Head motion during fMRI tasks is reduced in children and adults if participants take breaks

De elopmen al Cogni i e Neu oscience 44 (2020) 100803 A ailable online 29 May 2020 1878-9293/© 2020 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/). Head mo ion du ing MRI asks is educed in child en and adul s i pa icipan s ake b eaks Tobias W. Meissne a , b , *, Jon Walb in c , d , Ma isa No d b , e , Kami Koldewyn c , Sa ah Weigel a a TU Do mund Uni e si y, Facul y o Rehabili a ion Sciences, Depa men o Vision, Visual Impai men s & Blindness, Emil-Figge-S . 50, 44227, Do mund, Ge many b Ruh Uni e si y Bochum, Facul y o Psychology, Uni e si € a ss . 150, 44801, Bochum, Ge many c Bango Uni e si y, School o Psychology, De elopmen al Social Vision Lab, Pen all Road, Bango , LL57 2AS, Wales, Uni ed Kingdom d Uni e si y o Coimb a, Facul y o Psychology and Educa ion Sciences, P oac ion Lab, Rua Col� egio No o, 3000-115, Coimb a, Po ugal e S an o d Uni e si y, Psychology Depa men , 450 Se a Mall, S an o d, CA, 94305, USA ARTICLE INFO Keywo ds: Mo emen Da a quali y Pedia ic neu oimaging Mock scanne aining S udy design De elopmen ABSTRACT Head mo ion emains a challenging con ound in unc ional magne ic esonance imaging ( MRI) s udies o bo h child en and adul s. Mos pedia ic neu oimaging labs ha e de eloped expe ience-based, child- iendly s anda ds conce ning e.g. he maximum leng h o a session o he ime be ween mock scanne aining and ac ual scanning. Howe e , i is unclea which ac o s o child- iendly neu oimaging app oaches a e e ec i e in educing head mo ion. He e, we in es iga e h ee main ac o s including (i) ime lag o mock scanne aining o he ac ual scan, (ii) p io scan ime, and (iii) ask engagemen in a da ase o 77 child en (aged 6–13) and 64 adul s (aged 18–35) using a mul ile el modeling app oach. In child en, dis ibu ing MRI da a acquisi ion ac oss mul iple same-day sessions educes head mo ion. In adul s, mo ion is educed a e inside-scanne b eaks. Despi e hese posi i e e ec s o spli ing up da a acquisi ion, mo ion inc eases o e he cou se o a s udy as well as o e he cou se o a un in bo h child en and adul s. Ou esul s sugges ha spli ing up MRI da a acquisi ion is an e ec i e ool o educe head mo ion in gene al. A he same ime, di e en ways o spli ing up da a acquisi ion bene i child en and adul s. 1. In oduc ion “Please emembe : Relax and y no o mo e. He e we go.” In unc ional magne ic esonance imaging ( MRI) expe imen s, his is o en he las hing he esea che says be o e s a ing he scanne and he expe imen . Wha is le o do is hoping o good quali y da a. As MRI is e y suscep ible o head mo ion du ing he scan (F is on e al., 1996), hoping o good quali y da a is o en equi alen o hoping o da a wi h low head mo ion. The impo ance o educing head mo ion du ing da a acquisi ion has been imp essi ely documen ed in se e al s udies showing ha head mo ion can lead o misleading esul s (in some cases, e en a e e ospec i e mo ion co ec ion; Powe e al., 2012; Sa - e hwai e e al., 2012; an Dijk e al., 2012). In pedia ic neu oimaging s udies, he gene al p oblem o head mo ion is especially challenging. Despi e child en’s high mo i a ion o lie s ill, in mos s udies child en s ill mo e mo e han adul s. This poses a p oblem o da a quali y in pedia ic neu oimaging especially o g oup compa isons be ween child en and adul s. E en wi h e ospec i e head mo ion co ec ion (which is limi ed in i s abili y o co ec o mo ion Field e al., 2000; F ei e and Mangin, 2001; F is on e al., 1996), g oup di e ences in head mo ion o en pe sis . This usually leads o he exclusion o mo ion-a ec ed uns o pa icipan s om da a analysis (e.g. Meissne e al., 2019; No d e al., 2018; Walb in e al., 2020), hence equi ing addi ional ime and esea ch unds o achie e an adequa ely powe ed s udy design. To minimize mo ion du ing da a acquisi ion, a ious solu ions ha e been de eloped. Fo s uc u al MRI scans in clinical se ings, seda ion is o en used. Howe e , his is no an op ion o MRI s udies in esea ch se ings due o he need o a en i e pa icipan s as well as he unac- cep able isk o heal h- ela ed side e ec s (de Bie e al., 2010). One se o solu ions o MRI s udies aims a es aining head mo ion physically. Res ic i e app oaches including bi e ba s (Menon e al., 1997) o * Co esponding au ho a : TU Do mund Uni e si y, Facul y o Rehabili a ion Sciences, Depa men o Vision, Visual Impai men s & Blindness, Emil-Figge-S . 50, 44227, Do mund, Ge many E-mail add esses: [email p o ec ed] (T.W. Meissne ), [email p o ec ed] (J. Walb in), [email p o ec ed] (M. No d ), [email p o ec ed] (K. Koldewyn), [email p o ec ed] (S. Weigel ). Con en s lis s a ailable a ScienceDi ec De elopmen al Cogni i e Neu oscience jou nal homepage: www.else ie .com/loca e/dcn h ps://doi.o g/10.1016/j.dcn.2020.100803 Recei ed 23 Oc obe 2019; Recei ed in e ised o m 20 Ap il 2020; Accep ed 27 May 2020 De elopmen al Cogni i e Neu oscience 44 (2020) 100803 2 he moplas ic ace masks (G een e al., 1994), a e conside ed e ec i e o sho scan du a ions. Howe e , hey educe com o and a e he e- o e seldomly accep ed in pedia ic neu oimaging. In con as o he desi ed ou come, he discom o o hese me hods can also lead o mo e idge ing and wiggling in sea ch o com o (Ze i o, 1996), especially o longe scans. App oaches ha use indi idually 3D-p in ed s y o oam head molds a e e ec i e and said o be com o able (Powe e al., 2019), bu equi e addi ional ime and esea ch unds o each pa icipan . A less esou ce-in ensi e, ye e ec i e and accep ed solu ion, is o p o ide ac ile eedback abou pa icipan ’s head mo ion by applying a ape o ibbon ac oss he head coil ha ouches he pa icipan ’s o ehead (K ause e al., 2019). Aside om physical cons ain s, pedia ic neu oimaging g oups o en adap he s udy p ocedu e o educe child en’s mo ion du ing da a acquisi ion. The mos p ominen ool is o p ecede he ac ual MRI ses- sion(s) wi h a scanne aining session (Raschle e al., 2009; Sli e e al., 1993). Scanne aining sessions a e usually pe o med in a mock scanne , a cus om-buil o pu chased acsimile o a eal MRI, which lacks he echnical capabili y o acqui e ac ual da a, bu can play scanne sounds and has a simila se up o he eal scanne (e.g. head coil-moun ed mi o , esponse bu ons, e c.). Du ing hese sessions, beha io al aining is used o each lying s ill in he scanne bo e and esponding o he ask. The e is a gene al consensus ha a scanne aining session is bene icial o MRI da a acquisi ion as i educes child en’s anxie y (Du s on e al., 2009; Raschle e al., 2009; Rosenbe g e al., 1997). While a educing e ec on head mo ion is also assumed, he cu en li e a u e has no been able