Neu oImage 262 (2022) 119558
Con en s lis s a ailable a ScienceDi ec
Neu oImage
jou nal homepage: www.else ie .com/loca e/neu oimage
Rep oducible p o ocol o ob ain and measu e fi s -o de elay human
halamic whi e-ma e ac s
Mengxing Liu
a
,
∗
, Ga ikoi z Le ma-Usabiaga
a
,
c
, F ancisco Clascáb
, Ped o M. Paz-Alonso
a
,
c
,
∗
a
BCBL. Basque Cen e on Cogni ion, B ain and Language, Donos ia-San Sebas ián, Spain
b
Depa men o Ana omy and Neu oscience, School o Medicine, Au ónoma de Mad id Uni e si y, Mad id, Spain
c
Ike basque, Basque Founda ion o Science, Bilbao, Spain
a i c l e i n o
Keywo ds:
Thalamus
Thalamoco ical
Ce ebellum
Diffusion MRI
T ac og aphy
Rep oducibili y
a b s a c
The “p ima y ”o “fi s -o de elay ”nuclei o he halamus eed he ce eb al co ex wi h in o ma ion abou on-
going ac i i y in he en i onmen o he subco ical mo o sys ems. Because o he small size o hese nuclei and
he high specifici y o hei inpu and ou pu pa hways, new imaging p o ocols a e equi ed o in es iga e hala-
moco ical in e ac ions in human pe cep ion, cogni ion and language. The goal o he p esen s udy was wo old:
I) o de elop a econs uc ion p o ocol based on in i o diffusion MRI o ex ac and measu e he axonal fibe
ac s ha o igina e o e mina e specifically in indi idual fi s -o de elay nuclei; and, II) o es he eliabili y o
his econs uc ion p o ocol. In le and igh hemisphe es, we in es iga ed he halamoco ical/co ico halamic
axon bundles linking each o he fi s -o de elay nuclei and hei main co ical a ge a eas, namely, he la e al
genicula e nucleus (op ic adia ion), he medial genicula e nucleus (acous ic adia ion), he en al pos e io nu-
cleus (soma osenso y adia ion) and he en al la e al nucleus (mo o adia ion). In addi ion, we examined he
main subco ical inpu pa hway o he en al la e al pos e io nucleus, which o igina es in he den a e nucleus
o he ce ebellum. Ou p o ocol comp ised h ee componen s: defining egions-o -in e es ; p ep ocessing diffu-
sion da a; and modeling whi e-ma e ac s and ac ome y. We hen used compu a ion and es - e es me hods
o check whe he ou p o ocol could eliably econs uc hese ac s o in e es and hei p ofiles. Ou esul s
demons a ed ha he p o ocol had nea ly pe ec compu a ional ep oducibili y and good- o-excellen es - e es
ep oducibili y. This new p o ocol may be o in e es o bo h basic human b ain neu oscience and clinical s udies
and has been made publicly a ailable o he scien ific communi y.
1. In oduc ion
The halamus consis s o wo la ge g ay ma e masses loca ed on
bo h sides o he hi d en icle in he do sal pa o he diencephalon.
The halamus ecei es o de ly inpu s om subco ical senso y, mo o ,
and b ains em modula o y sys ems and i is massi ely in e connec ed
wi h all a eas o he ipsila e al neoco ex in complex and highly specific
pa e ns. Acco dingly, diffe en pa s o he halamus a e in ol ed in di-
e se oles, all o hem essen ial o co ical unc ion. These include he
elay and ga ing o pe iphe al and een an signals o he co ex o sen-
sa ion, a ousal, affec , olun a y mo emen , o language pe cep ion and
p oduc ion, he dynamic con ol o co ical ne wo k ac i i y in a en-
ion, pe cep ual decision-making and sho - e m memo y, as well as he
egula ion o consciousness and sleep s a es ( Guille y & She man, 2002 ;
Halassa & She man, 2019 ; Saalmann & Kas ne , 2011 ).
The s udy o halamic unc ions in humans is a ac ing inc easing
a en ion om he neu oimaging communi y ( Cunningham e al., 2017 ;
∗ Co esponding au ho s a : BCBL, Paseo Mikele egi 69, 2, 20009, Donos ia-San Sebas ián, Spain.
E-mail add esses: [email p o ec ed] (M. Liu), [email p o ec ed] (P.M. Paz-Alonso) .
Czisch e al., 2004 ; Fe nández-Espejo e al., 2010 ; Guye e al., 2003 ;
Kuma e al., 2017 ), no only o u he elucida e i s oles in pe cep-
ion, cogni ion, and mo emen bu also o explo e i s in ol emen in
he pa hogenesis and/o clinical exp ession o b ain diseases ( Clin on &
Meadow-Wood uff, 2004 ; Halassa & She man, 2019 ; K ol e al., 2018 ;
Kwak e al., 2021 ; Scheibel e al., 1997 ; Schiff, 2008 ; Schmi & Ha-
lassa, 2017 ).
Thalamus cells a e g ouped in o se e al disc e e nuclei, each cha -
ac e ized by a specific pa e n o inpu -ou pu wi ing mo i s wi h co i-
cal and subco ical egions ( Acsady, 2022 ; Clascá, 2022 ; Jones, 2007a ).
In asi e s udies wi h axonal ace s in expe imen al animals, includ-
ing non-human p ima es, ha e shown ha neu ons in some o hese
nuclei ecei e subco ical inpu s and elay hem o he ce eb al co -
ex in highly o de ed, poin - o-poin ashion, wi h minimal di e gence
o colla e al b anching. These nuclei a e collec i ely known as fi s -
o de elay ( Acsady, 2022 ; Guille y, 1995 ). The layou o hei con-
nec ions ai h ully p ese es opological in o ma ion wi hin he inpu
h ps://doi.o g/10.1016/j.neu oimage.2022.119558 .
Recei ed 4 Decembe 2021; Recei ed in e ised o m 25 July 2022; Accep ed 11 Augus 2022
A ailable online 13 Augus 2022.
