NITRC Diffusional Kurtosis Estimator Forum: dke-questions

Mouse datasets

Dominika Valášková — Thu, 24 Sep 2026 8:21:25 GMT

Hello, I am trying to process preclinical diffusion data using the DKE and would appreciate some advice.  I am working with in vivo mouse brain data acquired on a 7T Bruker BioSpec (70/30) scanner running ParaVision360 3.6, using a standard DWI EPI sequence.  I have tried importing both converted NIfTI files and raw DICOMs into DKE, but I am running into significant issues with the output maps. Initially, the default spatial smoothing settings were clearly tailored for human data and were far too strong for the small mouse brain. However, even after reducing the filter, the resulting kurtosis maps remain quite "holey", with zero-value voxels or empty regions throughout the brain tissue.  I am wondering if DKE is actually suitable for small animal data or if it is strictly optimized for human datasets. Additionally, I would like to know if there are specific pre-processing requirements or data formatting steps for Bruker/ParaVision data that I might be missing, and if anyone can share recommended parameter settings for mouse data.  Any advice on settings, data import would be greatly appreciated.  Thanks in advance! Dominika Valaskova

RE: Strage pxel blocks on fitting Kmeans maps and parts of the image are missing

ruicheng ba — Tue, 29 Mar 2022 8:30:19 GMT

[i]Originally posted by Emilie McKinnon:[/i][quote]Hi!

Check out the masking feature. By default, DKE uses a threshold value of 50 in the B-0 value to create a brain mask. Just by glancing at your image it looks like this is animal data and so your threshold might have to be adjusted?

Just so you know, we are phasing out support for DKE and have moved to a more modern software that will be monitored more closely. The documentation can be found here: https://pydesigner.readthedocs.io/en/latest/

I apologize for the delay!

Best,
Emilie[/quote]
[color=#000000]Thank you for your reply! [/color]I have tried the directional fitting instead of tensor fitting and I found these pixels disappear and the fitting results became coherent. So what's the difference between these two fitting methods? If I acquired data in >15 directions, which methods could be more accurate? By the way, is there any requirements on adjusting the B-0 threshold?

Best,
Ruicheng

RE: Strage pxel blocks on fitting Kmeans maps and parts of the image are missing

ruicheng ba — Tue, 29 Mar 2022 8:29:29 GMT

Thank you for your reply!

I have tried the directional fitting instead of tensor fitting and I found these pixels disappear and the fitting results became coherent. So what's the difference between these two fitting methods? If I acquired data in >15 directions, which methods could be more accurate? By the way, is there any requirements on adjusting the B-0 threshold?

Best,
Ruicheng

RE: Strage pxel blocks on fitting Kmeans maps and parts of the image are missing

Emilie McKinnon — Mon, 28 Mar 2022 14:56:18 GMT

Hi!

Check out the masking feature. By default, DKE uses a threshold value of 50 in the B-0 value to create a brain mask. Just by glancing at your image it looks like this is animal data and so your threshold might have to be adjusted?

Just so you know, we are phasing out support for DKE and have moved to a more modern software that will be monitored more closely. The documentation can be found here: https://pydesigner.readthedocs.io/en/latest/

I apologize for the delay!

Best,
Emilie

RE: Fail in installation of DKE

Emilie McKinnon — Mon, 28 Mar 2022 14:53:17 GMT

Hi Liu,

Couple of questions:

- What operating system are you using?
- Did you try running DKE using the command line?

Lastly, just so you know we are phasing out support for DKE and have moved to a different processing software that will be more active. I will post the link here for your convenience: https://pydesigner.readthedocs.io/en/latest/

Best,
Emilie

Fail in installation of DKE

Liu Kuiyuan — Mon, 28 Mar 2022 9:49:44 GMT

I installed MCR （R2012a (7.17) 64-bit），but I can't run DKE. Using DKEGUI.exe, I get message:

[b]idx_gradients{1} = [1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18];[/b]
[b]idx_gradients{2} = [1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18];[/b]
[b]idx_gradients{3} = [1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18];[/b]

[b]command line: dke DKEParameters.dat[/b]
[b]An Error has occurred while trying to initialize the MCR.[/b]
[b]The error is: Fatal error loading library C:\WINDOWS\system\libmx.dll Error: The specified module could not be found[/b]
[b]Error:mclmcr initialization failed[/b]

I don't know if I installed MCR correctly. When I installed MCR, I just selected installation path,should I select a specific path?

I tried to set environment variables,but it still didn't work(message above).I really need help.

Strage pxel blocks on fitting Kmeans maps and parts of the image are missing

ruicheng ba — Thu, 20 Jan 2022 12:18:28 GMT

I input dwi images with b=0,0.8,1.5 and 2.5, where b0 has one direction and the other b values have 18 directions. There are discontinuous squares on the Kmean maps of the output result. May I ask why this is caused? I checked the original image, and there were no artifacts of similar shapes.
In addition, I found that part of images on Kmean Map were missing. I attached the comparison images. Attachment 1 is the Kmean map, and Attachment 2 is the B0 image of the input data.

RE: Apply DKI model to multi-shell acquisition?

Cristina Roman — Fri, 08 Oct 2021 17:31:59 GMT

Thank you very much, Kayti!

RE: Apply DKI model to multi-shell acquisition?

Kayti Thorn — Fri, 08 Oct 2021 16:54:28 GMT

Hi Cristina,

The parameters you supplied are sufficient for DKI and DKE should work just fine with your dataset. However, DKE is now a legacy software that is no longer supported. Our lab has transitioned to a similar tool called PyDesigner, which we encourage all DKE users to consider using as well. It is a hands-off Python-based implementation of [url=https://github.com/NYU-DiffusionMRI/DESIGNER]NYU's DESIGNER[/url] dMRI preprocessing pipeline designed to maximize signal-to-noise ratio. You can read about it further on [url=https://github.com/m-ama/PyDesigner]the PyDesigner Github page here[/url] and in [url=https://pydesigner.readthedocs.io/en/latest/]the PyDesigner documentation here[/url].

Apply DKI model to multi-shell acquisition?

Cristina Roman — Wed, 06 Oct 2021 18:15:49 GMT

Hello,

I have a multishell diffusion protocol (see parameters below) and was curious if I can apply a DKI model to the data through DKE to obtain DKI metrics? Or, are the below parameters insufficient for DKI?

Thank you in advance for your help!
Cristina

3T Siemens Prisma Scanner, 64-channel head coil
Multiband factor of 3, no GRAPPA, bipolar acquisition
Diffusion directions calculated using INRIA (Caruyer et al., 2013)
Six separate scans
12 directions, 2 B0s (A>>P), b= 500 s/mm2
12 directions, 2 B0s (P>>A), b= 500 s/mm2
29 directions, 3 B0s (A>>P), b=1500 s/mm2
29 directions, 3 B0s (P>>A), b=1500 s/mm2
64 directions, 4 B0s (A>>P), b=3000 s/mm2
64 directions, 4 B0s (A>>P), b=3000 s/mm2
Data was concatenated and run through topup and eddy