o show his (due o a lack o con ol g oups ha did no ecei e scanne aining; bu see Ba nea-Go aly e al., 2014; de Bie e al., 2010; Eps ein e al., 2007 o success a es and educed mo ion a e scanne aining). Simila ly, many pedia ic neu oimaging g oups ha e es ablished expe ience-based guidelines o child- iendly s udy designs in e ms o he scanning p ocedu e. These usually include keeping uns sho , in e spe sing ana omical scans as ask- ee inside-scanne b eaks while p esen ing en e aining ideo clips, and limi ing he o al scanning session ime. Mo eo e , based on indings in adul s (Huijbe s e al., 2017), asks a e designed in an engaging and in e ac i e way a he han jus equi ing passi e pe cep ion o s imuli. In consequence, a small body o li e a u e has de eloped ha ecommends scanning p ocedu es o pedia ic neu oimaging (e.g. G eene e al., 2016; Habibi e al., 2015; Raschle e al., 2009). Howe e , as o scanne aining, he con ibu ion o p esumably child- iendly adap a ions in scanning p ocedu es on minimizing head mo ion has no been in es iga ed so a . The p esen s udy se ou o iden i y which MRI s udy p ocedu es con ibu e o educed head mo ion in MRI s udies. Ou aim was o gene a e da a-d i en sugges ions o op imize s udy p ocedu es o child en and adul s sepa a ely. To his end we u ilized head mo ion es ima es de i ed om a s anda d mo ion co ec ion pipeline applied o 77 child en and 64 adul s om h ee MRI s udies a wo si es, and no es ha yielded demog aphic da a, he da e o scanne aining, and he sequence o da a acquisi ion. Using sepa a e mul ile el linea models o child en and adul s, we in es iga ed he e ec o spli ing up da a acquisi ion in o se e al days o sessions, he e ec o in e spe sing unc ional da a acquisi ion wi h s uc u al uns and ideo clip b eaks, he e ec o ime be ween scanne aining and he ac ual scan, and he e ec o ask engagemen . 2. Me hods 2.1. De ini ion o session, un, and unc ional segmen In his a icle, we make impo an dis inc ions be ween “session”, “ un”, and “ unc ional segmen ”. Session co esponds o a con inuous pe iod o ime spen inside he scanne . A session begins upon en e ing he scanne and ends as a pa icipan lea es he scanne . Fo example, i a pa icipan en e s he scanne , akes a b eak ou side he scanne and e-en e s he scanne , his would cons i u e wo sessions. Run co e- sponds o a con inuous image acquisi ion sequence. Fo example, a pa icipan could comple e an expe imen wi h ou MRI uns wi hin a single session. Func ional segmen co esponds o a pe iod o consecu- i e acquisi ion o unc ional uns inside he scanne . Fo example, i a scan p ocedu e in ol es an ana omical T1-scan, h ee MRI uns, wo di usion-weighed imaging (DWI) sequences, and inally ou mo e MRI uns, his pa icipan has comple ed wo unc ional segmen s. 2.2. Pa icipan s Ou s udy included da a om wo neu oimaging cen e s (see 2.3, Neu oimaging) and h ee de elopmen al cogni i e neu oimaging s udies, a o al 680 uns om 77 child en and 624 uns om 64 adul s. Some o he da a has been used o answe ques ions conce ning he neu ocogni i e isual and social de elopmen in child en and adul s p e iously (Meissne e al., 2019; No d e al., 2018; Walb in e al., 2020). The inal analyzed da a se was educed o 626 uns om 77 child en and 470 uns om 54 adul s due o se e al exclusion c i e ia (see 2.5, Da a exclusion). In child en, he numbe o uns pe pa icipan anged om 1 o 14 (M ¼8.13, SD ¼3.21), in adul s, he numbe o uns pe pa icipan anged om 4 o 14 (M ¼8.70, SD ¼3.33). Child en’s age anged om 6.78 o 13.01 yea s (M ¼9.52, SD ¼1.68), adul ’s age anged om 18.41 o 35.02 yea s (M ¼21.76, SD ¼2.67). All pa ici- pan s we e heal hy, had no mal o co ec ed- o-no mal ision, and had been bo n a e m. No pa icipan epo ed pas o cu en neu ological o psychia ic condi ions, o had s uc u al b ain abno mali ies. All pa icipan s as well as child en’s pa en s ga e in o med and w i en consen o pa icipa e olun a ily. 2.3. Neu oimaging MRI ook place a he Neu oimaging Cen e o he Resea ch Depa men o Neu oscience a Ruh Uni e si y Bochum and he Bango Imaging Cen e a Bango Uni e si y. A Bochum, a o al o 338 uns in 50 child en and 346 uns in 40 adul s we e acqui ed—a Bango , a o al o 188 uns in 27 child en and 123 uns in 14 adul s we e acqui ed. A bo h si es, 3.0 T Achie a scanne s (Philips, Ams e dam, The Ne he lands) and 32-channel head coils we e used (Supplemen a y Fig. S1). Fo all MRI, we used blood oxygen le el dependen (BOLD) sensi i e T2*-weigh ed sequences. Ac oss si es and expe imen s, he ollowing MRI acquisi ion pa ame e s we e cons an : FOV ¼240 mm �240 mm, ma ix size ¼80 �80, oxel size ¼3 mm �3 mm �3 mm, TR ¼2000 ms, TE ¼30 ms. Howe e , slice gap (Bango : 0 mm, Bochum: 0.4 mm), numbe o slices (Bango : 32, Bochum: 33), slice scan o de (Bango : ascending, Bochum: ascending in e lea ed), and lip angle (Bango : 83�, Bochum: 90�) di e ed be- ween si es. Visual s imuli we e p esen ed ia a VisuaS im Digi al goggle sys em (FOV: 30��24�, 800 �600 pixel, Resonance Technology Inc., CA, USA) in Bochum and ia a MR-sa e moni o and mi o sys em (FOV: 25.92��16.90�, 1920 �1200 pixel; Camb idge Resea ch Sys ems, UK) in Bango . Due o he isual isola ion inhe en wi h he goggle sys em, a esea che was p esen nex o he scanne bo e opening a all imes o immedia e e bal con ac and mo ion eedback in Bochum, bu no in Bango . Fo each s udy, a ce ain numbe o unc ional uns and s uc u al scans we e planned. De ails on he expe imen s ha we e done du ing he unc ional uns as well as he numbe o pa icipan s and uns o each o he asks a e gi en in he Supplemen a y Tex S1 and Supple- men a y Table S1. In adul s, expe imen e s ollowed his p o ocol, checking in wi h pa icipan s a e he comple ion o each expe imen (i. e. a e mul iple uns) o explain wha would happen nex and only s opped o di e ed om p o ocol i pa icipan s ac i ely epo ed eeling unwell. In con as , in child en, we checked in wi h pa icipan s a e each un o ac i ely inqui e abou hei well-being. Mo eo e , nea ing he end o a session, we also ac i ely asked i hey s ill el good T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 3 and eady o do ano he un. This was done in o de o gi e child en he oppo uni y o exp ess any signs o discom o , which o ou expe ience child en do no necessa ily u e spon aneously bu some imes only a e encou aging hem o be open abou i o gi ing hem an explici op- po uni y. Thus, we dynamically decided when o b eak up acquisi ion in o sessions o days, when o in e spe se asks wi h a s uc u al scan accompanied by an en e aining ideo, o when o end he s udy. 2.4. Scanne aining All child en unde wen a scanne aining in o de o amilia ize hem wi h he