1053-8119/© 2022 The Au ho (s). Published by Else ie Inc. This is an open access a icle unde he CC BY license ( h p://c ea i ecommons.o g/licenses/by/4.0/ )
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
Fig. 1. Scheme o he fibe ac s o igina ing in o a ge ing he senso y / mo o fi s -o de elay nuclei o he halamic ac s ha we e analyzed in his s udy.
Colo lines a e used o indica e he opog aphic o de ing o hese fibe s. A) Axial iew o he igh op ic adia ion. B) Co onal iew o he igh acous ic adia ion.
C) Co onal iew o he igh soma osenso y adia ion. D) Co onal iew o he igh mo o adia ion and den a o halamic ac om he le den a e nucleus (in blue)
(please no e ha hese wo ac s canno be seen in hei en i e y in one single slice).
ma ix om senso y epi helia o mo o pa hways. The e is also e idence
ha unc ions pe o med by fi s -o de elay nuclei a e modula ed as a
unc ion o he na u e o he senso imo o in o ma ion being p ocessed
(e.g., Mihai e al., 2021 ; O’Conno e al., 2002 ). In con as , neu ons
in highe -o de elay nuclei a e mainly d i en by he ce eb al co ex
and p ojec back o i . Thei axons ypically show b anched, poin - o-
mul ipoin mo i s; c ea ing ci cui s sui able o signal b oadcas ing and
synch oniza ion o sepa a e co ical egions, such as hose in ol ed in
a en ion, cogni i e con ol, and sho - e m memo y ( A u e al., 2020 ;
Clascá, 2022 ; Halassa, 2020 ; Mukhe jee e al., 2021 ). In u n, each ha-
lamic nucleus ecei es ecip ocal connec ions om he co ical a eas i
inne a es ( B iggs, 2020 ; Nieuwenhuys e al., 2008 ). These co ico hala-
mic pa hways a e massi e; abou an o de o magni ude mo e nume ous
ha he ascending halamoco ical axons. Bo h axonal sys ems emain
igh ly in e mingled as hey ex end along he in e nal capsule and he
so-called halamic adia ions o he ce eb al whi e ma e (
Jones 2007a ;
Nieuwenhuys e al., 2008 ).
Thei well-known and o de ly layou , obus myelina ion and lim-
i ed di e gence make he fi s -o de elay nuclei inpu and ou pu ac s
ideal subjec s o de eloping and es ing a ep oducible p o ocol o se-
lec i ely ob ain and measu e halamic whi e ma e pa hways. He e,
we se ou o in es iga e he ecip ocal halamoco ical/co ico halamic
fibe ac s o igina ing in ou sepa a e fi s -o de elay nuclei; namely,
he fibe s om he la e al genicula e (known as he op ic adia ion, OR;
Fig. 1 A), om he medial genicula e (acous ic adia ion, AR, Fig. 1 B),
om he en al pos e io nucleus (which is pa o he supe io ha-
lamic adia ion and he e e e ed o as “soma osenso y adia ion ”, SR;
Fig. 1 C), om he en al la e al nucleus (likewise pa o he supe-
io halamic adia ion and he e e e ed o as he “mo o adia ion ”,
MR; Fig. 1 D), as well as he main subco ical inpu ac o he en al
2
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
la e al pos e io nucleus, which o igina es in he den a e nucleus o he
ce ebellum (known as he den a o halamic ac , DT; Fig. 1 D).
Thalamoco ical axons exi ing he la e al genicula e nucleus (LGN)
a ge he p ima y (s ia e) isual co ex along he banks and lips o he
calca ine sulcus, on he medial su ace o he occipi al lobe and occipi al
pole ( Fig. 1 A). A e lea ing LGN om i s os al aspec , he fibe s c oss
he e o- and in alen icula po ions o he in e nal capsule o pass o e
he oo o he la e al en icle empo al ho n and hen u n in e ola e -
ally owa ds he pole o he empo al lobe. The in e io fibe s o he OR
make a sha p loop a ound he ip o he la e al en icle empo al ho n
and a ium o e u n pos e omedially o ming wha is called Meye ’s
loop ( Dayan e al., 2015 ; Ebeling & Reulen, 1988 ; Meye , 1907 ; Wahle -
liiek e al., 1991 ). Along hei pa hs, OR fibe s p ese e he e ino opic
o de ing p esen in he LGN. Fibe s ca ying signals om bo h eyes e-
la ed o he lowe con ala e al isual field quad an adia e along a
do sal and mo e di ec ou e o e mina e in he uppe lip o he cal-
ca ine fissu e, wi h he o ea loca ed a he occipi al end o he fissu e.
Meanwhile, he fibe s ca ying inpu s o m he uppe quad an adia e
mo e in e io ly and ake inc easingly looped ajec o ies, Axons wi h
pe iphe al uppe quad an inpu s ake he longes and mos an e io
loop; lesion o hese an e io fibe s esul s in supe io quad an anopia
(see Ba on e al. 2005 , o a e iew). Co ico halamic axons om V1
ecip oca e hese pa hways in s ic opog aphic o de ( B iggs, 2020 ).
Axons a eling om he en al di ision o he medial genicula e
nucleus (MGN) o he ipsila e al p ima y audi o y co ex (A1, BA 41,
and pa ly 42 in he fi s s aigh (Heschl’s) gy us o he empo al op-
e culum) o m he AR ( Fig. 1 B). These fibe s exi om he os al pole
o he en al MGN, u n la e ally o c oss he in alen icula po ion
o he in e nal capsule and finally an do sally o each A1. The layou
o hese fibe s p ese es he o de ly seg ega ion p esen in he en al
MGN o signals ela ed o he p edominan exci a ion o specific egions
o he cochlea ( “ ono opy ”, see Che ches 2016 ). Func ionally, he AR
is he ga eway o audi o y in o ma ion o he ce eb al co ex; as such,
hese fibe s a e equi ed o conscious sound pe cep ion and o he dis-
c imina ion o speech signals ( Ojemann, 1991 ).
