Sunteți pe pagina 1din 4

Checking b-vector alignment

Chris Rorden, Michael Harms, Jolinda Smith, Nytavia Wallace :: March 31, 2011
In di!sion tensor ima"in" #$%I& 'e constr!ct tensors (y collectin" a series o direction)
sensitive di!sion ima"es* In addition to recordin" the ima"es, the scanner saves these
directions, 'hich are later !sed to reconstr!ct di!sion +ro+erties* %hereore, +o+!lar
$IC,M ima"e conversion tools #dcm2nii, MRIconvert& 'ill "enerate (oth NI%I ormat
ima"es #e*"* -dti*nii.& as 'ell as "radient vector #-dti*(vec.& and ()val!e #-dti*(val.& iles*
/ +otential concern is that the scanner man!act!rers can choose to either re+ort the
vectors 'ith reerence to the scanner (ore, or 'ith reerence to the ima"in" +lane #i*e*,
ima"in" "rid&* %his is not a +ro(lem i the ima"es are al'ays ac0!ired +recisely
ortho"onal to the scanner (ore #1i"!re 1&, as the ima"e and scanner have the same rame
o reerence* Ho'ever, +ro(lems can arise 'hen the ima"e +lane is not ali"ned 'ith the
scanner (ore #i*e*, o(li0!e ac0!isitions&* In this sit!ation, it is im+ortant to ens!re that the
vectors +rovided to 1S2 or MedInria are in the same rame o reerence as the ima"e,
re"ardless o ho' they 'ere stored in the $IC,M ile* It sho!ld (e noted that these
an"!lation errors 'ill have little inl!ence on $%I +arameters that are invariant to tensor
rotation, s!ch as /$C, M$ and 1/* Ho'ever, the ei"envectors o the tensor, and
conse0!ently tracto"ra+hy +ro"rams are sensitive to +ro+er inter+retation o the (vecs
relative to the ima"in" +lane*
3nort!nately, there remains some con!sion re"ardin" ho' dierent scanner
man!act!rers, and dierent sot'are versions or a "iven man!act!rer, store "radient
vectors 'ithin their $IC,Ms* %hereore i yo! +lan to !se the tensor ei"envectors or
cond!ct tracto"ra+hy 'e recommend that yo! collect test datasets to ens!re that yo!r data
is converted correctly or yo!r scanner ty+e and sot'are version* Here 'e descri(e an
easy 'ay to do this* Im+ortantly, small an"!lation errors in (vec orientation are hard to
detect as the vectors are three dimensional, and most tracts c!rve across yo!r ima"in"
+lane* %hereore, a "ood +lace to loo4 or errors in (vec orientation is the cor+!s
callos!m 5 a ma6or i(er (!ndle 'here the i(ers r!n in an almost +erect let)ri"ht
direction at the midline* %hereore, to chec4 tract directions, 'e recommend ac0!irin" an
ima"e 'here the roll an"le o the ima"in" +lane is dramatically dierent rom the scanner
(ore, (!t the +itch and ya' remain ortho"onal to the scanner (ore #1i"!re 2&* In this case,
a dramatic error 'ill (e seen in the cor+!s callos!m i the (vec directions created (y yo!r
ima"e conversion sot'are are not ali"ned 'ith the ima"e "rid #ass!min" that the $%I
sot'are yo! are !sin" e7+ects the (vecs to (e s+eciied relative to the ima"e "rid, 'hich
is the case or 1S2&*
Figure 1: Image acquired orthogonal to scanner bore
Figure 2: Image acquired with slice orientation rotated relative to the scanner bore (a
roll has been applied in this example).
We s!""est com+!tin" the irst ei"envectors o the tensor 'ith 1S2, and then dis+layin"
these vectors as lines, !sin" a +roced!re descri(ed on o!r 'e( +a"e
#'''*ca(iatl*com8Reso!rces8Co!rse8t!torial8html8dti*html&* 1i"!re 3 sho's an e7am+le
o an acc!rately ali"ned vers!s a misali"ned set o irst ei"envectors*
Figure 3: ines on each voxel show the estimated !iber direction based on the !irst
eigenvector o! the calculated di!!usion tensor. "ross#hairs indicate the corpus callosum$
which should run !rom the le!t to right hemisphere. %ote the !irst eigenvectors o! the
tensor in the middle panel are misaligned$ whereas the& are accuratel& aligned with the
anatom& in the right panel.

