I only started using DTK over the last few days to try and process data from a Philips scanner.  The data is in DICOM format, taken from a Philips 3T Achieva in HIGH DTI with 32 directions.  I have processed the same data in Matlab and the result looks sensible.  However, I am not getting anything sensible with DTK.  I am doing something wrong in DTK and wonder if someone can suggest what I have missed.

For DTK, I convert the DICOM data to NIFTI using MRIconvert.  This creates 34 img files, aaa_001_033.img, aaa_002_001.img to aaa_002_0033.img.  I assume aaa_001_033.img is b0 and aaa_002_0033.img is average.  Not sure if DTK is picking this up correctly.  For "Raw data files" in DTK, I point it to aaa_002_001.img.

 

The gradient table is below, a 32x3 matrix.  I use the same matrix in Matlab.

Number of b0 volumes = 1

Image orientation = auto

Filename paddings = 3

Filename start number = 1

Number of averages = 1

 

I have also tried inverting the X and Y in the gradient table, e.g. changing the sign in the second and third columns.

 

Not sure what else I can try.  Any suggestions greatly appreciated.

 

Gradient Table

-0.499998,   -0.499998,   -0.70711

-0.499998,   -0.499998,   0.70711

0.707107,   -0.707107,   0

-0.653288,   -0.270606,   -0.707098

-0.208664,   -0.67563,   -0.707095

0.0196577,   -0.706829,   -0.707112

0.421225,   -0.56795,   -0.707109

0.689925,   -0.154927,   -0.707108

-0.653495,   -0.270675,   -0.70688

-0.292882,   -0.707102,   -0.643604

0.294515,   -0.706411,   -0.643619

0.514993,   -0.486072,   -0.706056

0.707102,   -0.292882,   -0.643604

-0.707102,   -0.472486,   -0.526084

-0.472486,   -0.707102,   -0.526084

0.555492,   -0.64395,   -0.526077

0.707102,   -0.472486,   -0.526084

-0.707107,   -0.707107,   -0.000212132

-0.707102,   -0.472486,   0.526084

0.707102,   -0.472486,   0.526084

0.472486,   -0.707102,   0.526084

-0.707085,   -0.707085,   0.00784864

-0.636392,   -0.425181,   0.643604

-0.706047,   -0.706047,   0.0547301

-0.292882,   -0.707102,   0.643604

0.292882,   -0.707102,   0.643604

0.707085,   -0.707085,   0.00784864

0.707102,   -0.292882,   0.643604

-0.70626,   -0.70626,   0.0489319

0.0347188,   -0.706256,   0.707105

0.70706,   -0.70706,   0.0115262

0.707107,   0,         0.707107

 

Using Matlab

 

Using DTK

 

Ruopeng said:

Hi Ken,

Your file naming is too confusing. DTK would not be that smart to figure it out automatically. First you need to toss the average image. Then name all the files (b0 and the high b's) with the same prefix and number them in uniformed format, such as aaa_001.img, …, aaa_033.img. The way you did was actually using the first high b image as b0 and use the last (average) image as the last high b image. It will not reconstruct correctly.

Best,

Ruopeng

PS. from the screenshot it looks like your matlab result needs a orientation flip.

---

Thank you Ruopeng, I am now getting something sensible.  See attached image.  DTK looks like a nice piece of software.  For reference what I had to do with the DICOM data are

- Convert the data to NIFTI using MRIconvert.

- Remove the average image, the last image.

- Rename the 33 remaining images with same prefix and with increasing number.  B0 being the first image.

- Don't forget to rename the associated *.hdr files.

- Use the gradient file as above.

- Set Orientation Patch to :  Invert X

 

I am trying to work out the difference with the Matlab results.  The orientation seems to be a different convention somewhere.

 

Ken