BOLD MRI a 9.4T wi h 3D s ack-o -spi als
eadou s
Alejand o Mon eal-Mad igal1, Denizhan Ku ban1, Desmond H.Y. Tse2, Dimo I ano 1, Nicolas
Boulan 3, Benedik A. Pose 1
1Maas ich B ain Imaging Cen e, Facul y o Psychology and Neu oscience, Maas ich Uni e si y, Maas ich ,
The Ne he lands; 2Scannexus BV, Maas ich , The Ne he lands; 3Uni e si y Pa is-Saclay, CEA, CNRS, BAOBAB,
Neu oSpin, Gi -su -Y e e, F ance
Abs ac
In his wo k we in es iga e he use o spi al eadou s o sub-millime e BOLD MRI a 9.4T,
and e i y simula ions o he BOLD PSF wi h unc ional expe imen s. We con i m ha a TE < T2* can
be employed o spi al-ou eadou s wi hou comp omising BOLD sensi i i y. The sho e TE p o ides
a educed TR, imp o emen in SNR and minimizes o - esonance e ec s. We conclude ha spi al-in
eadou s a e mos ly use ul o lowe esolu ions a UHF, bu ha segmen ed spi al-ou eadou s can
play an impo an ole in mesoscopic BOLD MRI a ul a high ields (> 7T).
1. INTRODUCTION
Func ional MRI ( MRI) is an MRI modali y ha non-in asi ely measu es b ain ac i i y. I was
i s in oduced as Blood Oxygena ion Le el Dependen (BOLD) [1] con as , ollowing he
obse a ion ha changes in blood deoxyhemoglobin con en du ing neu onal ac i a ion al e he
MRI signal. Typical spa ial and empo al esolu ions o human MRI expe imen s ha e been on he
o de o a couple o millime e s and seconds, espec i ely.
The desi e o highe esolu ions in MRI has been one o he main d i e s o ul a-high ield
MRI [2–4], inc easing he ield s eng h p o ides a nea ly quad a ic inc ease in SNR [5]. G adien coils
wi h highe g adien ampli udes and slew a es a e also being de eloped [6, 7], p o iding signi ican
speed-up in echo-plana and spi al eadou s. Addi ionally, ecei e coils wi h high channel coun RF
ecei e a ays [8] a e being de eloped, allowing highe pa allel accele a ion by k-space
unde sampling. These ha dwa e ad ances oge he wi h new sequence de elopmen s ha e made i
possible o achie e sub-millime e esolu ions and sub-second acquisi ions. These ools can be used
by neu oscien is s o explo e he human b ain a high empo al and spa ial esolu ions wi h
unp eceden ed de ail.
BOLD MRI is an indi ec measu e o neu onal ac i a ion, i measu es he Hemodynamic
Response Func ion (HRF) which is a combina ion o di e en physiological p ocesses: Ce eb al Blood
Flow (CBF), Ce eb al Blood Volume (CBV) and CMRO2 (oxygen consump ion) [9]. Non-BOLD con as s
ha e been p oposed o measu e hose e ec s mo e di ec ly, such as calib a ed MRI o measu e
CMRO2 [10, 11], A e ial Spin Labeling (ASL) [12, 13] o CBF, and Vascula Space Occupancy (VASO)
[14, 15] o CBV. A Ul a High Field (UHF) howe e , hese pose di e en implemen a ion challenges,
o example, ASL and VASO make use o RF p epa a ion modules ha inc ease he minimum
achie able epe i ion ime (TR), and equi e a high use o RF powe o he sa u a ion and in e sion
pulses [2]. Fo hese easons, BOLD con as is s ill he mos widely used in MRI s udies also a UHF.
In unc ional imaging, small changes in T2*, caused by he local change in deoxyhemoglobin
concen a ion ha accompanies neu onal ac i a ion mani es and hus he highes BOLD signal
di e ence be ween ac i a ion and es is obse ed when TE ~ T2* o he g ay ma e [16]. The mos
commonly used eadou s a egy o BOLD MRI is Ca esian echo-plana imaging (EPI) [17], which
achie es high sampling e iciency and enables olume epe i ion imes on he o de o seconds a
he ypically desi ed esolu ions o MRI applica ions. In he pas decade, he inco po a ion o
CAIPIRINHA [18] accele a ion in o 2D simul aneous mul i-slice [19–21] and 3D EPI [22–24] has
imp o ed EPI and u he inc eased i s neu oscien i ic use.
Al e na i es o EPI o as imaging a e non-Ca esian eadou s, o which he mos common
a e spi al and adial [25, 26]. O he non-Ca esian eadou s ha ha e been p oposed o MRI can be
ound in [27–29]. Spi al eadou s wi h ull k-space sampling we e in oduced o BOLD MRI a 3T
[30–32]. Following ad ances in spi al econs uc ion and o - esonance co ec ion [33, 34], spi al
imaging has been e isi ed a 7T, by Engel e al. [35] and o 2D BOLD MRI by Kaspe e al. [36].
Recen ly, spi al eadou s ha e also been used o balanced SSFP BOLD in wo k by Valsala e al. [37],
in which he au ho s exploi ed he lexibili y o choose TE o imp o e he unc ional con as . Spi als
ha e also been used o di e en non-BOLD con as s such as ASL [38] and VASO [39], whe e a sho
echo ime is desi ed o imp o e he inhe en low SNR o hese echniques. Since hese ypes o
eadou s do no all in o a Ca esian g id, he use o he Fas Fou ie T ans o m is no possible; he
econs uc ion is hen ypically pe o med o line and i is a compu a ionally expensi e p ocess.
The simples design o a spi al eadou ollows an A chimedean spi al. In his case, an en i e
k-space plane (2D) is acqui ed a e an exci a ion (single-sho ). Each plane can also be acqui ed using
mo e exci a ions (mul i-sho ) which sho ens each eadou and educes ulne abili y o B0 o -
esonance, howe e a he cos o inc eased olume TR. The k-space can be ans e sed om he
cen e o he edge (spi al-ou ) o s a ing om he edge and mo ing inwa ds (spi al-in). A
combina ion o bo h (spi al-ou -in) acqui es wo echos pe exci a ion. The cen e o he k-space can
be sampled mo e densely han he ou e pa ( a iable densi y spi als). S acking spi al planes in he
pa i ion di ec ion p oduces a 3D s ack-o -spi als pa e n. To imp o e he unde sampling beha iou ,
k-space planes can be o a ed [40] . All hese pa ame e s a ec he con as , acquisi ion du a ion
and o e all image quali y and ca e ul selec ion o hem is impo an o a success ul implemen a ion
o spi al imaging.