scanne en i onmen and achie e high-quali y scans wi h as li le mo ion as possible (Supplemen a y Fig. S1). A bo h scanne si es, p e- eco ded audio o MRI acquisi ion sequences was played du ing scanne aining o simula e he eal scanne en i onmen and isual s imula ion was achie ed h ough a mi o sys em a ge ed a a moni o ou side o he mock scanne bo e. In o he aspec s, aining sessions di e ed be ween si es. In Bango , scanne aining was conduc ed igh be o e scanning and en ailed lying inside a mock scanne wi h a mo ion sensi i e elec ode placed on he o ehead o measu e mo emen ac oss h ee ansla ion and h ee o a ion axes (MoT ak Head Mo ion T acking Sys em; Psychology So - wa e Tools, 2017). Child en ecei ed isual mo ion eedback ia an on- sc een cu so ha was con olled by child en’s head mo emen s. Chil- d en we e ins uc ed o lie s ill and keep he cu so in he middle o a a ge ci cle, whose diame e allowed o 3 mm head ansla ions in any di ec ion. Once child en we e able o keep he cu so wi hin his a ge egion o a imed pe iod o 30 s, child en wa ched a sho anima ed ideo. When mo emen exceeded 3 mm ansla ion, ideo playback was paused, p o iding immedia e eedback ha hey had mo ed oo much. Once child en we e able o wa ch he ideo o a pe iod o 2 min wi hou a ideo pause, scanne aining was comple ed. In Bochum, scanne aining was conduc ed be ween one o en days be o e he MRI s udy o he majo i y o pa icipan s (n ¼48). A mi- no i y was ained on he same day as he i s scan (n ¼1) o be ween 12 and 35 days be o e he i s scan (n ¼7). T aining began wi h si ing on he ex ended mock scanne bed, wa ching a sho anima ed ideo ha in oduced a co e -s o y explaining he asks o be pe o med inside he scanne , and p ac icing hese asks (bu on-p ess). This was ollowed by explaining and demons a ing he p ocedu e o en e ing an MRI scanne wi h a la ge puppe . Nex , child en en e ed he mock scanne bo e, wa ched a sho anima ed ideo and pe o med se e al p ac ice ials o he asks p esen ed on a sc een ha was isible ia a mi o sys em. The esea che s ga e eedback wi h espec o he le el o mo emen and ask pe o mance h oughou he scanne aining ses- sion. I was es ablished ha a esea che would gen ly ouch he chil- d en’s shin in case o excessi e mo ion as a means o mo ion eedback. Ve bal eedback was g adually dec eased h oughou he session. As head mo ion du ing he mock scanne session in Bochum was no quan i ied, eedback was based on he expe imen e ’s obse a ion o child en’s head mo ion. Once he esea che decided ha mo ion le els we e accep able and he asks we e unde s ood, scanne aining was comple ed. 2.5. Da a exclusion Ou ini ial da ase encompassed all comple ed MRI uns o wo de elopmen al cogni i e neu oscience s udies conduc ed in Bochum and one in Bango . Abo ed uns (e.g. due o echnical e o s) we e no included, as hey we e no e ie ed om he MRI-con olling compu e and no eliably eco ded in handw i en no es. To ensu e ha ou e- sul s a e alid and in e p e able, we excluded da a ha would ha e biased ou analyses (Fig. 1). In a i s s ep, we excluded en pa icipan s (110 uns) om Bango ha we e eco ded wi h di e en acquisi ion pa ame e s, i.e. wi h a slice hickness o 3.5 mm ins ead o 3.0 mm. This di e ence in oxel size along he z-axis would a ec h ee ou o he six mo ion es ima ion pa ame e s ( o a ion a ound he y-axis, o a ion a ound he x-axis, and ansla ion along he z-axis) du ing mo ion co ec ion, leading o sys- ema ically la ge mo ion es ima es. In a second s ep, we excluded 56 uns ha did no e lec i s - ime MRI expe ience. Fou pa icipan s had pa icipa ed in bo h Bochum MRI s udies ha a e included in ou analysis. Pa icipa ion in bo h s udies was sepa a ed by 2–18 mon hs. To con ol o possible con ounding e ec s o p io scanning expe ience on mo ion, we excluded all uns o hei second MRI s udy pa icipa ion. Implemen ing his con ol s ep ensu ed ha we epo on he a he ypical case—a leas o child en—o a i s - ime MRI s udy pa icipan . No e ha we could no ule ou he possibili y ha pa icipan s ook pa in an MRI s udy o a di e en lab be o e, al hough his is highly unlikely o child en and mos adul s a ou neu oimaging cen e s. In a las s ep, we excluded 42 es ing-s a e uns. Res ing-s a e uns we e only acqui ed in one o he h ee s udies, only once pe pa icipan , and in he majo i y o cases he un was acqui ed a he end o a session. Mo eo e , es ing- s a e uns we e he only uns in which pa icipan s had hei eyes closed du ing acquisi ion. This unique se o ci cums ances makes es ing-s a e uns e y likely o bias ou esul s, as he e is no enough da a o e i- cien ly model his se o ci cums ances, e.g. as an independen a iable. 2.6. Head mo ion es ima ion To es ima e head mo ion du ing MRI scans, we used he neu o- imaging so wa e package B ainVoyage (Ve sion 20.2 o 64bi Win- dows, RRID: SCR_013057). Fi s , MRI un se ies in DICOM o ma we e con e ed o he p op ie a y STC and FMR o ma using B ainVoyage sc ip s. Then, we applied B ainVoyage MRI da a p ep ocessing ools in hei de aul se ings. Tha is, o slice scan ime co ec ion, we used cubic spline in e pola ion. Fo 3D mo ion co ec ion, we used a ilinea de ec ion and sinc in e pola ion app oach wi h a maximum o 100 i - e a ions o i a olume o he e e ence olume. Resul ing mo ion log iles con ained six imese ies ep esen ing he es ima ed olume-wise ins an aneous ansla ion and o a ion o axes x, y, and z in e e ence o he i s olume o he espec i e expe imen . Fo each o he six mo ion pa ame e s, we calcula ed a imese ies o olume- o- olume mo ion, i.e. he di e ence be ween each olume’s mo ion pa ame e and he p e ious olume’s mo ion pa ame e . Ro a- ional mo ion was con e ed om deg ee o millime e by calcula ing displacemen on he su ace o a sphe e o 50 mm adius (app oxima ing he dis ance om he cen e o he head o he co ex; Powe e al., 2012). The amewise displacemen o he head posi ion (FD) was Fig. 1. Da a exclusion s eps and esul ing sample sizes o child en (le , da k g ay) and adul s ( igh , ligh g ay). Main box wid hs a e in ela ion o he o iginal o al numbe o uns. Box segmen a ion mi o s he ela i e dis ibu- ion o adul s. child pa icipan s o uns, espec i ely. T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 4 calcula ed as he sum o he absolu e olume- o- olume mo ion alues (Powe e al., 2012). FD was shown o ha e a s ong associa ion wi h mo ion-induced a i ac s (Ci ic e al., 2017). Fo each un, we calcula ed he mean FD. Mo eo e , o disc e e one-minu e-in e als (i.e. 30 olumes) wi hin uns, we calcula ed he pe cen age o high-mo ion olumes wi h FD abo e he h eshold o 0.3 mm ( olume- o- olume). This h eshold o high-mo ion olumes was de e mined by ex ensi e explo a ion o ou da a be o e any s a is ical analysis, and aimed a cap u ing mo ion “spikes” wi hou impac ing he mo ion “ loo ” olumes as done p e i- ously (Powe e al., 2019, 2012) and esul ed in plausible a ios o high-mo ion olumes o child en (M ¼12.80%, SD ¼20.89%) and adul s (M ¼4.48%, SD ¼12.10%). Fo a isualiza ion o he high-mo ion olume h eshold, see Fig. 2. No e ha he mo ion “ loo ” is o en below 0.1 mm FD. Thus, ou h eshold o 0.3 mm allowed o some mo ion ha is no ca ego ized as a high mo ion “spike” (e.g. Fig. 2, bo om igh panel) and migh ha e been e en s ic e . The wo head mo ion measu es—mean FD pe un and equency o high-mo ion olumes pe minu e o a un—we e in es iga ed in sepa a e analyses. 