Index

• T1, T2 and PD images
• Rotation (Pitch, Roll and Yaw)
• Radians
• Planes of Section (Coronal, Axial, Sagittal)
• Translation of images between programs (see also Image Conversion on the glossary page)
• Artifacts: The Good, the Bad, and the Ugly
• 1.5T vs 3.0T Image Comparison
• Dimensions of 2D and SPGR images can be found on the Afni Anatomical Preprocessing page.

T1, T2 and PD images

Anatomical Features	T1	T2	PD (Proton Density)
White Matter	Bright	Dark	Bright
Grey Matter	Dark	Bright	Very Bright
CSF	Very Dark	Very Bright	Grey

Radians

Radians, rather than degrees, are used as a measure of rotation in some image processing programs (e.g., SPM).

1 radian=the number of degrees in the angle in a circle for which the arc and radius of the angle are equal.

• radians=degrees(pi/180).
• 1 degree=pi/180=0.0174532925 radians.
• 180°=3.14156 radians
• 90°=1.5708 radians
• 45°=0.7854 radians

The information and images included below about planes, have been taken directly from: http://www.madsci.org/~lynn/VH/planes.html. Please visit that site to learn more and to see helpful animations.

Coronal (X-Z)	Sagittal (Y-Z)	Transverse/Axial (X-Y)
Coronal sections follow front to back, as though cutting through a corona, or halo, around the head. In medical terms anterior means front, and posterior means back.	Sagittal sections follow from one side of the body to the other - left to right, or right to left. In medical terms lateral means 'outside', and medial 'inside' in the sense of lying in the middle. Your nose is medial (and inferior!) to your eyes, while your ears lie lateral to both your eyes and nose.	Transverse (or axial) sections form a series of circumferential slices - rather like slicing the body into a series of pancakces and stacking them atop one another. Transverse sections run top to bottom (superior to inferior). Your chin is inferior to your forehead but superior to your knees.

#	Condition	Up is Up	Eyeballs on Top	Right side up (sagittal)	Nose to Left (sagittal)	Right side up (Coronal)	L=R	Front is Front
1	Afni: 2d, 3d, fx	*	Y	Y	Y	Y	Y	*
2	SPM (M-files for 2d anat and fx)	Y	N	N	N	N	Y	Y
3	MRIcro	Y	Y	N	Y	N	Y	N
4	MRIcro-Dorsal	N	Y	Y	Y	Y	Y	N
5	MRIcro-Dorsal-LR	N	Y	Y	Y	Y	N	N
6	SPM Mfile, x flip	Y	N	N	N	N	N	Y
7	SPM Mfile, y flip	Y	Y	N	Y	N	Y	N
8	SPM Mfile, z flip	N	N	Y	N	Y	Y	Y
9	SPM Mfile, x-y flip	Y	Y	N	Y	N	N	N
10	SPM Mfile, y-z flip	N	Y	Y	Y	Y	Y	N
11	SPM Mfile, x-z flip	N	N	Y	N	Y	N	Y
12	SPM Mfile, x-y-z flip	N	Y	Y	Y	Y	N	N
13	SPM T1 Template	N	Y	Y	Y	Y	N	N

Table Interpretation

The table demonstrates how image orientation is affected by various flips and conversions. Note that such values can never be absolute, but only relative to the original image:

ROWS

1) Our original images were in the orientation described on line #1 (Afni). * in this case means that the criteria for judgement are different for afni than spm. In these particular cases:

Up is Up: Slice #1 is at the bottom of the head (in axial view) and
Front is front: Slice #1 is forward (towards the face/eyeballs) in the sagittal and coronal views.
2) We used a program called phx2spm.m to convert functional P-files to ANALYZE/SPM format. The difference in the image format results in several flips of the data. We used another program efl2ana.m (aka t12spm.m) to convert the 2D structural data. In fact, both the functional data and the structural data end up in the same orientation when they are converted (this assumes that you say "No" to flipping the 2D structural data).

3) MRIcro is a freeware image processing program for the PC. It does a nice job of converting the anatomical image files, but leaves them in a different orientation than the M-files. This line represents the default conversion choices for a 2D image.

4) In MRIcro, if you check "First file is dorsal" in the image conversion routine, the result resembles the SPM Template image EXCEPT that L-R flipping has not been done.