The widesp ead use o spi al eadou s has been limi ed due o p ac ical challenges in hei
implemen a ion. One o hem has been he need o accoun o ajec o y de ia ions as well as eddy
cu en s, which equi es knowledge o he ajec o y execu ed by he scanne . The ajec o y can be
measu ed wi h imaging echniques [41], by use o special ha dwa e such as ield came as [34], o
measu emen o he G adien Impulse Response Func ion [42] which can subsequen ly be applied o
p edic he achie ed g adien wa e o m. Ano he challenge is he econs uc ion p ocess, which
ypically makes use o he compu a ionally demanding Non-Uni o m Fou ie T ans o m NUFFT [43],
and hence o en imp ac ical o implemen on he scanne 's own image econs uc ion ha dwa e.
Finally, B0 o - esonance esul s in spi al image blu ing wi h a 2D poin sp ead ha is much ha de
o co ec o han he 1D phase-encoding dis o ion in EPI, which can be add essed wi h
s aigh o wa d pos -p ocessing (e.g. FSL op-up [44]). The con inuously changing phase encoding
di ec ion and a e o k-space a e sal in spi als cause a complex phase acc ual and blu ing ha
need o be add essed du ing he image econs uc ion [45], conside ably adding ex a
compu a ional demands and econs uc ion imes.
To a oid o - esonance e ec s due o B0 ield inhomogenei ies when using long spi al
eadou s, an o - esonance equency e m needs o be included in he SENSE signal model [46]:
𝑆𝛾(𝑡) = ∫𝑉𝑐𝛾(𝑟)𝑚(𝑟)𝑒−𝑖𝑘(𝑡)𝑟𝑒−𝑖𝛥𝜔0(𝑟)𝑡𝑑𝑉 (1)
whe e s( ) is he k-space da a om each ecei e channel ɣ, c( ) is he complex spa ial sensi i i y o
he ɣ coil, m( ) is he magne iza ion, k( ) is he k-space ajec o y, and he e m Δω0 is he angula
o - esonance equency, p opo ional o ield inhomogenei ies. Di e en me hods o o - esonance
co ec ion exis , such as ime segmen ed [47], mul i- equency in e pola ion [48], and g idding
app oaches [49]. Equa ion 1 can be disc e ized and a linea sys em o equa ions be c ea ed wi h he
k-space da a:
𝑠 = 𝐸𝑚 , 𝐸 = 𝑐𝛾(𝑟)𝑒−𝑖𝑘(𝑡)𝑟𝑒−𝑖𝛥𝜔0(𝑟)𝑡 (2)
This sys em o equa ions can be sol ed i e a i ely wi h a egula ized leas -squa es cos
unc ion:
𝑚 = 𝑎𝑟𝑔𝑚𝑖𝑛1
2||𝐸𝑚 − 𝑠||2
2+𝛽
2𝑅(𝑚), (3)
whe e β is he egula iza ion pa ame e and R(m) is a egula iza ion unc ion ha is used o ensu e
ha he algo i hm con e ges o a s able solu ion, common egula iza ion unc ions a e L1 no m, L2
no m and o al a ia ion (TV) [50]. Di e en algo i hms exis o sol e equa ion 3, some o he mos
widely used a e CG-SENSE [33], ADMM [51], and FISTA [52].
In BOLD MRI he highes con as be ween ac i a ion and es is obse ed a a TE
app oxima ely equal o T2*. Engel e al. [53] p oposed he concep o he BOLD poin sp ead unc ion
(PSF) o be e cha ac e ize he impac o he eadou and TE on BOLD unc ional expe imen s. Using
simula ions, hey showed ha a TE ~ T2* achie es he highes BOLD sensi i i y wi h spi al eadou s.
In MR Fou ie encoding, he poin sp ead unc ion (PSF) is he in e se Fou ie o he k-space
il e H(k):
𝑃𝑆𝐹(𝑟) = Ƒ−1(𝐻(𝑘)) (4)
wi h 𝑘 = [𝑘𝑥,𝑘𝑦,𝑘𝑧]𝑇 . The k-space il e H(k) comp ises he sampling o k-space a disc e e
poin s on a ini e suppo and a weigh ing unc ion:
𝐻(𝑘) = 𝐻𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔(𝑘) ⋅ 𝐻𝑤𝑒𝑖𝑔ℎ𝑡(𝑘) (5)
In g adien echo expe imen s, a mono-exponen ial signal decay weigh ing 𝐻𝑤𝑒𝑖𝑔ℎ𝑡
𝑇2∗is
imposed. In a BOLD expe imen , he measu e o in e es is he small di e ences a ising om
empo al changes in T2*. The BOLD weigh il e 𝐻𝑤𝑒𝑖𝑔ℎ𝑡
𝐵𝑂𝐿𝐷 is he de i a i e o he T2* weigh il e :
𝐻𝑤𝑒𝑖𝑔ℎ𝑡
𝑇2∗(𝑘) = 𝑒𝑥𝑝−𝑡(𝑘)/𝑇2∗ ,𝐻𝑤𝑒𝑖𝑔ℎ𝑡
𝐵𝑂𝐿𝐷 (𝑘) = 𝜕𝐻𝑤𝑒𝑖𝑔ℎ𝑡
𝑇2∗(𝑘)
𝜕𝑇2∗=𝑡(𝑘)
𝑇2
∗2 𝑒𝑥𝑝−𝑡(𝑘)/𝑇2∗ (6)
BOLD imaging is hen cha ac e ized by a di e en ial PSF:
𝑃𝑆𝐹𝐵𝑂𝐿𝐷(𝑥,𝑇2
∗,𝛥𝑇2
∗) = 𝑃𝑆𝐹𝐺𝐸(𝑥,𝑇2∗ +𝛥𝑇2
∗) − 𝑃𝑆𝐹𝐺𝐸(𝑥,𝑇2∗) = 𝜕𝑃𝑆𝐹𝐺𝐸(𝑥,𝑇2∗)
𝜕𝑇2∗𝛥𝑇2
∗
= Ƒ−1(𝐻𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔𝐻𝑤𝑒𝑖𝑔ℎ𝑡
𝐵𝑂𝐿𝐷 )𝛥𝑇2
∗ = Ƒ−1(𝐻𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔(𝑘)𝑡(𝑘)
𝑇2
∗2 𝑒𝑥𝑝−𝑡(𝑘)/𝑇2∗)𝛥𝑇2
∗ (7)
In his wo k we combine some o he p e iously men ioned ha dwa e and sequence
ad ances. We use a 3D s ack-o -spi al sequence o BOLD MRI on a 9.4T scanne , equipped wi h
high-pe o mance head-only g adien s o mo e e icien k-space co e age. We expe imen ally e i y
he BOLD poin sp ead unc ion simula ions by Engel e al. [53] o spi al eadou s and u he
in es iga e he bene i s o spi al eadou s o sub-millime e MRI a UHF. We also p o ide he
comple e pipeline o sequence, econs uc ion and analysis o spi al MRI da a.