2.7. P edic o a iables Ou s udy did no aim o examine age g oup di e ences be ween child en and adul s, bu o p o ide guidelines o p ac i ione s and e- sea che s who conduc pedia ic neu oimaging examina ions and/o expe imen s ha migh also include adul con ol g oups. Thus, we epo sepa a e analyses o child en and adul s. This app oach enabled us o include wo mo e p edic o a iables (PVs) o child en, as adul s did no pe o m scanne aining and we did no expec age e ec s in ou ange o 1835-yea -old adul s. Fo each age g oup, we assessed wo head mo ion measu es wi h up o eigh possible p edic o s. 2.7.1. Mean FD In ou i s analysis, we asked which ac o s would in luence he mean mo ion du ing a un. Se en PVs we e e alua ed: �PVs 1–4) P io scan ime encoded he ime in minu es ha a pa ici- pan had been in he scanne al eady. Tha is, he summed scan ime o all unc ional as well as s uc u al uns ha had been adminis e ed 1) since he beginning o he unc ional segmen , 2) since he begin- ning o he session, 3) since he beginning o he day, o 4) since he beginning o he s udy. Fo de ini ions o session and unc ional segmen , please e e o Sec ion 2.1, De ini ion o session, un, and unc ional segmen . No e ha p io scan ime since he beginning o he day was only analyzed in child en, because while child en we e scanned on ei he one (n ¼45), wo (n ¼31) o h ee days (n ¼1), only 2 ou o 54 adul s we e scanned on mul iple (he e: wo) days. The e o e, ou adul da a did no ha e he equi ed dis ibu ion o allow any in e ences on b eaking up MRI da a acquisi ion be ween days in adul s. Fo example, imagine he complex case o a pa icipan ha comes in o a second day o MRI scanning (a e 60 min o scanning on he i s day), akes a ba h oom b eak a e he comple ion o he day’s i s MRI expe imen (a e 15 min), e-en e s he scanne , com- ple es he second MRI expe imen (15 min), and is now scheduled o an inside-scanne b eak, du ing which a di usion-weighed im- aging sequence is being acqui ed while a ideo clip is p esen ed (5 min.). The nex MRI un ha is acqui ed would ha e di e en alues o each o he p io scan ime a iables, e.g. a) p io unc ional segmen scan ime ¼0 min, b) p io session scan ime ¼20 min, c) p io day scan ime ¼35 min, and d) p io s udy scan ime ¼95 min. In gene al, we hypo hesized ha pa icipan s’ mo ion would in- c ease wi h ime spen in he scanne , bu could be “ ese ” o lowe mo ion by b eaks. Thus, we es ed he e ec o ou di e en p io scan ime a iables o de e mine which—i any—way o b eaking up da a acquisi ion educes subsequen mo ion. Fig. 3 isualizes he possible main e ec s o p io scan ime a iables. To de e mine he p io scan ime a iables alues, we e alua ed s uc u ed handw i en MRI session no es ha s a ed he sequence o acqui ed unc ional and s uc u al uns and b eaks be ween sessions and days o each pa icipan . Then, he du a ion o each un was e ie ed om i s DICOM ile heade . This DICOM heade du a ion always exceeded he p oduc o epe i ion ime and numbe o olumes, i.e. i included scanne - and sequence-speci ic p epa a ion ime. We did no include po en ial b eaks be ween uns ha occu due o no mal ope a ion ime ha is equi ed o s a a new pa adigm on he s imula ion compu e o o s a he scanne , how-do-you- eel-inqui ies, o mino echnical issues, because we did no log hese e en s. Fu he , we did no include abo ed uns, because we did no sa e his da a no ook Fig. 2. F amewise displacemen (FD) o he head in mm in h ee example child en and adul s (g ay w i ing ¼pa icipan IDs). Fo each pa icipan s, one ep esen a i e un was chosen o show he high-mo ion olume h eshold ( ed ho izon al dashed line). G ay e ical lines indica e he one-minu e-in e als, o which he pe cen age o high-mo ion- olumes was calcula ed. (Fo in e - p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o his a icle.) Fig. 3. Visualiza ion o he e ec o p io scan ime wi h hypo he ical da a. Each ow shows he e ec o a di e en kind o p io scan ime. Each column shows a single session, consis ing o ou MRI uns. The do ed e ical line and poin s on he e ical line deno e inside scanne b eaks du ing which s uc u al scans we e acqui ed and ideo clips we e p esen ed. T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 5 eliable no es, ende ing a de e mina ion o he du a ion o hese scans impossible. �PV 5) Age was de ined in yea s a he i s day o scanning and was only in es iga ed o child en. This a iable was included o explo e possible in e ac ions wi h he o he PVs, no o main e ec s o age, as main e ec s o age a e well es ablished in he ield and would only lead o an unnecessa ily complex model wi h less powe o de ec e ec s o in e es . �PV 6) Scanne aining da e was eco ded as he numbe o days ha passed since he scanne aining o child en only. We hypo hesized ha a g ea e ime in e al be ween scanne aining and ac ual MRI scan would esul in highe mo ion. �PV 7) Task engagemen encoded how much ac i e engagemen a gi en MRI un equi ed. We dis inguished i pa icipan s jus had o passi ely wa ch he display, o i pa icipan s had o pe o m a ask ha included bu on-p essing. We expec ed ha an ac i e ask engagemen would esul in educed mo ion due o enhanced a en ion and less awa eness o a possibly uncom o able si ua ion, i ching, o o he dis ac ions. In addi ion o hese main e ec s, we in es iga ed he possible in- e ac ions be ween he p io scan ime a iables and he o he a iables. O he wo-way o h ee-way in e ac ions we e no in es iga ed o a oid o e ly complex and highly compu a ion-in ensi e models and as he de ec ion o high-o de e ec s would equi e a mo e powe ul s udy design (Simonsohn, 2014). 