Note that yo! need to ma4e s!re the lines ollo' the anatomy o the o(li0!e ac0!isition*
I a color ('()) dis+lay o the directions 'as !sed a(ove, the colors 'ill not match 'hat
yo! ty+ically e7+ect (eca!se o the o(li0!e nat!re o the ac0!isition #i*e*, +!re red 'o!ld
relect the direction o the irst a7is o the ima"e "rid, 'hich 'ill N,% (e e0!ivalent to
the tr!e let)ri"ht a7is o the s!(6ect.s o'n anatomy d!e to the roll a++lied to the
ac0!isition, and th!s vo7els in the center o the cor+!s callos!m 'o!ld N,% sho' as
+!re red&*
When creatin" ima"e conversion sot'are, 'e try to set dea!lt conversion val!es that 'e
(elieve 'ill yield (vecs 'hose direction is deined relative to the ima"e "rid* Ho'ever,
'e only have access to a limited n!m(er o scanners and systems, and thereore 'e
enco!ra"e all !sers 'ho 'ish to have acc!rately oriented tensors to cond!ct a 0!ic4
validation 'ith their system, and re+ort any +ro(lems to the develo+ers o dcm2nii and
MRIconvert*
/s re"ards Siemen.s scanners s+eciically, the latest versions o dcm2nii #rom /+ril
2011& no' incl!de three dierent val!es that allo' yo! to c!stomi9e ho' dcm2nii
converts yo!r Siemens $IC,M ima"es* %hese can (e seen (y editin" yo!r *ini ile 'ith a
te7t editor #the location o this ini ile is re+orted 'hen yo! r!n dcm2nii and dcm2nii"!i&*
%hese val!es and their c!rrent dea!lt settin"s are:
Siemens$%I3se001:I001;1020atleast<1=
Siemens$%INo/n"!lationCorrectionI001;1020atleast<1000
Siemens$%IStac4I001;1020atleast<1=
%he n!m(er or each val!e allo's yo! to ad6!st ho' ima"es 'ill (e converted, relative to
the version o Siemens sot'are yo! are !sin"* When yo! convert a Siemens $%I ima"e
'ith dcm2nii yo! 'ill (e told the version o Siemens sot'are detected and the c!rrent
conversion dea!lts* 1or e7am+le:
%umber o! di!!usion directions * 31
%ote: detected +iemens +o!tware version ,--1.:1-2-/ * 10
#1ill use --12:---3 or --12:1--3 instead o! --22:1-2- i! version 4* 15
#1ill stac6 across 7cquisitions i! version 4*15
#%o slice angulation correction o! vectors i! version 4*1---
8o ad9ust$ edit "::pas:mricron:dcm2nii:dcm2niigui.ini
%he meanin" o these settin"s are as ollo's:
Siemens$%I3se001:I001;1020atleast<1=: Siemens stores the vector val!es in
(oth the main $IC,M header #ta"s 001::000> or 001::100>& as 'ill as a hidden
shado' header #+art o the CS/ header in 002::1010&* In older versions o
Siemens sot'are the $IC,M val!es in 001::100> 'ere either a(sent or
considered inacc!rate*
Siemens$%INo/n"!lationCorrectionI001;1020atleast<1000: /s lon" as the
system version is less than 1000, the (vec val!es e7tracted rom the $IC,M
header 'ill (e rotated to com+ensate or the ima"e +lane* 1or hi"her versions,
ima"es 'ill not (e rotated #note that the c!rrent Siemens sot'are version is
?@1A #and ?$11&, so (y dea!lt all Siemens $%I data 'ill (e rotated&*
Siemens$%IStac4I001;1020atleast<1=: While most se0!ences store data rom a
sin"le se0!ence as a sin"le ac0!isition 'ith a series o ima"es, Siemens scanners
can act!ally save each direction and each avera"e as a dierent ac0!isition* %his
eat!re sho!ld a!tomatically colla+se across these, creatin" a sin"le NI%I ile* I
this is not set correctly, yo! may ind that yo!r $IC,M ima"e is converted into a
series o NI%I ima"es #e*"* one ima"e +er n>7&* %hese +ro(lems are !s!ally
+retty easy to detect*
2ast, t'o additional notes:
1& %he iss!e o (vec rotation is not strictly limited to o(li0!e ac0!isitions* 1or e7am+le, i
yo! collect a strictly coronal ac0!isition !nder Siemens ?@13, ?@1=, or ?@1A, and yo!
don.t a++ly an an"!lation correction to the directions stored in the $IC,M, then yo!r
ens!in" (vecs 'ill not (e correctly s+eciied relative to the ima"in" "rid* BS+eciically,
the NI%I ile ret!rned (y dcm2nii 'ill (e oriented C2SD #let)s!+erior)+osterior&, 'hile
the directions coded in the $IC,M itsel 'ill ollo' a C2DS convention #consistent 'ith
the $IC,M standard&E* Similarly, strictly sa"ittal ac0!isitions 'o!ld re0!ire an"!lation
correction #i*e*, rotation& o the (vecs stored in the Siemen.s $IC,Ms as 'ell*
2& I yo! man!ally reorient the ima"in" "rid ollo'in" !se o dcm2nii #e*"*, !sin"
-slreorient2std.&, then yo! also need to acco!nt or that reorientin" on the (vecs as 'ell*

S-ar putea să vă placă și