The goal o his p ojec was o be e cha ac e ize he adeo s o spi al BOLD a lamina
esolu ion and gain insigh s in o he challenges when aiming o ye highe esolu ions (< 0.6 mm)
MRI and ield s eng hs (> 9.4T) as en isaged by he AROMA p ojec . Wi h u he sho ening o T2*
and inc eases in B0 inhomogenei ies single-sho EPI will also become inc easingly challenging a high
esolu ion wi h he p ohibi i ely long eadou p ecluding sui ably sho TE, unless using
segmen a ion which comes a TR and SNR penal ies.
2. METHODS
Simula ions o he BOLD poin sp ead unc ion we e pe o med o a ange o T2* alues
(T2*= 20-26 ms) o de e mine a sui able TE o 0.6 mm iso opic esolu ion BOLD MRI wi h dual-sho
spi al-ou (DS-SO) and dual-sho spi al-in (DS-SI) eadou s. In p e ious wo k we concluded ha high-
esolu ion single-sho spi al-ou images a 9.4T su e se e e o - esonance e ec s and a e no
usable [54], o his eason we ocus on dual-sho spi al eadou s in his wo k. Echo imes anging
om 2 o 30 ms we e used o he spi al-ou simula ions, and 30 o 50 ms o spi al-in.
Figu e 1: BOLD imaging il e (HBOLD) schema ic depic ion and main componen s o he BOLD poin sp ead unc ion om a
spi al-ou eadou . (a) k-space s HBOLD and c oss-sec ion. (b) Real pa o he BOLD PSF and i s sec ions. The main lobe is
di ided in o nominal main lobe (nominal oxel wid h) and esidual main lobe. Side lobes include bo h he nega i e and
posi i e ones.
The eal pa o he PSF was di ided in h ee di e en sec ions: nominal main lobe, esidual
main lobe and side lobes (Figu e 1b), which we e used o cha ac e ize each PSF simila ly o [53, 55].
BOLD esolu ion (FWHM o main lobe), sensi i i y (in eg al o e he nominal main lobe), and
speci ici y ( ela ion o he in eg al o e nominal main lobe and he in eg al o he absolu e alue o
he esidual main lobe and side lobes) we e used as me ics:
𝑟𝑒𝑠𝐵𝑂𝐿𝐷 = 𝐹𝑊𝐻𝑀𝑃𝑆𝐹 , 𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 = ∑𝑛𝑜𝑚.𝑚𝑎𝑖𝑛 𝑙𝑜𝑏𝑒 , (8)𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 =
∑ 𝑛𝑜𝑚.𝑚𝑎𝑖𝑛 𝑙𝑜𝑏𝑒
∑|𝑠𝑖𝑑𝑒 𝑙𝑜𝑏𝑒𝑠| +∑| 𝑟𝑒𝑠.𝑚𝑎𝑖𝑛 𝑙𝑜𝑏𝑒|
Signal- o-noise a io es ima es we e also ob ained, he SNR o an image is p opo ional o:
𝑆𝑁𝑅 ∝ 𝑥𝑦𝑧√𝑛𝑠ℎ𝑜𝑡𝑠𝑇𝑎𝑐𝑞 𝑓(𝑇𝑅,𝑇𝐸) (9)
𝑓(𝑇𝑅,𝑇𝐸) ∝ (1−𝐸)𝑠𝑖𝑛(𝛼)
1−𝑐𝑜𝑠(𝛼)𝐸 𝑒
−𝑇𝐸
𝑇2
∗, 𝐸 = 𝑒−𝑇𝑅
𝑇1 (10)
whe e x, y, z a e he oxel dimensions, nsho s is he numbe o in e lea es (lines o segmen s pe
plane), Tacq is he du a ion o he acquisi ion eadou and (TR,TE) is a unc ion desc ibing he
dependence on he elaxa ion cha ac e is ics o he issue o g adien echo.
Da a we e acqui ed on a Siemens 9.4T scanne equipped wi h a 16Tx-31Rx head coil [56] and
an AC-84II head-only g adien wi h 80 mT/m peak ampli ude and 330 T/m/s slew a e. When using
spi al eadou s co e ing a ci cula k-space, an a ea simila o ha o a squa e k-space should be used
o achie e he desi ed esolu ion. In his wo k, we used a k-space adius which is 1.13 imes la ge
han he nominal one as p oposed in [57]. To do so, an in-plane esnom=0.54x0.54x0.60 mm3 was
used, esul ing in an e ec i e esolu ion ese =0.6 mm iso opic. The spi al eadou s we e designed
wi h a a iable densi y 𝛼=1.3, in-plane unde -sampling Rxy=3.3 and a 120° o a ion o segmen s
be ween planes. An op imal g adien design algo i hm [58] was used o achie e he desi ed k-space
ajec o y in he sho es ime wi h Gmax=50 mT/m, SRmax=250 T/m/s and BW=312 kHz.
A o al o nine heal hy olun ee s we e scanned a e p o iding w i en consen ollowing
he p o ocols o he local e hics commi ee. Fou olun ee s we e scanned o alida e he BOLD
poin sp ead unc ion simula ions. Fi e olun ee s we e la e scanned o assess he ela i e u ili y o
spi al-in and spi al-ou eadou s. To add ess he B1+ inhomogenei ies in he isual co ex, we used
p e-calcula ed RF shim se ings (phase-only), p e iously ob ained as an a e age o se e al subjec s.