Neu ons in he en opos e io halamic complex (VP) inne a e he
soma osenso y a eas in he pos e io banks and lips o he cen al sul-
cus (B odmann a eas 3, 1 and 2, e e ed o collec i ely as p ima y so-
ma osenso y co ex, S1) in a highly opog aphic ashion ( Fig. 1 C). These
pa hways con ey a di e se a ay o ac ile, he mal, nocicep i e and
p op iocep i e inpu s, and each o hese senso y submodali ies is p e -
e en ially di ec ed o one o he abo e a eas ( Jones & F iedman, 1982 ).
Axons om cells in he medial VP subnucleus elay igeminal in o -
ma ion om he ace and mou h and a ge he la e al and ope cula
po ions o S1, whe eas axons o igina ing in he la e al subnucleus ca y
in o ma ion om he hands, a ms, unk and lowe limbs and a ge do -
sola e al and medial egions o S1 ( e iewed in Jones 2007b ). The VP
axons a el h ough he pos e io limb o he in e nal capsule and he
supe io halamic adia ion in close icini y o he axons o he en o-
la e al (VL) nucleus. The descending co ico halamic axons o igina ing
in S1 inne a e VP in he same opog aphic o de ( Nieuwenhuys e al.,
2008 ).
The pos e io di ision o he en al la e al nucleus (VLN) elays in-
pu s om deep ce ebella nuclei, mainly he den a e nucleus, o he p i-
ma y mo o co ex (M1, BA4) ia he MR ( Ilinsky & Kul as-Ilinsky, 2002 ;
Fig. 1 D). The o de o p ojec ions p ese es he muscula soma o opy
p esen in each o he deep ce ebella nuclei ( Jones, 2007b ). Neu ons in
he medial and en al pa o VLN inne a e he ope cula egion o M1
con olling mou h and ace muscles, while mo e la e o-do sally loca ed
neu ons inne a e p og essi ely mo e do sal and medial egions o M1
con olling he con ala e al hand and unk and lowe limb muscles
( Kelly & S ick, 2003 ).
Inpu s om he deep ce ebella nuclei o he VLN a el ia he
DT ( Fig. 1 D). This fibe bundle o igina es mainly in he den a e nu-
cleus, wi h smalle con ibu ions om he in e posi us and as igial
( Jones, 2007b ; Middle on & S ick, 1997 ; no shown in Fig. 1 D). This
fibe ac exi s he ce ebellum in he supe io peduncle ( b achium con-
junc i um ) and decussa es o he con ala e al mesencephalic egmen-
um o ascend medially o he ed nucleus, whe e i issues nume ous
sho colla e al b anches. The DT fibe s finally en e he VLN om i s
in e ola e al aspec ( Coenen e al., 2014 ; Kwon e al., 2011 ; Pelze e al.,
2017 ). A he unc ional le el, he DT, VLN and MR oge he p o-
ide a low-con e gence and as -conduc ion pa hway ha is belie ed
o eed M1 high- esolu ion e o signals abou he execu ion o mo-
o commands. Such signals a e c ucial o lea ning and flexible adjus -
men o skilled mo o beha io s, including speech ( Kawai e al., 2015 ;
Ojemann, 1975 ). Lesions o he DT can p oduce abno mal mo emen ,
including a axia, emo , and dys onia ( Kwon e al., 2011 ).
P e ious s udies ha e ied o econs uc fi s -o de elay hala-
mic ac s om in i o diffusion da a using diffe en pa ame e s o
he OR (e.g., Bassi e al., 2008 ; Beh ens e al., 2003 ; Benjamin e al.,
2014 ; Mülle -Ax e al., 2017 ; She bondy e al., 2008 ), o he AR (e.g.,
Beh ens e al., 2007 ; Be man e al., 2013 ; Ja ad e al., 2014 ; Maffei e al.,
2018 , 2019; P o an e al., 2014 ; Tschen sche e al., 2019 ), o he SR
(e.g., Be man e al., 2005 ; Kamali e al., 2009 ; Ru land e al., 2020 ;
S aud e al., 2006 ; Sudhyadhom e al., 2013 ), and o he MR and he DT
(e.g., Coenen e al., 2014 ; Hyam e al., 2012 ; Ji e al., 2019 ; Kwon e al.,
2011 ; Meola e al., 2016 ; Nowacki e al., 2019 ; Samma ino e al., 2016 ;
Vo e al., 2015 ). Howe e , o he bes o ou knowledge, no single s udy
has ied o econs uc all o hese fi s -o de elay halamic ac s. In
addi ion, as summa ized below, hese p e ious s udies diffe ed in hei
aims, me hodologies and popula ions and ha e some impo an limi a-
ions.
Se e al s udies ha e epo ed success ul econs uc ion o OR wi h
diffusion da a wi h a a ie y o app oaches ( Bassi e al., 2008 ;
Beh ens e al., 2003 ; Benjamin e al., 2014 ; She bondy e al., 2008 ).
The econs uc ion me hods and pa ame e s adop ed in hese s udies
diffe in se e al ways, such as he seeding s a egy, ac ional aniso opy
h eshold, s eamline leng h and angle es ic ion. S udies ying o iden-
i y he AR ha e encoun e ed difficul ies when using a single fibe ac-
og aphy model because he AR fibe g oup c osses wi h o he fibe
g oups, which esul s in mul iple-o ien a ion signals in oxels. . Fo in-
s ance, Beh ens e al’s (2007) s udy defined he MGN as a cuboid me-
dial o he LGN and s a ed acking om he e o he p ima y audi o y
co ex using a p obabilis ic algo i hm. In his s udy hey ailed o econ-
s uc he AR wi h single-fibe ac og aphy, and only succeeded in do-
ing so when hey used mul i-fibe ac og aphy. In a mo e ecen s udy,
Maffei e al. (2019) explo ed how he DWI acquisi ion and acking pa-
ame e s can affec he econs uc ion o he AR. They ound ha highe
b- alues and mo e g adien di ec ions inc eased he accu acy o econ-
s uc ion o bo h p obabilis ic and de e minis ic acking algo i hms,
bu wi h low b- alues ( ≤ 3000 s/mm
2
) only he p obabilis ic algo i hm
was able o success ully econs uc he AR.
