frames. I used DTIstudio for visual inspection. In DTIstudio the volumes are listed in a dropdown menu on the right. Click once this ara and use up and down arrows to loop through volumes. It gives also an idea of subject motion, because the quick motion of volumes is perfect for the eye to detect motion. It doesn't take more than 30 seconds to have a sense of whats going on with the dataset. You can also see stripes or lines in single volumes, which may indicate a single corrupt slice.

My advise is to remove volumes that are entirely corrupt. Some say 2-3 volumes out of 32 is the limit; of course your DTI deteriorates with gradually more orientations removed. It is also important at this point to make sure different groups of subjects do not have a different level of image corruption. You may find out differences in FA that are in fact related to the difference in number of orientations left after cleanup (yes, the number of orientations seriously impacts FA). The degree to which this is true may vary, depending if you remove 5 orientations from a dataset with 60 or with 22 orientations. It may depend also whether those 5 are randomly distributed or fall on the  same angulation.

Hope this helps.

There is a recent publication in Medical Physics titled: Signal-to-noise assessment for diffusion tensor imaging with single data set and validation using a difference image method with data from a multicenter study (by Wang ZJ, Chia JM, Ahmed S, Rollins NK; Med Phys. 2014 Sep;41(9):092302. doi: 10.1118/1.4893195).

The authors describe  a technique for noise measurements using high-pass filtered images and combine the noise properties from all available diffusion directions; the approach appears reasonable and sufficiently simple if no repeated measurements are available.

I will check if I am able to do this. 

Thank you again.

Don't know at which anatomical level you see the increase, but the splenium does not have much crossing fibers. Is actually one of the purest single orientation regions, if measured straight at the midline.

There is a measure called "mode of anisotropy", which is used as a marker of crossing fibers. Mode is not as established but may be useful to understand effects caused by crossing fibers when used to complement FA findings.

If you have a MacOS based workstation you can use OsiriX to do that:

<http://www.osirix-viewer.com/Downloads.html>

(download the free version)

Install the app and than use plugin manager to add the free plugin "ADCmap".

Attached file

These could be interesting for you:

Draganski B, Kherif F, Kloppel S, Cook PA, Alexander DC, Parker GJ, Deichmann R, Ashburner J, Frackowiak RS (2008) Evidence for segregated and integrative connectivity patterns in the human Basal Ganglia. The Journal of neuroscience : the official journal of the Society for Neuroscience 28:7143-7152.

Zhang D, Snyder AZ, Shimony JS, Fox MD, Raichle ME (2010) Noninvasive functional and structural connectivity mapping of the human thalamocortical system. Cereb Cortex 20:1187-1194.

Zhang D, Snyder AZ, Fox MD, Sansbury MW, Shimony JS, Raichle ME (2008) Intrinsic functional relations between human cerebral cortex and thalamus. J Neurophysiol 100:1740-1748.

Behrens TE, Johansen-Berg H, Woolrich MW, Smith SM, Wheeler-Kingshott CA, Boulby PA, Barker GJ, Sillery EL, Sheehan K, Ciccarelli O, Thompson AJ, Brady JM, Matthews PM (2003) Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging. Nature neuroscience 6:750-757.

And this for cerebellum:

Buckner RL, Krienen FM, Castellanos A, Diaz JC, Yeo BT (2011) The organization of the human cerebellum estimated by intrinsic functional connectivity. Journal of neurophysiology 106:2322-2345.

You may have a look at the jhu DTI database http://gforge.icm.jhu.edu/gf/project/dtmri_data_sets/docman/?subdir=39

Yes, of course.

I read on the TBSS forum that "though of course we always try to discourage people

from using ANOVA in general" and I read some instructions on the web but it seems that there is a little bit of confusion about this argument.

This phenomenon can be observed in cell culture:

Other:

Regards,

Joachim

Dear Shoji,

There is no simple answer, of course, as there're plenty of software packages and you will probably get different suggestions here.

As you already have some experience with neuroimaging data analysis, I would generally recommend using software that you are already familiar with (SPM/FSL?).

Currently, I do my DTI analysis mostly in FSL and Tracula (Freesurfer implementation), depending on whether I want to run it voxel- or ROI-wise. The first one (TBSS in FSL: http://fsl.fmrib.ox.ac.uk/fsl/fsl4.0/tbss/index), I think, is substantially easier for beginners, but if you're already familiar with Freesurfer and comfortable with shell scripting - then I would suggest Tracula.

If you like python, then you might be interested in looking at Nipype (http://nipy.sourceforge.net/nipype/index.html), which is a meta-package with a possibility to combine multiple toolboxes into one workflow. It also has some ready-to-use pipelines for DTI.

Recently, I was at the course, where we also had BrainSuite lab (http://brainsuite.loni.ucla.edu/). The software is well-documented and seemed relatively straightforward to work with as well.