To alida e he BOLD PSF simula ions, a sequence wi h a dual-sho spi al-ou (DS-SO) s ack-
o -spi als eadou and wo di e en nominal echo imes (6 and 12 ms) was implemen ed in Pulseq
[59]. The echo ime o 6 ms was selec ed as a good comp omise be ween he di e en quali y
me ics and he 12 ms one since simula ions sugges i gi es he highes sensi i i y a T2*=22 ms,
expec ed alue o GM a 9.4T [60]. Budde e al. [61] epo ed an a e age GM T2* o 28.3 ± 6.8 ms a
9.4T, ob ained om ME-GRE scans wi h 0.35x0.35x2.00 mm3 esolu ion. Plane pa i ions we e
in e lea ed be ween echo imes esul ing in e ec i ely simul aneous acquisi ion o he MRI BOLD
esponse a he wo echo imes. Fa sa u a ion was applied be o e each exci a ion. The sequence
u he mo e con ained a 1 ms long na iga o module consis ing o wo 0.3 ms non-phase encoded
FID 0.4 ms apa , applied a TE=2 ms. The ollowing pa ame e s we e used: FOV=140x140x18 mm3,
30 kz pa i ions wi h linea encoding o de , TE1/TE2 6/12 ms, TRsho =45 ms, TR olume=2.7 s,TRpai =5.41 s,
FA=15° and BWTP=25. This esul ed in a eadou du a ion o 27 ms pe sho . The sequence used RF
spoiling wi h quad a ic phase inc emen s and g adien spoiling in h ee axes. A schema ic depic ion
o his sequence can be ound in igu e 2a.
Se en 15 minu es unc ional uns we e acqui ed ( h ee olun ee s unde wen 2 uns, and
ano he one 1 un). A ull-sc een licke ing (app ox. 8Hz) black and whi e adial checke boa d
p og ammed in Psychopy [62] was used as s imuli. Each block consis ed o 7 TRpai o es (38 s)
ollowed by 7 TRs o ac i a ion (38 s), e ec i ely acqui ing 14 olumes in each block (7 pe echo
ime); each un consis ed o 154 olumes pe TE. The block design, s imuli and slice posi ioning can
be ound in igu e 2e.
To u he in es iga e he pe o mance o di e en spi al eadou s, h ee olun ee s we e
scanned wi h a dual-sho spi al-ou (DS-SO) and dual-sho spi al-in (DS-SI) sequences. Mos o he
pa ame e s we e he same as desc ibed abo e, excep o FOV=140x140x24 mm3 and 40 kz
pa i ions. The DS-SO had a TE=6 ms, TRsho =48 ms, TR olume=3.88 s, FA=11° and a eadou du a ion o
27 ms (pe sho ) whe eas he DS-SI had a TE=31 ms, TRsho =47 ms, TR olume=3.76 s, FA=11° and a
eadou du a ion o 27 ms (pe sho ). The TE=6 ms ul ima ely was chosen ollowing he simula ions
which show ha i is a easonable comp omise be ween he di e en me ics used in his wo k.
To es he acquisi ion o a la ge FOV, wo addi ional olun ee s we e scanned wi h he DS-
SO sequence as desc ibed, and in addi ion wi h doubled FOVpa i ion (48 mm ins ead o 24 mm), bu
wi h Rxy=3, Rz=2 o achie e a simila TR and 8% phase o e sampling o a oid oldo e in he cen e
slices; all o he pa ame e s we e kep cons an . Since a co onal iew p o ides limi ed sensi i i y
changes in he AP di ec ion, a kz blip was applied be ween sho s o imp o e he aliasing pa e n in
he pa i ion di ec ion. This esul ed in a CAIPIRINHA-like [18] pa e n wi h shi Δ=1 along kz,
e ec i ely esul ing in an unde sampling o Rxy=6 and Rz=1. A schema ic depic ion o hese
sequences and k-space ajec o ies can be ound in igu e 2 b,c,d.
Each unc ional un las ed app oxima ely 12 minu es, wi h he block design consis ing o 8
TRs o es ollowed by 8 TRs o ac i a ion. Fo DS-SO and DS-SI , his esul ed in 31 s and 30 s block
du a ion, espec i ely, and 192 olumes we e acqui ed o each un. The same ull-sc een licke ing
checke boa d was used o he BOLD PSF expe imen s.
A 3D low- esolu ion ully-sampled mul i-echo GRE acquisi ion wi h segmen ed spi al-ou
eadou was acqui ed be o e each unc ional un and used o calcula e sensi i i y and B0 maps. This
sequence was FOV ma ched o he MRI scans and used he ollowing pa ame e s:
esnom=1.1x1.1x0.6 mm3, TE1/TE2/TE3=2.3/6.5/10.7 ms, TR=62 ms ( o SAR es ic ions), FA=11° and
BWTP=25. Spi al pa ame e s: 32 sho s, Rxyz=1 , a iable densi y 𝛼=1, esul ing in a eadou du a ion
o 3.2 ms (pe sho ) and o al scan du a ion o 1.5 and 3 minu es o he 24 mm and 48 mm pa i ion
FOV, espec i ely.
Figu e 2: Schema ic depic ion o sequences, k-space ajec o y, s imuli and slice posi ioning. (a) Sequence used o he
BOLD PSF expe imen s. The pa i ions o bo h TEs a e acqui ed in e lea ed (e.g. TE1 (blue): 1s sho , 1s pa i ion -> TE1
(blue): 2nd sho , 1s pa i ion -> TE2 (pu ple): 1s sho , 1s pa i ion -> TE2 (pu ple): 2nd sho , 1s pa i ion, e c). A a sa u a ion
module is used be o e he i s exci a ion o each pa i ion. The DS-SO (b) and DS-SI (c) sequence used o he spi al-in and
spi al-ou expe imen s. A a sa u a ion module is used be o e each exci a ion. (d) schema ic depic ion o he k-space
ajec o ies used in his wo k, o bo h he FOVpa i ion=24 mm (Rz=1) and he FOVpa i ion=48 mm (Rz=2), o he la e , a blip
is applied a e each sho o achie e CAIPIRINHA-like k-space co e age in he pa i ion (kz) di ec ion. (e) FOV o ien a ion,
and pa adigm block-design using a licke checke boa d.
Image econs uc ion was pe o med o line using a pipeline based on he one p esen ed in
[39], using Neu odesk [63]. The i s echo o he low- esolu ion ME-GRE scan was used o calcula e
he sensi i i y maps using he ESPiRIT [64] implemen a ion in MRIReco.jl [65]. These maps we e hen
comp essed o 16 i ual channels [66]. The h ee echoes we e used o calcula e he o - esonance
map using he egula ized ield map es ima ion [67] as implemen ed in MRIFieldmaps.jl.