The exis ing ac og aphy s udies on he SR ha e mainly ocused
on clinical and special popula ions, such as Pa kinson pa ien s wi h
essen ial emo ( Sudhyadhom e al., 2013 ), pa ien s wi h b ain le-
sions ( S aud e al., 2006 ), pa ien s wi h igeminal neu algia ( Ru land
e al., 2020 ) o p e e m bo n in an s ( Be man e al., 2005 ). Simila ly,
mos s udies in es iga ing he MR ha e ocused on clinical popula ions,
and o en on pa ien s unde going deep b ain s imula ion (DBS) su ge y
( And ade e al., 2020 ; Pou a ian e al., 2011 ; Riskin-Jones e al., 2021 ;
Samma ino e al., 2016 ). These s udies econs uc ed only pa s o he
MR depending on he pa hological condi ions o hose pa ien s. Fo ex-
ample, And ade e al’s (2020) s udy econs uc ed pa o he MR, using
he en al o alis in e nus nucleus as he seed, in pa ien s who unde -
wen DBS su ge y due o Tou e e synd ome. As he DT a els h ough
deep and small nuclei i is difficul o iden i y wi h DWI echniques.
The e a e only a hand ul o s udies ha ha e success ully econs uc ed
he DT om DWI da a in heal hy ( Kwon e al., 2011 ; Meola e al., 2016 )
o clinical popula ions ( Coenen e al., 2011 , 2014 ; Nowacki e al., 2019 ).
Nowacki e al’s (2019) s udy es ed ou diffe en acking p o ocols o
iden i y he DT, which led o di e gen esul s. All ou me hods we e
3
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
based on de e minis ic algo i hms, bu no p obabilis ic algo i hms we e
es ed. Some s udies ha e also econs uc ed he DT and MR as one sin-
gle ac ( Ji e al., 2019 ; Vo e al., 2015 ).
The p esen s udy aimed o de elop and es a ep oducible p o ocol
o ob aining all o hese fi e fi s -o de elay halamic inpu and ou pu
whi e-ma e ac s. The no el y o his p o ocol capi alizes on 4 aspec s:
(1) I is ocused on well-known whi e-ma e ac s cons i u ed by myeli-
na ed axons ha o igina e and/o a ge he fi s o de elay nuclei o
he halamus, es ing hem wi hin he same s udy, and using simila
me hods and econs uc ion p ocedu es ac oss hem. 2) Diffe en om
mos p e ious s udies, he e we specifically in es iga ed in a la ge da ase
he eliabili y o he p o ocol in e ms o bo h compu a ional and es -
e es ep oducibili y. 3) The p esen p o ocol uses s a e-o - he-a MRI
p o ocols (mul iband, mul i-shell) and ac og aphy me hods wi h he
aim o de eloping an ad anced p o ocol ha can be applied o cu en
ongoing s udies and u u e esea ch. 4) The p o ocol is designed o be
ep oducible, easy o use and au oma ized, which un o una ely has no
been he no m in he pas . Also, i builds on p e ious well- alida ed ools
including he fi s p obabilis ic a las o he halamus based on combin-
ing high- esolu ion ex i o MRI and his ology ( Iglesias e al., 2018 ) and
he ep oducible- ac -p ofiles (RTP2) con aine ized ool which is based
on s a e-o - he-a echniques implemen ed on op o Vis aso ’s code,
which ha e been es ed and used in many publica ions o e he las
15 yea s ( h ps://www.gi hub.com/ is alab/ is aso ; Le ma-Usabiaga
e al., 2022 ). The ul ima e goal o his wo k was o p o ide a eliable
p o ocol o ob aining and es ima ing fi s -o de elay halamic pa h-
ways o basic esea ch and clinical s udies.
To his end, we fi s defined mul iple pa ame e s o op imally econ-
s uc he abo e-men ioned fi e halamic pa hways (i.e., OR, AR, SR,
MR and DT; Fig. 1 ) in le and igh hemisphe es. Second, we es ed
he compu a ional and es - e es ep oducibili y o ou p o ocol by ex-
amining a ange o whi e-ma e p oxies ela ed o he mic os uc u al
and mac os uc u al p ope ies o hese ac s. To examine he elia-
bili y o he p o ocol we ob ained ac s om diffusion-weigh ed im-
ages o 113 no mal adul s. The p o ocol consis ed o h ee componen s:
Defining he egions-o -in e es (ROI); p ep ocessing diffusion-weigh ed
images (DWI) da a; modeling whi e-ma e ac s and ac ome y. Re-
p oducibili y was es ed using wo app oaches: 1) Compu a ional ep o-
ducibili y , es ed by iden i ying each ac using he same pa ame e s 10
independen imes o all 113 subjec s, and 2) Tes - e es ep oducibili y ,
es ed by e-scanning a subse o 24 pa icipan s using he same MRI
p o ocol wice wi hin an a e age in e al o 15 days. Ou hypo hesis
was ha we would ob ain a high deg ee o ep oducibili y o he mi-
c os uc u al and mac os uc u al p ope ies o hese ac s. Howe e ,
we expec ed some a iabili y in specific ac s, and hypo hesized ha
his a iabili y would be highe o es - e es han o compu a ional
ep oducibili y.
2. Me hods
2.1. Subjec s
A o al o 113 heal hy olun ee s (mean age = 24.5 yea s, SD = 4.33
yea s; 65 emales) pa icipa ed in he s udy. Twen y- ou o he ol-
un ee s (mean age = 24.7 yea s, SD = 4.06 yea s; 13 emales) e-
u ned o a second session in which hey we e scanned using exac ly
he same MRI p o ocol (mean in e al = 15 days, SD = 21.82 days,
ange: 7-104 days). All pa icipan s we e igh -handed and had no -
mal o co ec ed- o-no mal ision. No pa icipan had a his o y o ma-
jo medical, neu ological, o psychia ic diso de s. The s udy p o ocol
was app o ed by he E hics Commi ee o he Basque Cen e on Cogni-
ion, B ain and Language (BCBL) and was ca ied ou in acco dance
wi h he Code o E hics o he Wo ld Medical Associa ion (Decla a-
ion o Helsinki) o expe imen s in ol ing human pa icipan s. P io
o hei inclusion in he s udy, all pa icipan s p o ided in o med w i -
en consen . Pa icipan s ecei ed mone a y compensa ion o hei
pa icipa ion.