2.7.2. High-mo ion olumes In ou second analysis we asked which a iables would in luence head mo ion o e he cou se o a un. As decisions on e aining o dis- ca ding uns o u he analysis a e o en in o med by high-mo ion olumes wi hin a un a he han a un’s mean mo ion, ins ead o mean FD, we in es iga ed he pe cen age o high-mo ion olumes o each minu e o a un (i.e. FD olume >0.3 mm; see 2.6, Head mo ion es ima ion). Consequen ly, eigh PVs we e e alua ed: �PVs 1–7) All p e ious PVs, i.e. p io unc ional segmen scan ime, p io session scan ime, p io day scan ime, p io s udy scan ime, age, scanne aining da e, ask engagemen �PV 8) Minu e o un coded disc e e one-minu e-in e als wi hin uns, allowing us o in es iga e he cou se o high- olume mo ion occu - ence ac oss he cou se o a un. In addi ion o main e ec s, we in es iga ed he possible in e ac ions be ween minu e o un and he o he a iables. Again, we did no in es iga e o he wo- o h ee-way in e ac ions (see 2.7.1, Mean FD) 2.8. S a is ical model 2.8.1. Hie a chical da a s uc u e We analyzed ou da a using mul ile el linea models (MLMs) due o he hie a chical ou - o i e-le el g ouping s uc u e inhe en in ou da a (c . Engelha d e al., 2017; Fig. 4): Mo ion es ima ions o mean FD pe un a e nes ed wi hin sessions (le el 4), sessions a e nes ed wi hin days (le el 3), days a e nes ed wi hin pa icipan s (le el 2), and pa - icipan s a e nes ed wi hin s udies (le el 1). Mo ion es ima ions o pe cen age o high-mo ion olumes pe minu e o a un a e nes ed wi hin uns, es ablishing a i h le el. This hie a chical da a s uc u e is likely o in oduce dependency o obse a ions wi hin a g ouping a i- able. Fo example, mo ion om wo andomly selec ed uns om he same pa icipan is likely o be much mo e simila han mo ion om di e en pa icipan s. In echnical e ms, he esiduals wi hin a g ouping a iable a e likely o be co ela ed. Thus, he assump ion o independen esiduals, c ucial o pa ame ic es s, is iola ed. While dependen ob- se a ions o one le el, e.g. wi hin pa icipan s, could be handled easily by a con en ional epea ed-measu es ANOVA, his case is a complex nes ed mul ile el dependency s uc u e ha canno be adequa ely and alidly handled by a epea ed-measu es ANOVA—bu can be by an MLM. 2.8.2. MLM c ea ion s ep 1: iden i ying ele an g ouping s uc u es MLMs we e c ea ed in a da a-d i en p ocess. Fi s , we assessed he possibili y ha he g ouping a iables would in oduce dependencies in he da a—and hus con i m he need o an MLM. To his end, we calcula ed he in aclass co ela ion (ICC, i.e. he p opo ion o he o al a iance ha is explained by he espec i e g ouping ac o ) o each g ouping le el o he model and libe ally inco po a ed all g ouping le els in o ou model ha would explain a leas 1% o he o al a iance (i.e. ICC �0.01, Supplemen a y Tex S2). Nex , we used he chi-squa e likelihood a io es and Akaike’s in o ma ion c i e ion (AIC) o es i he model ha inco po a ed he g ouping s uc u e ac ually had a be e i o ou da a han he model wi hou he g ouping s uc u e and included o igno ed he g ouping s uc u e in all subsequen models acco dingly (Supplemen a y Tex S3). ICC analysis o mean FD o e he cou se o a s udy indica ed ha in bo h child en and adul s, he g ouping ac o s pa icipan (child en: 0.487, adul s: 0.376) and session (child en: 0.229, adul s: 0.021) explained a subs an ial amoun o a iance in he da a, while he g ouping ac o s s udy and day did no (bo h age g oups <0.001). In bo h child en and adul s, models ha allowed andom in e cep s o pa ici- pan and session i he da a signi ican ly be e han a model wi h ixed in e cep s (child en: Supplemen a y Table S4, adul s: Supplemen a y Table S5). Thus, he MLMs we e buil wi h andom in e cep s o pa icipan and session. ICCs o he pe cen age o high-mo ion olumes ac oss he cou se o a un in bo h child en and adul s e ealed subs an ial dependency wi hin he g ouping le els pa icipan (child en: 0.136, adul s: 0.292), session (child en: 0.275, adul s: 0.037), and un (child en: 0.143, adul s: 0.309), bu no wi hin he g ouping le els s udy (bo h age g oups: <0.001) and day (child en: <0.001, adul s: 0.004). A andom in e cep model o pa icipan , session, and un i ou da a be e ha a ixed in e cep model (child en: Supplemen a y Table S6, adul s: Supple- men a y Table S7). Consequen ially, he MLMs we e buil wi h andom in e cep s o pa icipan , session and un. 2.8.3. MLM c ea ion s ep 2: s epwise inclusion o ele an p edic o s Second, we in oduced he PVs, i.e. main e ec s and selec ed in- e ac ions ( ixed e ec s), in o he model s epwise, o es i hey imp o ed he model i signi ican ly. The o de o in oduc ion o mean FD and high-mo ion olumes MLMs is lis ed in Supplemen a y Tables S2 and S3, espec i ely. I a model including a ixed p edic o led o a be e model i han he p e ious model wi hou i , he p edic o was included Fig. 4. Hie a chical da a s uc u e wi h ou le els. Fo isualiza ion pu poses, only one ic ional subjec is displayed o each s udy. T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 6 in all subsequen models, o he wise i was le ou o subsequen models. Fo child en’s mean FD MLM main e ec s o p io unc ional segmen ime, p io session ime, p io day ime, and scanne aining da e, as well as he in e ac ion e ec s o age �p io session ime and age �p io s udy ime we e included in he model (Supplemen a y Table S4). Fo child en’s high-mo ion olumes MLM, main e ec s o minu e o un, p io segmen ime, p io session ime, p io day ime, and scanne aining da e we e added o he model (Supplemen a y Table S6). Fo adul ’s mean FD MLM, main e ec s o p io unc ional segmen ime, p io s udy ime, and a p io unc ional segmen ime � ask engagemen in e ac ion we e added o he model (Supplemen a y Table S5). Fo adul ’s high-mo ion olumes MLM main e ec s o minu e o un, p io segmen ime, and p io session ime, we e added o he model (Supplemen a y Table S7). 2.8.4. MLM c ea ion s ep 3: assessing he need o pa icipan -speci ic p edic o e ec s Thi d, we es ed i he model i imp o ed u he i we le he e ec o each included main e ec p edic o a y o e pa icipan s ( andom slopes) and added signi ican andom e ec s o he model acco dingly. P edic o s scanne aining da e and age we e no conside ed as hey a e cons an o each pa icipan ’s un. Fu he , we did no i andom slopes o in e ac ion e ms as his p e en ed he MLMs o con e ge—possibly because ou da a did no ha e enough powe o i a model o such high le el o complexi y. Adding andom slopes ac oss pa icipan s o all main e ec s u he imp o ed model i in all MLMs and hus was inco po a ed in he inal models (Supplemen a y Tables S4–S7). 2.9. So wa e Da a handling and s a is ical da a analysis was pe o med using R ( e sion 3.6.0, RRID: SCR_001905, R Co e Team, 2019) in RS udio ( e sion 1.2.1335; RRID: SCR_000432). Fo MLMs, we used he nlme package ( e sion 3.1–140, RRID:SCR_015655, Pinhei o e al., 2019). Figu es we e c ea ed using he ggplo 2 package ( e sion 3.2.0, RRID: SCR_014601, Wickham, 2016). Reg ession lines and con idence in e al bands o ixed e ec s isualiza ions, i.e. p edic o e ec plo s, we e c ea ed using he e ec s package ( e sion 4.1-1, Fox and Weisbe g, 2018). Ma ginal R 2 ( o ixed e ec s) and condi ional R 2 ( o ixed e - ec s and andom e ec s combined) goodness-o - i was es ima ed using he .squa edGLMM unc ion o he MuMIn package ( e sion 1.43.6, Ba on, 2019), which is based on me hods by Nakagawa and Schielze h (2013), Johnson (2014), and Nakagawa e al. (2017). 