The GIRF p edic ed ajec o y and k0 modula ions we e es ima ed om he nominal
ajec o y using p e iously acqui ed g adien impulse esponse unc ion (GIRF) da a ollowing he
me hod desc ibed in [68]. Dynamic o - esonance co ec ion in k-space (DORK) [69] was applied o
he aw da a o co ec o scanne d i . The phase o he i s FID na iga o o each olume was
compa ed o he second one o he MRI imese ies and any phase di e ence was emo ed om he
aw da a. The DORK co ec ed unc ional aw da a was me ged wi h he GIRF p edic ed ajec o y
in o an MRD ile [70].
The sensi i i y maps, o - esonance maps and MRD ile we e he inpu o he econs uc ion
pe o med in MRIReco.jl using an ADMM sol e and “L1” egula iza ion, including a densi y
compensa ion unc ion [71] and mul i- equency in e pola ion B0 co ec ion [48]. The numbe o
i e a ions and s opping c i e ia was se o 30 and 1.2x10-7, espec i ely. The numbe o bins was
selec ed as 4𝛥𝜔𝑚𝑎𝑥𝑇
𝜋, whe e 𝛥𝜔𝑚𝑎𝑥 is he maximum (absolu e) o - esonance equency and 𝑇 is
he eadou du a ion, esul ing in 16 bins o he 27 ms long eadou . Recons uc ion was pe o med
on a dedica ed se e unning Ubun u 22.04, wi h 1 TB o memo y and wo In el Xeon Scalable
P ocesso “Skylake” Gold 6140 wi h 18 co es each, o a o al o 36 co es. Figu e 3 shows an
o e iew o he acquisi ion, econs uc ion and analysis pipeline.
The da a analysis pipeline was implemen ed in Neu odesk [63] and consis ed o : (1) Mo ion
co ec ion using AFNI’s 3dAllinea e command [72], he ime se ies was egis e ed o he second
olume o he acquisi ion. (2) High-pass empo al il e using FSL wi h an FWHM o he du a ion o
es /ac i i y blocks. (3) Mean imese ies and SNR we e compu ed as quali y me ics using AFNI. (4)
Ac i a ion maps we e compu ed using a Gene al Linea Model (GLM) and clus e ed using AFNI’s
3dDecon ol e and 3dclus commands. BOLD sensi i i y in e ms o BOLD con as o empo al noise
a io ( CNR) was calcula ed o each oxel a e age block as CNR = Δs/σn, wi h Δs = mean BOLD signal
change (mean ac i a ion - mean es ) and σn = s anda d de ia ion o GLM esidual. No spa ial
smoo hing was applied.
Figu e 3: O e iew o he 3D s ack-o -spi als acquisi ions, econs uc ion, and analysis pipeline. Sequence: he Pulseq
sequences a e gene a ed in MATLAB; nominal ajec o ies and pa ame e dic iona ies a e sa ed. The nominal ajec o ies
a e used o ob ain he GIRF p edic ed ones. Fully sampled, highly segmen ed low- esolu ion GRE-ME and spi al unc ional
da a a e acqui ed; bo h da a se s a e con e ed in o MRD iles. Recons uc ion: The DORK-co ec ed spi al aw da a is
me ged wi h he GIRF p edic ed ajec o y and o he pa ame e s in o an MRD ile. Sensi i i y and B0 maps a e calcula ed
om he low- esolu ion ME-GRE scan. The MRD ile, sensi i i y and B0 maps a e he inpu o he econs uc ion in
MRIReco.jl. Analysis: Rigid mo ion co ec ion is pe o med wi h AFNI, high-pass il e in FSL and he s a is ical analysis is
done wi h AFNI. The Recons uc ion and analysis is done in Neu odesk.
Decla a ion o compe ing in e es
The au ho s ha e no compe ing in e es s o epo .
Funding
This wo k has ecei ed inancial suppo om he Eu opean Union Ho izon 2020 Resea ch and
Inno a ion p og am unde g an ag eemen no. 885876 (AROMA).
Supplemen a y ma e ial
Supplemen a y igu e 1: 0.8 mm iso opic da a. (a) mean imese ies, (b) e ec i e SNR and (c) ac i a ion maps. Single-sho
spi al-ou images p esen se e e o - esonance e ec s e en a e co ec ion, dual-sho acquisi ions ha e good image
quali y. Fo his esolu ion, he dual-sho spi al-in acquisi ion has good SNR and ac i a ion, hese esul s di e om he
0.6 mm iso opic da a. Addi ional de ails can be ound in [54].
REFERENCES
1. Ogawa S, Tank DW, Menon R, Elle mann JM, Kim SG, Me kle H, Ugu bil K (1992) In insic signal
changes accompanying senso y s imula ion: unc ional b ain mapping wi h magne ic
esonance imaging. P oc Na l Acad Sci 89:5951–5955.
2. I ano D, De Ma ino F, Fo misano E, F i z FJ, Goebel R, Hube L, Kashyap S, Kempe VG, Ku ban D,
Roeb oeck A, Sengup a S, So ge B, Tse DHY, Uludağ K, Wiggins CJ, Pose BA (2023) Magne ic
esonance imaging a 9.4 T: he Maas ich jou ney. Magn Reson Ma e Phys Biol Med
36:159–173.
3. Sadeghi-Ta akameh A, DelaBa e L, Lago e RL, To ado-Ca ajal A, Wu X, G an A, Ad iany G,
Me zge GJ, Van de Moo ele P-F, Ugu bil K, A ala E, E yaman Y (2020) In i o human head
MRI a 10.5T: A adio equency sa e y s udy and p elimina y imaging esul s. Magn Reson
Med 84:484–496.
4. Boulan N, Que ie L, Aube G, Amadon A, Belo gey J, Be iaud C, Bonnelye C, B edy Ph, Chazel E,
Dilasse G, Dubois O, Giacomini E, Gilg ass G, G as V, Guiha d Q, Janno V, Jus e FP, Lannou
H, Lep ê e F, Le man C, Le S e C, Luong M, Maucondui F, Molinié F, Nunio F, Scola L,
Sinanna A, Touze y R, Véd ine P, Vignaud A, he Iseul Conso ium (2023) Commissioning o
he Iseul CEA 11.7 T whole-body MRI: cu en s a us, g adien –magne in e ac ion es s and
i s imaging expe ience. Magn Reson Ma e Phys Biol Med 36:175–189.
5. Le S e C, G an A, Van de Moo ele P-F, Mon eal-Mad igal A, Ad iany G, Vignaud A, Maucondui
F, Rab ai -Le man C, Pose BA, Uğu bil K, Boulan N (2022) Magne ic ield s eng h
dependen SNR gain a he cen e o a sphe ical phan om and up o 11.7T. Magn Reson Med
88:2131–2138.