2.2. Da a acquisi ion
Whole-b ain MRI da a acquisi ion was conduc ed on a 3-T Siemens
P isma Fi whole-body MRI scanne (Siemens Medical Solu ions) using
a 64-channel whole-head coil. The MRI acquisi ion included one T1-
weigh ed s uc u al image (T1w) and DWI sequences. High- esolu ion
MPRAGE T1-weigh ed s uc u al images we e collec ed wi h he ol-
lowing pa ame e s: ime- o- epe i ion (TR) = 2530 ms, ime- o-echo
(TE) = 2.36 ms, flip angle (FA) = 7°, field o iew (FoV) = 256 mm, oxel
size = 1 mm iso opic, 176 slices. In o al 100 diffusion-weigh ed images
we e acqui ed wi h he an e io o pos e io phase-encoding di ec ion
and 50 iso opically dis ibu ed diffusion-encoding g adien di ec ions.
The 100 diffusion weigh ed images included 50 images wi h b- alue o
1000 s/mm
2
and 50 images wi h b- alue o 2000 s/mm
2
. Twel e images
wi h no diffusion weigh ing (b- alue o 0 s/mm
2
) we e ob ained o mo-
ion co ec ion and geome ical dis o ion co ec ion, which comp ised
fi e images wi h he same phase-encoding di ec ion as he DWI images
and se en images wi h he e e sed phase-encoding di ec ion (pos e io
o an e io ). Bo h DWIs and b0 images sha ed he ollowing pa ame e s:
TR = 3600 ms, TE = 73 ms, FA = 78°, oxel size = 2 iso opic, 72 slices
wi h no gap and a mul iband accele a ion ac o o 3.
2.3. T ac og aphy pipeline
T ac econs uc ion was conduc ed using a cus om con aine ized
wo kflow called RTP2 (Rep oducible T ac P ofiles 2) pipeline ( Le ma-
Usabiaga e al., 2019 , 2020 , 2022), which gua an ees da a p o enance
and ep oducibili y. The RTP2-pipeline di ides his p ocess in o h ee
main pa s: (1) ROI defini ion, (2) DWI p ep ocessing, and (3) ac iden-
ifica ion and ac ome y. In s ep 3, we used he p ep ocessed diffusion
da a om s ep 2 o model s eamlines based on he ROIs c ea ed in
s ep 1. De ails a e gi en below. The code and pa ame e s a e a ailable
h ough Gi Hub (gi hub.com/MengxingLiu/Tha ac -pape ) and Docke
Hub ( h ps://hub.docke .com/u/ga ikoi z ).
2.3.1. ROI defini ion
The fi s s ep o he RTP2-pipeline (called RTP2-ana ROIs) in ol es
p ocessing he subjec ’s ana omical T1w image o ob ain he ROIs o be
used in ac og aphy. The inpu o his s ep is he subjec ’s T1w file and
ROIs defined in MNI space; he ou pu is a segmen ed T1w image and
he ROIs o in e es in indi idual subjec T1w space.
The ROIs used o he iden ifica ion o each fibe g oup (see
Table 1 ) we e ob ained as ollows. Fi s , F eesu e ( h p://su e .nm .
mgh.ha a d.edu/ ) was used o pe o m co ical/subco ical segmen a-
ion and pa cella ion. Nex , he halamic nuclei we e ob ained by un-
ning he halamic segmen a ion module implemen ed in F eesu e on a
p obabilis ic a las buil wi h his ological and high- esolu ion ex i o MRI
da a ( Iglesias e al., 2018 ). Fo his s udy, we only conside ed fi s -o de
elay nuclei as ROIs: he LGN, MGN, VP, VLN, and VLp. The VP was de-
fined om he subdi ision o la e al en al pos e io nucleus (VPL) in
he halamic segmen a ion because in Iglesias e al. (2018) his subdi i-
sion combines bo h la e al and medial en al pos e io nuclei. The VLN
was c ea ed by combining he an e io and pos e io en al la e al nu-
clei (VLa and VLp). All ac s desc ibed in his pape a el om hese
halamic nuclei o o he loca ions in he b ain. We nex desc ibe how
we ob ained he ROIs.
To pa cella e he isual co ex we an he Neu opy hy ( Benson &
Winawe , 2018 ; h ps://gi hub.com/noahbenson/neu opy hy ) ool on
he F eesu e esul s. A combina ion o he esul ing V1 and V2 ROIs
was used o ou isual co ex ROI when defining he OR. The OR p o-
jec ions o V1 a e well es ablished (see Rokem e al., 2017 , o a e iew
on OR ac og aphy), whe eas OR p ojec ions o V2 a e ela i ely less
s udied. The e is ace e idence om non-human p ima es sugges ing
4
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
Table 1
Pa ame e s used o iden i y each ac o in e es .
ac name ROI#1 ROI#2 ROI#3 ROI #4
Le OR Le LGN Le V1V2 OR_ oi3_L
∗ -
Righ OR Righ LGN Righ V1V2 OR_ oi3_R
∗ -
Le AR Le MGN Le A1 - -
Righ AR Righ MGN Righ A1 - -
Le SR Le VP Le S1 - -
Righ SR Righ VP Righ S1 - -
Le MR Le VLN Le M1 - -
Righ MR Righ VLN Righ M1 - -
Le DT Le VLp Righ Den a e Righ Thalamus
∗ ∗ Le Ce ebellum Co ex
∗ ∗
Righ DT Righ VLp Le Den a e Le Thalamus
∗ ∗ Righ Ce ebellum Co ex
∗ ∗
Abb e ia ions: VLp = pos e io en al la e al nucleus; V1V2 = p ima y and seconda y
isual co ex.