3. Resul s We e alua ed which a iables would p edic head mo ion du ing MRI in child en and adul s. We assessed wo measu es o head mo ion in sepa a e analyses: Fi s , we in es iga ed he cou se o head mo ion ac oss an MRI s udy in e ms o mean amewise displacemen pe un. Then, we in es iga ed he cou se o head mo ion o e he cou se o a un in e ms o he equency o high-mo ion olumes pe minu e. Each analysis was pe o med sepa a ely o child en and adul s. Reg ession coe icien s o all main and in e ac ion e ec s included in he MLMs we e es ed agains ze o using one-sample - es s wi h a signi icance h eshold o α ¼0.05. 3.1. Spli ing up da a acquisi ion educes mo ion 3.1.1. Child en E alua ion o ixed e ec s p edic o s in he inal mean FD MLM e ealed ha p io session scan ime and a p io s udy scan ime �age in e ac ion signi ican ly p edic ed FD in child en (Fig. 5, Table 1). Child en showed g ea e mo ion wi h ongoing session leng h (Fig. 5, le ). No e ha he o al numbe o sessions in child en anged om 1 o 4 (16 child en did only one session, 55 child en did wo sessions, ou child en did h ee sessions, and wo child en did ou sessions). In addi ion, mo ion inc eased wi h s udy leng h in olde child en, bu no as much in younge child en (Fig. 5, igh ). As ou da a in child en shows one ob ious ou lie , a pa icipan whose mean FD is a ound 2 mm o all o he /his i e uns, we es ed i his case would in luence he obse ed e ec s. Howe e , a eanalysis wi hou his pa icipan only sligh ly inc eased he e ec o p io session scan ime and did no change he age �p io s udy scan ime in e ac ion no ably. Thus, he e, we epo he ull da a se . 3.1.2. Adul s Fo adul s, e alua ion o ixed e ec s p edic o s in he inal model e ealed ha p io unc ional segmen scan ime and p io s udy scan ime signi ican ly p edic ed mo ion (Fig. 6, Table 2). Adul s’ mo ion inc eased wi h ongoing s udy scan ime (no e ha in 52/54 adul s s udy scan ime is equal o day scan ime). Mo eo e , he longe unc ional scans a e acqui ed wi hou an inside-scanne b eak (e.g. h ough a s uc u al scan wi h elaxing ideo), he mo e adul s mo ed. O iginally, we also ound ha a ask engagemen �p io unc ional segmen scan ime in e - ac ion signi ican ly p edic ed mo ion, i.e. o passi e asks, adul s showed a s eepe inc ease in mo ion han o ac i e asks (Fig. 6, middle, hick black dashed line). Howe e , as i seemed possible ha his ask engagemen �p io unc ional segmen scan ime in e ac ion was d i en by an ex eme inc ease in mo ion a he end o a unc ional segmen in Fig. 5. P edic o e ec plo s o chil- d en’s head mo ion ac oss he cou se o a s udy.Thick magen a lines a e i ed linea eg ession lines o he ixed e - ec s p edic o s deno ed on he x-axes. Shaded magen a a eas show he 95% con idence band o he eg ession lines. Each g ay ci cle ep esen s one un. Le : Head mo ion ac oss he cou se o a ses- sion. Each ci cle-connec ing line ep e- sen s a pa icipan ‘s session. Pa icipan s ha we e scanned in mul iple sessions a e ep esen ed by mul iple lines. Righ : Head mo ion ac oss he cou se o a s udy o 6-9-yea -old and 10-13-yea old child en. Age bins we e chosen a bi- a ily o isualize he age �p io s udy scan ime in e ac ion. Each ci cle- connec ing line ep esen s one pa icipan . T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 7 passi e asks in a single pa icipan only (Fig. 6, hin black dashed lines and squa es), we eanalyzed he MLM wi hou da a om his pa ici- pan . While all o he e ec s emained s able (Fig. 6, see o e lapping g een solid and black dashed lines), he ask engagemen �p io unc- ional segmen scan ime in e ac ion was ende ed insigni ican by he exclusion o he pa icipan (Fig. 6, see black dashed line ou side o g een con idence band, Supplemen a y Table S8). 3.1.3. Summa y Ou esul s demons a e ha head mo ion in child en and adul s can be educed by spli ing up da a acquisi ion. Howe e , depending on he age g oup, di e en s a egies seem o be e ec i e. Fo child en, ou da a sugges s a bene i o spli ing up da a acquisi ion in o mul iple, sho sessions on he same day and keeping he o e all s udy leng h as sho as possible. In con as , adul s bene i ed om in e spe sing expe imen al uns wi h inside-scanne b eaks. 3.2. High-mo ion e en occu ence inc eases wi h un leng h 3.2.1. Child en Minu e o un signi ican ly p edic ed mo ion in child en, i.e. chil- d en’s mo ion inc eased wi h inc easing un leng h (Fig. 7, Table 3). Fo example, in he hi d minu e o a un, he a e age isk o a high-mo ion olume was a 16.5%, while i inc eased o 18.3% in he six h minu e o a un. P io day scan ime was also a signi ican p edic o o mo ion. Howe e , each alue o his un-wise p edic o a ec s each minu e o a un equally and he e o e does no con ibu e o explaining how mo ion de elops o e he cou se o a un. Fo example, minu e o un 1, 2, and 3 always ha e he same alues o p io day ime. Thus, as he un-wise p io day scan ime did no in e ac wi h minu e o un, i s e ec is no o in- e es in his analysis. 3.2.2. Adul s Minu e o un signi ican ly p edic ed mo ion in adul s, i.e. adul ’s Table 1 Fixed e ec s pa ame e es ima es o inal model o child en’s mean mo ion ac oss he cou se o a s udy. pa ame e β Lowe 95% CI Uppe 95% CI SE d p In e cep 0.04895 0.01800 0.07990 0.01584 509 3.090 0.002 P io unc ional segmen scan ime 0.00147 0.00170 0.00464 0.00162 509 0.904 0.366 P io session scan ime 0.01564 0.00307 0.02821 0.00644 509 2.430 0.015 P io day scan ime 0.00212 0.00043 0.00467 0.00130 509 1.627 0.104 Scanne aining da e 0.00560 0.01790 0.02909 0.01203 509 0.465 0.642 Age �p io session scan ime 0.00121 0.00245 0.00003 0.00063 509 1.905 0.057 Age �p io s udy scan ime 0.00016 0.00005 0.00028 0.00006 509 2.721 0.007 No e: CI ¼con idence in e al, SE ¼s anda d e o , d ¼deg ees o eedom. Fig. 6. P edic o e ec plo s o adul ’s head mo ion ac oss he cou se o a s udy. Thick g een lines a e i ed linea eg ession lines o he ixed e ec s p edic o s deno ed on he x-axes. Shaded g een a eas show he 95% con- idence band o he eg ession lines. Each g ay ci cle ep esen s one un. Le : Head mo ion ac oss he cou se o a s udy. Each ci cle-connec ing line ep- esen s a pa icipan ‘s s udy. Righ : Head mo ion ac oss he cou se o a unc ional segmen o passi e and ac i e MRI asks. Each