6. Feinbe g DA, Becke AJS, Vu AT, S ockmann J, Hube L, Ma S, Ahn S, Se sompop K, Cao X, Pa k S,
Liu C, Wald LL, Polimeni JR, Ma eyam A, G ube B, S i nbe g R, Liao C, Yacoub E, Da ids M,
Bell P, Rumme E, Koehle M, Po has A, Gonzalez-Insua I, S ocke S, Gunamony S, Die z P
(2023) Nex -gene a ion MRI scanne designed o ul a-high- esolu ion human b ain imaging
a 7 Tesla. Na Me hods 20:2048–2057.
7. Huang SY, Wi zel T, Keil B, Scholz A, Da ids M, Die z P, Rumme E, Ramb R, Ki sch JE, Yendiki A,
Fan Q, Tian Q, Ramos-Llo dén G, Lee H-H, Nummenmaa A, Bilgic B, Se sompop K, Wang F,
A am AV, Komlosh M, Benjamini D, Magdoom KN, Pa hak S, Schneide W, No iko DS,
Fie emans E, Tounek i S, Mekkaoui C, Augus inack J, Be ge D, Shapson-Coe A, Lich man J,
Basse PJ, Wald LL, Rosen BR (2021) Connec ome 2.0: De eloping he nex -gene a ion ul a-
high g adien s eng h human MRI scanne o b idging s udies o he mic o-, meso- and
mac o-connec ome. Neu oImage 243:118530.
8. Wiggins GC, Polimeni JR, Po has A, Schmi M, Alagappan V, Wald LL (2009) 96-Channel ecei e-
only head coil o 3 Tesla: Design op imiza ion and e alua ion. Magn Reson Med 62:754–
762.
9. Uludağ K (2023) Physiological modeling o he BOLD signal and implica ions o e ec i e
connec i i y: A p ime . Neu oImage 277:120249.
10. Da is TL, Kwong KK, Weissko RM, Rosen BR (1998) Calib a ed unc ional MRI: Mapping he
dynamics o oxida i e me abolism. P oc Na l Acad Sci 95:1834–1839.
11. Hoge RD, A kinson J, Gill B, C elie GR, Ma e S, Pike GB (1999) In es iga ion o BOLD signal
dependence on ce eb al blood low and oxygen consump ion: The deoxyhemoglobin
dilu ion model. Magn Reson Med 42:849–863.
12. Alsop DC, De e JA, Golay X, Gün he M, Hend ikse J, He nandez-Ga cia L, Lu H, MacIn osh BJ,
Pa kes LM, Smi s M, an Osch MJP, Wang DJJ, Wong EC, Zaha chuk G (2015) Recommended
implemen a ion o a e ial spin-labeled pe usion MRI o clinical applica ions: A consensus
o he ISMRM pe usion s udy g oup and he Eu opean conso ium o ASL in demen ia.
Magn Reson Med 73:102–116.
13. Calaman e F, Thomas DL, Pell GS, Wie sma J, Tu ne R (1999) Measu ing Ce eb al Blood Flow
Using Magne ic Resonance Imaging Techniques. J Ce eb Blood Flow Me ab 19:701–735.
14. Lu H, Golay X, Peka JJ, an Zijl PCM (2003) Func ional magne ic esonance imaging based on
changes in ascula space occupancy. Magn Reson Med 50:263–274.
15. Lu H, Hua J, an Zijl PCM (2013) Nonin asi e unc ional imaging o ce eb al blood olume wi h
ascula -space-occupancy (VASO). NMR Biomed 26:932–948.
16. Bande ini PA, Wong EC, Hinks RS, Tiko sky RS, Hyde JS (1992) Time cou se EPI o human b ain
unc ion du ing ask ac i a ion. Magn Reson Med 25:390–397.
17. Mans ield P (1977) Mul i-plana image o ma ion using NMR spin echoes. J Phys C Solid S a e
Phys 10:L55.
18. B eue FA, Blaime M, Muelle MF, Seibe lich N, Heidemann RM, G iswold MA, Jakob PM (2006)
Con olled aliasing in olume ic pa allel imaging (2D CAIPIRINHA). Magn Reson Med
55:549–556.
19. Se sompop K, Gagoski BA, Polimeni JR, Wi zel T, Wedeen VJ, Wald LL (2012) Blipped-con olled
aliasing in pa allel imaging o simul aneous mul islice echo plana imaging wi h educed g -
ac o penal y: Blipped-CAIPI o Simul aneous Mul islice EPI. Magn Reson Med 67:1210–
1224.
20. Feinbe g DA, Se sompop K (2013) Ul a- as MRI o he human b ain wi h simul aneous mul i-
slice imaging. J Magn Reson 229:90–100.
21. Ba h M, B eue F, Koopmans PJ, No is DG, Pose BA (2016) Simul aneous mul islice (SMS)
imaging echniques. Magn Reson Med 75:63–81.
22. Na sude M, Gallichan D, an de Zwaag W, G ue e R, Ma ques JP (2016) Th ee-dimensional
echo plana imaging wi h con olled aliasing: A sequence o high empo al esolu ion
unc ional MRI: 3D-EPI-CAIPI: A Sequence o High Tempo al Resolu ion MRI. Magn Reson
Med 75:2350–2361.
23. Pose BA, Kempe VG, I ano D, Kannengiesse SA, Uludag K, Ba h M (2013) CAIPIRINHA-
accele a ed 3D EPI o high empo al and/o spa ial esolu ion EPI acquisi ions. P oc.
ESMRMB Cong . Toulouse F . p 226
24. S i nbe g R, S öcke T (2021) Segmen ed K-space blipped-con olled aliasing in pa allel imaging
o high spa io empo al esolu ion EPI. Magn Reson Med 85:1540–1551.
25. Dela e BMA, Heidemann RM, C owe LA, Vallée J-P, Hyacin he J-N (2010) Spi al demys i ied.
Magn Reson Imaging 28:862–881.
26. Feng L (2022) Golden-Angle Radial MRI: Basics, Ad ances, and Applica ions. J Magn Reson
Imaging 56:45–62.
27. Zahneisen B, G o z T, Lee KJ, Ohlendo S, Reise M, Zai se M, Hennig J (2011) Th ee-
dimensional MR-encephalog aphy: Fas olume ic b ain imaging using ose e ajec o ies.
Magn Reson Med 65:1260–1268.
28. Asslände J, Zahneisen B, Hugge T, Reise M, Lee H-L, LeVan P, Hennig J (2013) Single sho
whole b ain imaging using sphe ical s ack o spi als ajec o ies. Neu oImage 73:59–70.