∗
waypoin ROI, only s eamlines ha pass h ough his ROI we e kep ;
∗ ∗
excluding ROI, s eamlines ha pass h ough his ROI we e excluded.
ha LGN p ojec s beyond V1 ( Ga ey & Powell, 1971 ; Maciewicz, 1975 ;
Mange & Rosa, 2005 ), specifically o V2 (see Bullie & Kennedy, 1983 ,
and Kennedy & Bullie , 1985 ). In- i o ac og aphy e idence om hu-
mans econs uc ing he OR and measu ing he e mina ing poin s sug-
ges ed ha a conside able numbe o s eamlines also e mina ed in V2
( Al a ez e al., 2015 ; A igo e al., 2016 ). Thus, bo h V1 and V2 a e
included in he cu en p o ocol o econs uc he OR. The emaining
ROIs we e ob ained om a lases defined in MNI space. To con e hem
o indi idual subjec space, we fi s pe o med a non-linea egis a ion
o a 1mm
3
MNI empla e using Ad anced No maliza ion Tools (ANTs,
h p://s na a.gi hub.io/ANTs/ ). A1, M1 and S1 we e con e ed om
he human connec ome p ojec (HCP) a las ( Glasse e al., 2016 ). The
S1 was defined by combining A eas 1, 2, 3a and 3b in he HCP a las. The
ce ebella den a e nucleus ROIs we e ob ained and ans o med om he
ce ebella pa cella ion p oposed by Died ichsen e al. (2011) .
To make su e he ROIs ex ended o he in e ace o g ay and whi e
ma e , all ROIs we e dila ed by one cubic oxel. Also, we used inclusion
o exclusion ROIs o imp o e neu oana omical accu acy when ob ain-
ing bo h he OR and DT. Fo he OR, he inclusion ROI was a waypoin ,
fi s d awn in MNI space (co onal plane o y = -80 ha anspasses he
pos e io limb o he in e nal capsule) and hen ans o med o na i e
space, o selec fibe s passing h ough he in e nal capsule. In con as ,
o he DT, wo ROIs, he ipsila e al ce ebella co ex and he con ala -
e al halamus, we e used wi h “NO ”logic, ha is, excluding fibe s ha
passed h ough hem. These we e gene a ed om one o he de aul pa -
cella ions o F eesu e (i.e., apa c + aseg.mgz).
2.3.2. DWI da a p ep ocessing
The second s ep in he RTP2-pipeline (called RTP2-p ep oc) con-
sis ed in p ep ocessing he diffusion da a and egis e ing i in ana om-
ical space. This s ep was mainly based on MR ix’s ( Tou nie e al.,
2019 ) ecommenda ions and used MR ix ools, he ANTs ool desc ibed
abo e and FSL ( Jenkinson e al., 2012 ). The da a was p ep ocessed us-
ing se e al MR ix unc ions in he ollowing s eps: fi s , da a denois-
ing based on andom ma ix heo y, which exploi s da a edundancy in
he pa ch-le el p incipal componen analysis domain ( Co de o-G ande
e al., 2019 ; Ve aa e al., 2016 ) using dwidenoise ; second, Gibbs Ring-
ing co ec ion ( Kellne e al., 2016 ) using m degibbs ; hi d, suscep ibili y
induced dis o ions and mo ion co ec ion wi h he FSL’s opup and eddy
ools ( Smi h e al., 2004 ) called by dwi slp ep oc ; ou h, B1 field inho-
mogenei y co ec ion wi h dwibiasco ec and Rician backg ound noise
emo al wi h m calc ; fi h, a igid ans o ma ion ma ix o align he
DWI images o he co esponding T1w image using ANTs.
To make su e he DWI da a om es and e es sessions we e in
he same space, DWI da a om bo h sessions we e aligned o he same
T1w collec ed in he ini ial es session. The e o e, bo h es and e es
sessions used he same ROIs; only DWI p ep ocessing and he s eamline
acking we e session-specific.
2.3.3. T ac iden ifica ion and ac ome y
In he hi d and final main s ep, he con aine RTP2-pipeline was
used o ob ain he final whi e-ma e ac s. This con aine used he ROIs
and p ep ocessed DWI da a o sys ema ically iden i y he ac s o in e -
es bila e ally: OR, AR, SR, MR and DT. We ini ially an his s ep wi h
4 andom subjec s. A e ca e ully checking he econs uc ion o each
ac in line wi h he neu oana omical and neu ophysiological li e a u e
(see Nieuwenhuys e al. 2008 , o a e iew), he expe neu oana omis
(FC), who has mo e han 30 yea s o esea ch expe ience in he s udy
o halamoco ical p ojec ions, e alua ed and guided he i e a ion o
diffe en pa ame e s, such as inclusion/exclusion ROIs, acking algo-
i hm, s eamline cleaning, e ce e a. I he econs uc ed fibe bundles
we e no compa ible wi h he knowledge o he neu oana omis (e.g.,
he pa hway, hickness o he bundle, e mina ing loca ion e ce e a),
we adjus ed ou pa ame e s acco dingly, un il he bundle was as close
as possible o he known neu oana omy.
We fi s modeled he diffusion in o ma ion o ob ain a map o
possible di ec ions wi h weigh s, called fibe o ien a ion dis ibu ions
(FODs), o e e y oxel, using MR ix3’s mul i- issue cons ained sphe -
ical decon olu ion (CSD; Jeu issen e al. 2014 ). This ool can dis-
ce n c ossing fibe s and p o ide mo e han one di ec ion in each
oxel. Nex , s eamline ac og aphy was pe o med on he es ima ed
fibe o ien a ion dis ibu ions using a p obabilis ic algo i hm (iFOD2;
Tou nie e al. 2010 ) wi h he ollowing pa ame e s: s ep size 1mm,
maximum fibe leng h 200mm, minimum fibe leng h 20mm, FODs
ampli ude h eshold 0.05, angle h eshold 45 deg ees. The ROIs used
in he s eamline ac og aphy a e desc ibed in Table 1 (please no e
ha he den a o halamic ac is a c oss-hemisphe ic ac ; in he
p esen s udy we defined he hemisphe e o his ac based on he
halamic hemisphe e o he con enience o he desc ip ion). The
s eamlines we e seeded om ROI#1 and e mina ed in ROI#2 o
all ac s excep he OR, which combined s eamlines om bo h
di ec ions o limi he impac o olume diffe ences be ween seed
and a ge . Ma lab u ili ies de eloped in Vis aso ( h ps://gi hub.
com/ is alab/ is aso ) we e used o emo e he ou lie s eamlines
om each ac gene a ed in MR ix and o ob ain he main ac
me ics.