Finally, I heard some positive feedback from people working with Camino (http://cmic.cs.ucl.ac.uk/camino//index.php). Haven't tried it myself so far, but I think I will, as Nipype has a pipeline, combining it with Freesurfer.

Good luck with your analysis!

The general rule of thumb is that if you predicted changes in particular areas before you collected the data, you are justified in reporting uncorrected results. But you must explain your a priori hypothesis. If you didn't have any idea where you would see changes appear in the brain, you must use corrected data. For TBSS presentation, a threshold of p<0.05 corrected is considered acceptable.

Tim

Article on "Pediatric diffusion tensor imaging: normal database and observation of the white matter maturation in early childhood" may help you to get controls for your study , if the database of this study is still maintained.

The whole question of how to select ROIs is not terribly straightforward. I'm generally a big fan of ROI-based approaches, as I think they can be much more powerful than whole-brain analyses, but there are essentially two schools of though that are best exemplified by this pair of papers by Karl Friston et al. (http://www.ncbi.nlm.nih.gov/pubmed/16635579) and Saxe/Brett/Kanwisher (http://www.sciencedirect.com/science/article/pii/S1053811905025796)

Essentially the Kanwisherian (!) position is that the optimal way of defining fROIs is by a completely independent (i.e. separate set of data) localiser task. Friston's argument is that fROIs can be defined by using an orthogonal contrast from within the same task/experiment. Both methods have advantages and drawbacks, and both are pretty much acceptable these days.

The clear advantage of the Fristonian view is that you don't need an extra localiser task. It's also very useful in being able to decompose complex designs and view statistical interactions in a simple way. For instance, say you're interested in face processing, and have a balanced 2x2 factorial design with male vs. female faces as one factor, and happy vs. sad expressions as the other factor. You could, using appropriate contrasts, view the interaction of the two factors as a statistical map, however viewing interaction effects as SPMs is not generally that informative, as you can never be sure exactly which lower-level effect(s) is/are producing the activated voxels.

A better approach with that kind of design would be to define a simple t contrast of All faces > baseline. This would likely give you a nice fat blob in the fusiform face area. You could then use that ROI to pull out amplitude data related to each of your four conditions for each subject, average the data, and plot it in a simple histogram; it should be instantly clear what's going on in the data. Alternatively depending on your research question, it might also be preferable to use one of the main-effects of the factors (e.g. All male vs. female faces) to define the ROIs - this would likely give your more specific ROIs. These contrasts are still orthogonal to the lower-levels of your design and therefore independent; in fact, I had a long conversation with Tom Nichols about this, and he said that they're what statisticians call 'conditionally independent' but it boils down to the same thing.

An important aspect of this approach is that the statistical inference is done on the amplitude data pulled out of the ROI - not on the brain images. In a sense then, it doesn't matter how you define your ROIs - they can be defined at any (reasonable) corrected or uncorrected statistical threshold, and by any (reasonable) method. To (eventually - sorry) return to your original question then, I think selecting ROIs in 'unpredicted' locations is fine - and in fact it might even be somewhat scientifically irresponsible to ignore results from these locations. Clearly though, if you end up with 100 ROIs your results are going to be very onerous to interpret, and you'd have to do some kind of (Bonferroni-type) correction on your inference stats; this would be an argument for thresholding your ROI definitions at a 'standard' level e.g. p<0.001 for individual data, or p<0.05 (corrected) for group data.

Hope that helps! Sorry for the long-windedness, but these are difficult questions, and don't have short answers!

Cheers,

Matt.

PS. This site: http://web.mit.edu/evelina9/www/funcloc.html has a lot more good info on ROI methods - by one of Nancy Kanwisher's research group - Ev Fedorenko.

DTI, when only one tensor is used to characterize the fibre orientation in each voxel, is incapable of providing an accurate whole brain structural network for any kind of analysis. The reason for this is that most white matter voxels have crossing fibres, which cannot be fitted with a single tensor. At least 60 diffusion gradient directions, and a technique such as spherical harmonic decomposition, must be used to resolve multiple fibre directions. Use of a single tensor model generates nonsense. See: Jones DK, Knösche TR, Turner R. White matter integrity, fiber count, and other fallacies: The do's and don'ts of diffusion MRI. Neuroimage. 2012 Jul 23. pii: S1053-8119.

Dear Luca and Lindsay,

in my experience, you need to rotate the gradients according to the image slice orientation every time this orientation is not the identity (i.e. transverse slice aligned with the axes).

The overplus option only changes the gradient orientations to maximize the gradient strength. In both overplus and non-overplus situations, the directions in the par header are given in the "LPS" frame (Left-Anterior-Superior) and need to be re-oriented onto the diffusion weighted image frame.

I hope this helped.

Regards,

Nicolas Toussaint