29. Laza us C, Weiss P, Chau e N, Maucondui F, Bo laende M, Vignaud A, Ciuciu P SPARKLING:
No el Non-Ca esian Sampling Schemes o Accele a ed 2D Ana omical Imaging a 7T Using
Comp essed Sensing.
30. Lai S, Glo e GH (1998) Th ee-dimensional spi al MRI echnique: A compa ison wi h 2D spi al
acquisi ion. Magn Reson Med 39:68–78.
31. Glo e GH, Law CS (2001) Spi al-in/ou BOLD MRI o inc eased SNR and educed suscep ibili y
a i ac s. Magn Reson Med 46:515–522.
32. Hu Y, Glo e GH (2007) Th ee-dimensional spi al echnique o high- esolu ion unc ional MRI.
Magn Reson Med 58:947–951.
33. P uessmann KP, Weige M, Bö ne P, Boesige P (2001) Ad ances in sensi i i y encoding wi h
a bi a y k -space ajec o ies: SENSE Wi h A bi a y k -Space T ajec o ies. Magn Reson Med
46:638–651.
34. Ba me C, Zanche ND, P uessmann KP (2008) Spa io empo al magne ic ield moni o ing o MR.
Magn Reson Med 60:187–197.
35. Engel M, Kaspe L, Ba me C, Schmid T, Vionne L, Wilm B, P uessmann KP (2018) Single‐sho
spi al imaging a 7 T. Magn Reson Med 80:1836–1846.
36. Kaspe L, Engel M, Heinzle J, Muelle -Sch ade M, G aedel NN, Rebe J, Schmid T, Ba me C, Wilm
BJ, S ephan KE, P uessmann KP (2022) Ad ances in spi al MRI: A high- esolu ion s udy wi h
single-sho acquisi ion. Neu oImage 246:118738.
37. Iyyappan Valsala P, Veldmann M, Bosch D, Sche le K, Ehses P Submillime e balanced SSFP
BOLD– unc ional MRI accele a ed wi h 3D s ack-o -spi als a 9.4 T. Magn Reson Med. doi:
10.1002/m m.30064
38. Ku ban D, I ano D, Kashyap S, Hube L, Libe man G, Pose BA (2022) Concu en CBF and BOLD
MRI wi h dual-echo spi al simul aneous mul i-slice acquisi ions a 7T. Neu oImage
247:118820.
39. Mon eal-Mad igal A, Ku ban D, Hube L, I ano D, Boulan N, Pose BA (2024) Combining he
bene i s o 3D acquisi ions and spi al eadou s o VASO MRI a UHF. Imaging Neu osci 2:1–
14.
40. Deng W, Zahneisen B, S enge VA (2016) Ro a ed s ack-o -spi als pa ial acquisi ion o apid
olume ic pa allel MRI: S ack-o -Spi als, CAIPIRINHA, GRAPPA. Magn Reson Med 76:127–
135.
41. Duyn JH, Yang Y, F ank JA, an de Veen JW (1998) Simple Co ec ion Me hod o k-Space
T ajec o y De ia ions in MRI. J Magn Reson 132:150–153.
42. Vannesjo SJ, Haebe lin M, Kaspe L, Pa an M, Wilm BJ, Ba me C, P uessmann KP (2013) G adien
sys em cha ac e iza ion by impulse esponse measu emen s wi h a dynamic ield came a:
G adien Sys em Cha ac e iza ion wi h a Dynamic Field Came a. Magn Reson Med 69:583–
593.
43. Fessle JA, Su on BP (2003) Nonuni o m as Fou ie ans o ms using min-max in e pola ion.
IEEE T ans Signal P ocess 51:560–574.
44. Ande sson JLR, Ska e S, Ashbu ne J (2003) How o co ec suscep ibili y dis o ions in spin-echo
echo-plana images: applica ion o di usion enso imaging. Neu oImage 20:870–888.
45. Haskell MW, Nielsen J-F, Noll DC (2022) O - esonance a i ac co ec ion o magne ic esonance
imaging: a e iew.
46. P uessmann KP, Weige M, Scheidegge MB, Boesige P (1999) SENSE: Sensi i i y encoding o
as MRI. Magn Reson Med 42:952–962.
47. Noll DC, Meye CH, Pauly JM, Nishimu a DG, Maco ski A (1991) A homogenei y co ec ion
me hod o magne ic esonance imaging wi h ime- a ying g adien s. IEEE T ans Med
Imaging 10:629–637.
48. Man L-C, Pauly JM, Maco ski A (1997) Mul i equency in e pola ion o as o - esonance
co ec ion. Magn Reson Med 37:785–792.
49. Knopp T, Egge s H, Dahnke H, P es in Jü, Senegas J (2009) I e a i e O -Resonance and Signal
Decay Es ima ion and Co ec ion o Mul i-Echo MRI. IEEE T ans Med Imaging 28:394–404.
50. Fessle JA (2010) MODEL-BASED IMAGE RECONSTRUCTION FOR MRI. IEEE Signal P ocess Mag
27:81–89.
51. Boyd S, Pa ikh N, Chu E, Pelea o B, Ecks ein J (2011) Dis ibu ed Op imiza ion and S a is ical
Lea ning ia he Al e na ing Di ec ion Me hod o Mul iplie s. Found T ends® Mach Lea n
3:1–122.
52. Beck A, Teboulle M (2009) A Fas I e a i e Sh inkage-Th esholding Algo i hm o Linea In e se
P oblems. SIAM J Imaging Sci 2:183–202.
53. Engel M, Kaspe L, P uessmann KP (2022) BOLD PSF: Impac o k-space sampling on T2* con as .
P oc J Annu Mee ISMRM-ESMRMB abs ac :2446.
54. (2024) Book o Abs ac s ESMRMB 2024 Online 40 h Annual Scien i ic Mee ing 2–5 Oc obe
2024. Magn Reson Ma e Phys Biol Med 37:1–781.
55. Feizollah S, Ta di CL (2023) High- esolu ion di usion-weigh ed imaging a 7 Tesla: Single-sho
eadou ajec o ies and hei impac on signal- o-noise a io, spa ial esolu ion and
accu acy. Neu oImage 274:120159.
56. Shajan G, Kozlo M, Ho mann J, Tu ne R, Sche le K, Pohmann R (2014) A 16-channel dual- ow
ansmi a ay in combina ion wi h a 31-elemen ecei e a ay o human b ain imaging a
9.4 T. Magn Reson Med 71:870–879.
57. Pipe JG, Zwa NR (2014) Spi al ajec o y design: A lexible nume ical algo i hm and base
analy ical equa ions. Magn Reson Med 71:278–285.