ci cle- connec ing line ep esen s one pa ici- pan ’s unc ional segmen da a acquisi- ion cou se. Pa icipan s ha we e scanned on mul iple days, on mul iple sessions, o wi h mul iple unc ional segmen s a e ep esen ed by mul iple lines. Black dashed lines and squa es show he pa icipan ha d o e he ask engagemen �p io unc ional segmen ime in e ac ion. (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o his a icle.) Table 2 Fixed e ec s pa ame e es ima es o inal model o adul ’s mean mo ion ac oss he cou se o a s udy. pa ame e β Lowe 95% CI Uppe 95% CI SE d p In e cep 0.05030 0.03328 0.06732 0.00869 377 5.787 <.001 P io unc ional segmen scan ime 0.00250 0.00098 0.00401 0.00077 377 3.230 0.001 P io s udy scan ime 0.00087 0.00030 0.00144 0.00029 377 2.990 0.003 Task engagemen �P io unc ional segmen scan ime 0.00185 0.00058 0.00428 0.00124 377 1.491 0.137 No e: CI ¼con idence in e al, SE ¼s anda d e o , d ¼deg ees o eedom. T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 8 mo ion inc eased wi h inc easing un leng h (Fig. 8, Table 4). Fo example, in he hi d minu e o a un, he a e age isk o a high-mo ion olume was a 5.3%, while i inc eased o 6.5% in he six h minu e o a un. P io unc ional segmen scan ime and p io session scan ime we e also signi ican p edic o s o mo ion. Howe e , as in child en, hese un- wise p edic o s did no con ibu e o explaining how mo ion de elops o e he cou se o a un as hey did no in e ac wi h minu e o un and hus a e no o in e es in his analysis. 3.2.3. Summa y Ou esul s indica e ha long uns ha e a nega i e impac on da a quali y: The equency o high-mo ion e en s inc eases wi h ongoing un leng h o bo h child en and adul s. 3.3. Con ol analysis: Age—bu no mo ion—in luenced he da a acquisi ion p ocedu e in child en Da a acquisi ion in e ms o which ask was acqui ed when, he po- si ions o s uc u al scans, i.e. inside-scanne b eaks, and he numbe o o al uns was la gely p ede e mined o each s udy and pa icipan . Howe e , we adap ed ou planning o he equi emen s o he gi en pa icipan —mainly o child en and only seldomly o adul s. Speci - ically, pa icipan s we e able o e mina e he s udy, day o session a any momen . Mos o he ime, e mina ion o a session was done a e ac i e inqui ing abou he well-being by he esea che s. He e, chil- d en’s age migh ha e in luenced esea che s’ sensi i i y owa ds well- being epo s and esea che s’ decision o e mina e. C ucially, subjec- i e well-being (o bo edom) migh be associa ed wi h mo ion in he scanne . Thus, i espec i e o epo ed well-being o he child, obse ed mo ion by he esea che s ( isible om he con ol oom in Bango and om s anding nex o he scanne bo e in Bochum) migh ha e in luenced he decision o e mina e he s udy, day, session, o unc- ional segmen ea ly. To in es iga e he possible e ec o age and mo ion on he p ocedu e, we an sepa a e mul iple eg ession analyses o he o al s udy leng h as well as he o al numbe o acqui ed uns, sessions, and days using pa icipan s’ age and hei mean pe cen age o high-mo ion- olumes as p edic o s. We chose he la e p edic o o e mean FD because only high-mo ion e en s would be isible o he esea che and ac as a possible indica o o change he p ede ined s udy p o ocol. The numbe o comple ed uns, sessions, o days was nei he signi ican ly p edic ed by age, no by high-mo ion olumes (all ps o age >.0664; all ps o high-mo ion olumes >.1676). Howe e , age—bu no he mean pe cen age o high-mo ion olumes—was a signi ican p edic o o o al s udy leng h (age: β (SE) ¼2.91 (1.29), ¼2.26, p ¼.027; high-mo ion olumes: β (SE) ¼  0.17 (0.16), ¼  1.10, p ¼.275). Thus, i is unlikely ha obse able mo ion du ing MRI scans was ac o ed in o he decision o de ia e om ou p ede ined s udy p o ocol, while young age migh ha e been a ac o . 4. Discussion We iden i ied ac o s ha p edic pa icipan ’s head mo ion in h ee neu ode elopmen al MRI s udies including da a o 77 child en and 64 adul s. Using MLMs, we in es iga ed he e ec o scanne aining da e, ask engagemen , as well as p io scan ime since he beginning o he s udy, day, session, o unc ional segmen . In child en, spli ing MRI da a acquisi ion in o mul iple sessions educed mo ion. Howe e , mo ion s ill inc eased ac oss he cou se o a s udy, especially in olde child en. In adul s, mo ion was educed a e ask- ee inside-scanne b eaks bu —as Fig. 7. Child en’s head mo ion ac oss he cou se o a un. Each g ay ci cle ep esen s one minu e bin. Each ci cle-connec ing line ep esen s a pa icipan ‘s un. Pa icipan s ha we e scanned o mo e han one unc ional un a e ep- esen ed by mul iple lines. Thick magen a line shows he i ed linea eg ession line o he minu e o un e ec . Shaded magen a a ea show he 95% con idence band o he eg ession line. Table 3 Fixed e ec s pa ame e es ima es o inal model o child en’s equency o mo ion peaks ac oss he cou se o a un. pa ame e β Lowe 95% CI Uppe 95% CI SE d p In e cep 1.12857 2.87326 5.13041 2.04301 2290 0.552 0.581 Minu e o un 0.60187 0.15491 1.04884 0.22818 2290 2.638 0.008 P io unc ional segmen scan ime 0.22521 0.01436 0.46478 0.12231 2290 1.841 0.066 P io session scan ime 0.08717 0.18705 0.36138 0.13999 2290 0.623 0.534 P io day scan ime 0.41294 0.22250 0.60339 0.09722 2290 4.247 <.001 Scanne aining da e 0.81829 0.71812 2.35470 0.78436 2290 1.043 0.297 No e: CI ¼con idence in e al, SE ¼s anda d e o , d ¼deg ees o eedom. Fig. 8. Adul ’s head mo ion ac oss he cou se o a un. Each g ay ci cle ep- esen s a one minu e bin. Each ci cle-connec ing line ep esen s a pa icipan ‘s un. Pa icipan s ha we e scanned o mo e han one unc ional un a e ep- esen ed by mul iple lines. Thick g een line shows he i ed linea eg ession line o he minu e o un e ec . Shaded g een a ea show he 95% con idence band o he eg ession line. (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o his a icle.) T.W. Meissne e al. De elopmen al Cogni i e Neu oscience 44 (2020) 100803 9 in child en—mo ion s ill inc eased ac oss he cou se o a s udy. In bo h child en and adul s, mo ion inc eased wi h un leng h. 4.1. Spli ing up da a acquisi ion educes mo ion 4.1.1. Child en Child en’s head mo ion seems o inc ease ac oss he cou se o a s udy, as p io s udy scan ime p edic ed head mo ion, especially o olde child en. A he same ime, we did no ind ha p io day scan ime p edic ed head mo ion, sugges ing ha child en do no seem o bene i om spli ing up s udies in o se e al days. Al e na i ely, a possible e - ec o spli ing up s udies in o se e al days was oo small o be de ec ed wi h he powe o ou s udy. To specula e on possible explana ions, o child en, he ini ial exci emen o pa icipa ing in an MRI s udy and he commi men o do e e y hing igh migh be