We gene a ed along- ac p ofiles using he ac me ics ob ained
om Vis aso . Al hough CSD was used o model he fibe s because i
disce ns c ossing fibe s, classical diffusion enso (DTI) modeling was
used o ob ain he ypical diffusion summa y s a is ics, such as ac ional
aniso opy (FA), axial diffusi i y (AD), mean diffusi i y (MD), and a-
dial diffusi i y (RD). To gene a e ac p ofiles, we ob ained he cen al
loca ion o all he s eamlines in he ac and sampled his as 100 same-
leng h segmen s. We hen summa ized he diffusion p ope ies o each
segmen by aking a weigh ed a e age o he diffusion p ope ies co -
esponding o a disc cen e ed in he segmen . Finally, an along- ac
p ofile was gene a ed o each ac using Vis aso .
5
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
Fig. 2. The ep oducibili y measu emen scheme. A) Compu a ional ep oducibili y ( ep oducibili y ac oss compu a ions); es - e es ep oducibili y ( ep oducibili y
ac oss es and e es sessions). B) The Dice o e lap o MR econs uc ed a he fi s and second compu a ions, om subjec S038. C) Co ela ion o he FA p ofile o
MR om subjec S038’s es and e es sessions.
2.4. Rep oducibili y measu emen
Rep oducibili y was measu ed in wo diffe en ways ( Fig. 2 ). Fi s ,
o es compu a ional ep oducibili y , we epea ed he ac og aphy e-
cons uc ion using he RTP2-pipeline 10 imes. Ou aim was o measu e
how consis en ly ou pipeline gene a ed he ac s o in e es . Second, o
check o es - e es ep oducibili y, we used a subse o 24 pa icipan s,
who e u ned o a second acquisi ion session wi hin a mean empo al
in e al o 15 days. We examined how consis en ly ou pipeline gene -
a ed hei ac s a hese wo diffe en ime poin s.
To e alua e hese wo ypes o ep oducibili y a he mic os uc u al
scale, we pe o med pai wise co ela ions on ac p ofiles o all possi-
ble pai s om he 10 epea ed compu a ions o measu e compu a ional
ep oducibili y, and ac oss es and e es o measu e es - e es ep o-
ducibili y. Fo simplici y, we only show he co ela ions o he FA al-
ues ( o esul s wi h o he me ics please see he Supplemen a y ma e-
ial).
A he mac os uc u al le el, we quan i a i ely analyzed ac ol-
ume o e lap, s eamline densi y and dis ance o check he ep oducibil-
i y o ac shapes. These analyses included: (1) Dice simila i y index o
check o olume-based o e lap o all ac pai s; (2) densi y co ela ion
o he oxel-le el s eamline densi y o all ac pai s; and, (3) bundle
adjacency, he a e age dis ance be ween s eamlines om wo ac s.
The measu emen s used o examine compu a ional ep oducibili y and
es - e es ep oducibili y we e compu ed using he package scilpy (see
de ails in Schilling e al., 2021 and h ps://gi hub.com/scilus/scilpy ).
These analyses we e conduc ed ac oss all possible pai s o compu a-
ional ep oducibili y and es - e es ep oducibili y, as well as o each
ac .
3. Resul s
In he p esen s udy, we ob ained and measu ed fibe s bundles con-
nec ing ou fi s -o de senso y (LGN, MGN), soma osenso y (VP) and
mo o (VLN) halamic nuclei wi h hei main co esponding co ical a -
ge a eas. In addi ion, we econs uc ed he subco ical inpu pa hway
o VLp om he den a e nucleus o he ce ebellum. These fi e ac s
we e iden ified as homologous ac pai s in he le and he igh hemi-
sphe e. Figs. 3 and 4 show hese ac s in a ep esen a i e subjec (see
also supplemen a y ideos showing 3D o a ions o he main ac s s ud-
ied, as well as supplemen a y Figu e S1 o ac s om fi e mo e andom
subjec s). To examine he ep oducibili y o ou p o ocol, we ollowed
a double analy ical app oach es ing: (1) compu a ional ep oducibili y
by epea ing he compu a ion on he same diffusion da a 10 imes and
quan i ying changes om compu a ion o compu a ion o he same
ac ; and, (2) es - e es ep oducibili y , by ob aining DWI da a om he
same subjec s and using he same MRI p o ocols ac oss wo diffe en
sessions o quan i y es - e es changes in he same ac s.
3.1. Compu a ional ep oducibili y
Fo he fi e pai s o whi e-ma e fibe s used in he es ablished p o o-
col, he epea ed compu a ions on same diffusion da a esul ed in almos
6
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
Fig. 3. The OR (A) and AR (B) econs uc ed in a ep esen a i e subjec . A1 and B1 show he 3D ep esen a ions o he OR and AR in yellow. A2 and B2 show he
posi ions om which he slices in A3 and B3 a e d awn. A3 and B3 depic axial and co onal iews o he co e subcomponen s o OR and AR, espec i ely. G een
colo is used o co ical ROIs V1/V2 and A1.
7
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
Fig. 4. The SR (A) and MR/DT (B) econs uc ed in a ep esen a i e subjec . A1 and B1 show he 3D ep esen a ions o he SR and he MR/DT. A2 and B2 show he
posi ions om which he slices in A3, B3 and B4 a e d awn. A3 and B3 depic co onal iews o he co e subcomponen s o SR, MR and DT. G een colo is used o
he co ical ROIs S1 and M1. Yellow s eamlines in A and B a e SR and MR, espec i ely. Blue s eamlines in B ep esen he DT. B4 shows he axial iew o he co e
subcomponen s o he DT.