58. Lus ig M, Kim S-J, Pauly JM (2008) A as me hod o designing ime-op imal g adien wa e o ms
o a bi a y k-space ajec o ies. IEEE T ans Med Imaging 27:866–873.
59. Lay on KJ, K obo h S, Jia F, Li in S, Yu H, Leupold J, Nielsen J-F, S öcke T, Zai se M (2017)
Pulseq: A apid and ha dwa e-independen pulse sequence p o o yping amewo k: Rapid
Ha dwa e-Independen Pulse Sequence P o o yping. Magn Reson Med 77:1544–1552.
60. No is DG, Ladd ME (2023) Chap e 37 - New ho izons: Human MRI a ex emely high ield
s eng hs. In: Ma ken o h Bloch K, Guye M, Pose BA (eds) Ad . Magn. Reson. Technol. Appl.
Academic P ess, pp 577–592
61. Budde J, Shajan G, Ho mann J, Uğu bil K, Pohmann R (2011) Human imaging a 9.4 T using T2*-,
phase-, and suscep ibili y-weigh ed con as . Magn Reson Med 65:544–550.
62. Pei ce JW (2007) PsychoPy—Psychophysics so wa e in Py hon. J Neu osci Me hods 162:8–13.
63. Ren on AI, Dao TT, Johns one T, Ci ie O, Sulli an RP, Whi e DJ, Lyons P, Slade BM, Abbo DF,
Amos TJ, Bollmann S, Bo ing A, Campbell MEJ, Chang J, Close TG, Dö ig M, Ecks ein K, Egan
GF, E as S, Flandin G, Ga ne KG, Ga ido MI, Ghosh SS, G igna d M, Halchenko YO, Hannan
AJ, Heins eld AS, Hube L, Hughes ME, Kaczma zyk JR, Kaspe L, Kuhlmann L, Lou K, Man illa-
Ramos Y-J, Ma ingley JB, Meie ML, Mo is J, Na ayanan A, Pes illi F, Puce A, Ribei o FL,
Rogasch NC, Ro den C, Schi a MM, Shaw TB, Sowman PF, Spi z G, S ewa AW, Ye X, Zhu JD,
Na ayanan A, Bollmann S (2024) Neu odesk: an accessible, lexible and po able da a
analysis en i onmen o ep oducible neu oimaging. Na Me hods 1–5.
64. Uecke M, Lai P, Mu phy MJ, Vi ue P, Elad M, Paeuly JM, Vasanawala SS, Lus ig M (2014)
ESPIRiT-an eigen alue app oach o au ocalib a ing pa allel MRI: Whe e SENSE mee s
GRAPPA. Magn Reson Med 71:990–1001.
65. Knopp T, G osse M (2021) MRIReco.jl: An MRI econs uc ion amewo k w i en in Julia. Magn
Reson Med 86:1633–1646.
66. Huang F, Vijayakuma S, Li Y, He el S, Duensing GR (2008) A so wa e channel comp ession
echnique o as e econs uc ion wi h many channels. Magn Reson Imaging 26:133–141.
67. Lin CY, Fessle JA (2020) E icien Regula ized Field Map Es ima ion in 3D MRI. IEEE T ans Compu
Imaging 6:1451–1458.
68. Co ec ion o B0 eddy cu en e ec s in spi al MRI - Robison - 2019 - Magne ic Resonance in
Medicine - Wiley Online Lib a y.
h ps://onlinelib a y.wiley.com/doi/ ull/10.1002/m m.27583. Accessed 27 Ma 2023
69. P eu e J, Van de Moo ele P-F, Ugu bil K, Hu X, Glo e GH (2002) Co ec ion o physiologically
induced global o - esonance e ec s in dynamic echo-plana and spi al unc ional imaging.
Magn Reson Med 47:344–353.
70. Ina i SJ, Naegele JD, Zwa NR, Roopchansingh V, Lizak MJ, Hansen DC, Liu C, A kinson D, Kellman
P, Koze ke S, Xue H, Campbell‐Washbu n AE, Sø ensen TS, Hansen MS (2017) ISMRM Raw
da a o ma : A p oposed s anda d o MRI aw da ase s. Magn Reson Med 77:411–421.
71. Pipe JG, Menon P (1999) Sampling densi y compensa ion in MRI: Ra ionale and an i e a i e
nume ical solu ion. Magn Reson Med 41:179–186.
72. Cox RW (1996) AFNI: So wa e o Analysis and Visualiza ion o Func ional Magne ic Resonance
Neu oimages. Compu Biomed Res 29:162–173.
73. Gao J-H, Liu H-L (2012) In low e ec s on unc ional MRI. Neu oImage 62:1035–1039.
74. Ren all V, Nangini C, Ha i R (2014) All ha gli e s is no BOLD: inconsis encies in unc ional MRI.
Sci Rep 4:3920.
75. Quan i a i e assessmen o blood in low e ec s in unc ional MRI signals - Gao - 1996 - Magne ic
Resonance in Medicine - Wiley Online Lib a y.
h ps://onlinelib a y.wiley.com/doi/abs/10.1002/m m.1910360219. Accessed 17 Feb 2025
76. Huang F, Vijayakuma S, Li Y, He el S, Duensing GR (2008) A so wa e channel comp ession
echnique o as e econs uc ion wi h many channels. Magn Reson Imaging 26:133–141.
77. Die ich BE, B unne DO, Wilm BJ, Ba me C, G oss S, Kaspe L, Haebe lin M, Schmid T, Vannesjo
SJ, P uessmann KP (2016) A ield came a o MR sequence moni o ing and sys em analysis.
Magn Reson Med 75:1831–1840.
78. Zhao B, Se sompop K, Adals einsson E, Gagoski B, Ye H, Ma D, Jiang Y, Ellen G an P, G iswold
MA, Wald LL (2018) Imp o ed magne ic esonance inge p in ing econs uc ion wi h low-
ank and subspace modeling. Magn Reson Med 79:933–942.
79. Cao X, Liao C, Iye SS, Wang Z, Zhou Z, Dai E, Libe man G, Dong Z, Gong T, He H, Zhong J, Bilgic B,
Se sompop K (2022) Op imized mul i-axis spi al p ojec ion MR inge p in ing wi h subspace
econs uc ion o apid whole-b ain high-iso opic- esolu ion quan i a i e imaging. Magn
Reson Med 88:133–150.