5) If you check "First file is dorsal" AND "flip L-R" in the image conversion routine, the result is equivalent to the SPM Template image.

6-12) If you have created an image file as in line #2, you can then use SPM to display it and flip it in various directions (x,y,z). These detail the results of all possible flips. Note that in our case, flipping on x,y, and z results in a file in exactly the same orientation as the SPM template.

COLUMNS

I've tried to examine all of the ways in which image orientation could change, and ended up with seven parameters explained below. On line #1, the assumption is that the image is being viewed in Afni. In all other cases, I assume the image is being viewed using the <Display> feature in SPM. This feature allows one to move "up", "forward" and to flip the images on x, y, and z axes:

A) Up is up: "Yes" if putting a positive value in the "up" box results in moving toward the top of the head.

B) Eyeballs on top: "Yes" if the face/eyeballs are oriented toward the top of the screen, as if the subject is staring up at the ceiling.

C) Right side up (sagittal): "Yes" if the orientation of the sagittal image is such that the top of the head is nearer the top of the computer screen (i.e., the subject does not appear to be "standing on his head")

D) Nose to left (sagittal): "Yes" if the subject is facing the left side of the computer screen. For example, the letter "X" is to the left of the letter "Y" in this sentence.

E) Right side up (coronal): "Yes" if the subject does not appear to be standing on his head in the coronal view.

F) L=R: "Yes" if the image is in radiological orientation. "No" if the image is in neurological orientation.

G) Front is front: "Yes" if putting a positive value in the "forward" box results in moving toward the face/eyeballs.

Using images created in mricro (without dorsal or left-right flip flags checked), we make the following changes to put the file into correct SPM orientation:

• pitch (in radians) 1.5708 (i.e., this is 90 degrees)
  
• roll (in radians) 1.5708
  
• resize (y) -1 (This appears necessary for images that are reconstructed with mricro only)
  

This reorients the sagittal images correctly for our purposes. Left remains left.

To flip the image into radiological orientation: resize (z) -1.

WARNING: changing the orientation this way only works for programs like SPM that read the mat file. To make a permanant change to the image, that you can export to another program (like MrGray or MRIcro) do one of the following:

MRIcro's "save as" function will allow you to correctly describe the orientation of the image (ours are sagittal-left) at which point you can save it and it will then correctly understand how to display the image. SPM will also correctly display the image after this.

Alternatively, reorient.m will reslice the image into 1 mm planes and then permanantly change the orientation of the images. Reorient.m reads the mat file and applies those changes to the image, creating a new img and hdr with "r" prepended to the front of the name. Get a copy of reorient.m from the Mrgray Download page. The benefit of this program is that it handles nonisotropic voxels by reslicing, thus it prevents distortion.

WARNING 2: SPM's normalization procedure assumes by default that you are giving it images in radiological orientation. At the point of normalization, it automatically flips the functional images L-R so that they will match the template image. If you do not normalize your functional images, then you must either explicitly flip them L-R before coregistering them to the 3D structural OR you must L-R flip the 3D structural before coregistering.

Nothing substitutes for experience when evaluating an image for movement, artifacts etc. But below you will find a collection of normal and abnormal images that should provide a guide to evaluating something of your own.

Normal fMRI image	Byte swapped fmri image (Learn more about byte swapping):

Image with artifact	Equivalent slice without artifact

Subject 1	Subject 2

April 26, 2005 Artifacts

Note the fluctuating character of the radio frequency noise (background speckles). These are seen in a problem dataset as one scrolls through the timepoints with afni or spm_movie. You must make sure the contrast is high enough to reveal the background noise (this is especially true for spm_movie). Then bad timepoints should flash at you as you scroll past them.

bad timepoint	fairly good timepoint	another bad timepoint

1.5T vs 3T

We created image sets in both the 3T and the 1.5T for comparison purposes. Images in the 1.5T scanner were collected with our regular spiral sequence. Images from the 3T were collected with the spiral-in-out sequence. Images were preprocessed and analyzed in afni using the same choices and the same P-value. These images are as follows (Thanks due to Ted Trouard who donated his brain for our purposes):

Complex finger tapping on the left and then the right side. The images were analyzed to contrast left and right, with the left being positive. Images show activation on the appropriate side in the motor and sensory strips.

finger tapping: 1.5T

finger tapping: 3.0T