8
M. Liu, G. Le ma-Usabiaga, F. Clascá e al. Neu oImage 262 (2022) 119558
Table 2
Rep oducibili y indices and hei s anda d de ia ions (in pa en heses) o all measu es and fibe bundles.
compu a ional es - e es
FA p ofile co ela ion bundle adjacency dice index densi y co ela ion FA p ofile co ela ion bundle adjacency dice index densi y co ela ion
L OR 0.9996(0.0016) 0.09(0.01) 0.92(0.01) 0.998(0.001) 0.9956(0) 0.11(0.01) 0.90(0.01) 0.990(0.005)
R OR 0.9996(0.0005) 0.09(0.01) 0.92(0.01) 0.998(0.001) 0.9944(0) 0.11(0.01) 0.90(0.01) 0.989(0.005)
L AR 0.9976(0.0074) 0.18(0.09) 0.85(0.05) 0.990(0.008) 0.9265(0.12) 0.39(0.27) 0.76(0.09) 0.907(0.082)
R AR 0.9989(0.0013) 0.11(0.09) 0.90(0.04) 0.997(0.002) 0.9473(0.07) 0.34(0.41) 0.80(0.12) 0.940(0.047)
L SR 0.9993(0.0005) 0.12(0.01) 0.89(0.01) 0.995(0.001) 0.9908(0.01) 0.13(0.02) 0.88(0.01) 0.987(0.006)
R SR 0.9991(0.0007) 0.130.01) 0.88(0.01) 0.995(0.001) 0.9851(0.02) 0.14(0.02) 0.87(0.01) 0.986(0.004)
L MR 0.9985(0.0012) 0.09(0.01) 0.91(0.01) 0.993(0.001) 0.9787(0.02) 0.13(0.05) 0.89(0.01) 0.971(0.010)
R MR 0.9982(0.0017) 0.09(0.01) 0.91(0.01) 0.993(0.001) 0.9713(0.03) 0.12(0.03) 0.89(0.01) 0.971(0.014)
L DT 0.9988(0.0038) 0.07(0.03) 0.93(0.02) 0.994(0.004) 0.9638(0.03) 0.21(0.11) 0.82(0.08) 0.853(0.097)
R DT 0.9986(0.0032) 0.06(0.01) 0.94(0.01) 0.994(0.004) 0.9458(0.06) 0.22(0.14) 0.82(0.09) 0.782(0.182)
Fig. 5. E alua ion o compu a ional ep oducibili y. A) Examples o g oup a e age FA p ofiles o le MR om he fi s (g ay con inuous line) and second (g een
dashed line) compu a ions. The ligh g een shaded a ea indica es he s anda d de ia ion. B) S ip plo s showing he dis ibu ion o co ela ion coefficien s be ween all
possible pai s compu ed o each ac and each subjec (ligh e colo columns ep esen he le hemisphe e, da ke colo columns ep esen he igh hemisphe e).
Each do ep esen s he co ela ion coefficien o a specific compu a ion pai o one pa icipan . C) Ag eemen indices dis ibu ion: bundle adjacency ( op), Dice
coefficien (middle), and densi y co ela ion (below) o all possible compu a ion pai s o each ac and each subjec (ligh e colo columns ep esen he le
hemisphe e, da ke colo columns ep esen he igh hemisphe e).
iden ical ac p ofiles and high ag eemen on s eamlines. Mean co e-
la ions o he FA p ofile we e abo e 0.99 o all ac s examined (see
Table 2 ; Fig. 5 A). A an indi idual le el, co ela ion coefficien s we e
highe han 0.97 o each fibe ac oss all possible compu a ion pai s,
excep o le AR, which had a long ail ha eached 0.82 ( Fig. 5 B).
Bea in mind o 10 epea ed compu a ions, ha will esul in a leas
9 ela i ely low coefficien s, e en i only one compu a ion esul ed in a
diffe en ac han all he o he s.
Ag eemen indices also showed ha he iden ified whi e-ma e
fibe s had consis en shapes and densi y ac oss epea ed compu a ions
( Table 2 ). Fig. 3 C shows ag eemen indices o each indi idual pai o
compu a ions. These h ee ag eemen indices e ealed a simila pa e n
as ha obse ed o co ela ion coefficien s o he FA p ofile ( Fig. 5 B),
wi h mo e a iabili y in le AR. A simila pa e n was also ound o
he homologous igh AR. I is no ewo hy ha some o his a iabili y
in ag eemen indices de i es om he same single subjec (e.g., ou ly-
ing clus e s o bundle adjacency and Dice coefficien in igh AR, and
o Dice coefficien and densi y co ela ion in le OR). Bila e al DT also
showed mo e a iabili y in densi y co ela ion, bu less a iabili y in
bundle adjacency and Dice coefficien . The co ela ion coefficien s o
AD, MD, and RD ac oss all possible pai s o compu a ions a e desc ibed
in he Supplemen a y ma e ial (Figu e S1).
3.2. Tes - e es ep oducibili y
To examine es - e es ep oducibili y, 24 pa icipan s came back o
a e es session whe e we used exac ly he same MRI p o ocol. The mean
o FA p ofile co ela ions was abo e 0.9 ac oss he en ac s o in e es ,
al hough as expec ed he alues we e nume ically lowe han hose ob-
se ed in he compu a ional ep oducibili y analyses. As in he compu a-
ional ep oducibili y analysis, he le AR also showed highe a iabili y
in he es - e es ep oducibili y analysis, wi h lowe alues wi hin he
mean co ela ion coefficien s (0.93, Table 2 and Fig. 6 A & B). Ne e -
heless, i is impo an o highligh ha all o hese alues eflec a high
deg ee o ep oducibili y.
Tes - e es ep oducibili y was also confi med by he ag eemen in-
dices. The g oup a e ages o bundle adjacency we e all unde 0.4 o
he en ac s, indica ing ha he s eamlines iden ified in he es we e
e y close o he s eamlines iden ified in he e es (see Table 2 and
Fig. 6 C). High ep oducibili y was also eflec ed by he Dice index and
9
