MRG: add Gregory Lee's new PARREC data

[matthew-brett]

Test parrec data with multiple rotations.

Modify translation order to minimize error between our rotations and
those from the NIfTI images in the conversion.

[matthew-brett]

The slight offset in the translations is still very puzzling to me.

I see that the amount that the afffine is off, depends on the rotations. Here is a print of the filename, the rotations in FH, AP, RL order, and the error in the translations:

T1W_cor_0_0_0_SENSE_13_1.PAR 0.0 0.0 0.0 [-0.0943233  -0.00043228 -0.0945    ]
T1W_cor_5_15_30_SENSE_12_1.PAR 5.0 15.0 30.0 [-0.08251608 -0.07266869 -0.07619181]
T1W_sag_0_0_0_SENSE_8_1.PAR 0.0 0.0 0.0 [ 0.00018147 -0.09492847 -0.0945    ]
T1W_sag_5_15_30_SENSE_11_1.PAR 5.0 15.0 30.0 [ 0.03278593 -0.1254854  -0.03404345]
T1W_trans_0_0_0_SENSE_4_1.PAR 0.0 0.0 0.0 [-0.0943233  -0.09492847  0.        ]
T1W_trans_5_15_30_SENSE_10_1.PAR 5.0 15.0 30.0 [-0.06638231 -0.10357589  0.05282434]

Has anyone got any ideas how to track this down?

[matthew-brett]

Here's a script to investigate - https://gist.github.com/d251bd4f6c5aab653161 - nothing much came to me looking at the output.

[matthew-brett]

Any thoughts on the incorrect translations? I can't think of anything to try.

[grlee77]

It is good to see that the rotations are a very close match now.

I played with it for awhile, but I couldn't get a better result for the translations either. The peak translation error across the test cases is about 0.3 voxels, which I could live with. Without access to the vendor's conversion code, it probably won't be easy to track down the source of the discrepancy.

[matthew-brett]

I'm surprised by the apparent order of the rotations - and it's very difficult to check visually with the one-slice images here.

@grlee77 - should you be bored one day in front of the scanner - would you consider taking some images with a fair few slices and three biggish rotations (maybe 15 20 25 degrees or something) so we can see visually if the rotations are right?

[matthew-brett]

I will merge this one soon unless I hear otherwise.