less p onounced a e he i s day. On he second day, hey migh be less mo i a ed o lie s ill, e.g. due o he lacking no el y o he si ua ion (simila isual appea ance o ask demands). Olde child en showed a la ge inc ease in mo ion wi h inc easing s udy leng h. Ini ially, his seems unexpec ed. In ui ion would sugges ha olde child en a e be e a lying s ill o ex ended pe iods o ime. Howe e , i seems ha his age �p io s udy scan ime in e ac ion could be d i en by a highe baseline mo ion o young child en a he beginning o he s udy, which hen does no change as much o e ime. In con as , olde child en seem o be able o supp ess he u ge o mo e a he beginning o a s udy qui e well, and hen g adually elax in o mo ion le els simila o ha o younge child en. Thus, we cau ion agains o e in e p e ing his in e ac ion as a cause o plan sho e s udies jus because olde child en a e scanned, as p edic ed mo ion alues a maximum s udy leng h we e compa able and ew s udies will plan s udies wi h mo e han 80 min o pu e scan ime. Ou esul s indica e ha pedia ic neu oimaging s udies may bene i om b eaking up da a acquisi ion in o se e al sessions on he same day, as child en’s head mo ion inc eases wi h ongoing session leng h. So a , no o he s udies ha e in es iga ed mo ion ac oss mul iple sessions on he same day. Ne e heless, single-session s udies in child en ha e ound ha mo ion inc eases wi h inc easing session leng h in child en (Ach- e be g and an de Meulen, 2019; Engelha d e al., 2017). Ou s udy poin s o an impo an dis inc ion ega ding he kind o b eaks ha e- sea che s can schedule in child en’s MRI scanning p o ocols. While mo ion can be e ec i ely educed by allowing child en o exi he scanne , inside-scanne b eaks in MRI da a acquisi ion du ing which child en canno exi he scanne , bu ins ead wa ch a ideo while a s uc u al scan is acqui ed, do no seem o educe mo ion a e he b eak. Possibly, i is impo an o child en o be able o mo e eely, ge ace- o- ace social in e ac ions, o jus elie e he desi e o oid hei bladde —none o which a e possible wi hou exi ing he scanne . Al e na i ely, showing a ideo migh ha e e ec s opposi e o hose desi ed: Because he ideo is highly engaging, i may lead o educed mo ion du ing he b eak (Can lon and Li, 2013; Vande wal e al., 2015), bu any subsequen no mal MRI ask migh be pe cei ed as bo ing due o a con as e ec , leading o highe mo ion in u n. In ou expe ience, implemen ing ou side-scanne b eaks (i.e. spli - ing da a acquisi ion in o mul iple sessions) is e y easible. In ou h ee s udies, b eaks we e designed o las 5 min ou side o he scanne and we did no expe ience any p oblems wi h upholding his ime limi . I a co e s o y o child en is used, he b eak can be inco po a ed in ha s o y (e.g. in a space jou ney heme he e migh be limi ed ime o a space walk, main enance, plane isi , e c.). Ou 5-minu e b eaks usually p olonged he o al s udy by 15 min due o ge ing pa icipan s ou o and back inside he scanne , and unning ano he su ey scan. In con as o longe b eaks, such as 20 min o 2 h, hese sho b eaks limi he addi ional scanne cos s and a oid bo h un easible s udy da e a - angemen s wi h pa icipan s and pa en s as well as complica ed scan- ne ese a ion a angemen s. Thus, o s udies wi h mo e han 30 min o o al aw scan ime, implemen ing a spli in o wo sessions seems o be a easible and e ec i e way o educing mo ion in child en. Ano he app oach is o schedule wo o mo e child en back o back, who hen ake u ns in being in he scanne and aking b eaks. I should be no ed hough ha his app oach migh equi e addi ional s a o supe ise child en du ing scanning and b eaks in pa allel. In ou s udy, we did no ind an e ec o ask engagemen on head mo ion in child en. This con adic s p e ious indings, which showed ha child en show less mo ion in engaging asks in con as o less engaging asks o es ing s a e (Engelha d e al., 2017). This dispa i y migh be due o di e en de ini ions and aims o he s udies. In ou analysis, we de ined asks as engaging i pa icipan s had o pe o m a ask ha included bu on-p essing, and we de ined asks as non-engaging i pa icipan s jus had o passi ely wa ch he display. We chose his de ini ion as ou aim was o iden i y aspec s in MRI s udy designs ha can be ac i ely manipula ed by he esea che . Engelha d e al. (2017) de ined engaging asks as as -paced, cogni i ely demanding, o socially engaging. These aspec s a e o en inhe en in a ask and canno be changed wi hou changing he esea ch ques ion. Fo example, expe imen s on social de elopmen in child en will need o ha e a socially engaging componen . Howe e , while esea che s can decide i a gi en ask will equi e a manual esponse o no , his choice alone does no seem o a ec head mo ion. Possibly, mo ion- educing e ec s due o engaging asks ha equi e bu on-p essing o high a en ion a e also cancelled ou by mo ion-inc easing e ec s due o exci emen o bu on-p ess- ela ed mo ion. The ime be ween scanne aining and he ac ual scan does no seem o be a majo in luence on child en’s head mo ion du ing MRI scans. Al hough scanne aining iming imp o ed (and hus was added o) he MLM, i did no eme ge as a signi ican p edic o in he inal model. This assump ion i s wi h he obse ed in aclass co ela ion o he g ouping ac o s udy. As all child en in s udy C we e ained on he same day as he ac ual MRI scan bu all child en (excep o one) in s udy A and B we e ained be ween 1 and 28 (mean ¼4.30) days be o e he ac ual MRI scans, a high in aclass co ela ion o he g ouping ac o s udy would ha e indica ed di e ences be ween s udies (and hus si es) such as he scanne aining p ocedu e o he scanne aining da e. Howe e , we obse ed an ex emely low in e class co ela ion o he g ouping ac o s udy. Thus, planning scanne aining as close o he ac ual scan as possible o s ong ecency e ec s, o planning scanne aining some days be o e he ac ual scan o le i “sink in” seem o be equally e ec i e (bu also see 4.4, Fu u e di ec ions). 4.1.2. Adul s Fo adul s, we ound ha head mo ion seems o inc ease ac oss he cou se o a s udy, as p io s udy scan ime p edic ed head mo ion. While his esul seems o mi o ou indings in child en, no e ha all bu wo Table 4 Fixed e ec s pa ame e es ima es o inal model o adul ’s equency o mo ion peaks ac oss he cou se o a un. pa ame e β Lowe 95% CI Uppe 95% CI SE d p In e cep 0.80019 1.46836 3.06874 1.15779 1642 0.691 0.490 Minu e o un 0.42562 0.10116 0.75008 0.16559 1642 2.570 0.010 P io unc ional segmen scan ime 0.18341 0.00516 0.36165 0.09097 1642 2.016 0.044 P io session scan ime 0.11631 0.04504 0.18757 0.03637 1642 3.198 0.001 No e: CI ¼con idence in e al, SE ¼s anda d e o , d ¼deg ees o eedom. T.W. Meissne e al.