{"paper_id":"291ce354-1a48-455f-a7ee-58d2e97f4b69","body_text":"1\t\nInvestigating\t the\t effect\t of\t channel\t pruning\t on\t functional\t near-1 \ninfrared\tspectroscopy\tdata\tcollected\tfrom\tchildren\taged\t5-24\tmonths\t\t2 \n\t3 \nAuthors\t&\tAffiliations\t4 \nSamuel\tBeaton!,\tBorja\tBlanco\",\tChiara\tBulgarelli#,\tClare\tElwell$,\tSarah\tLloyd-Fox\",&,\t5 \nEbrima\tMbye',\tSamantha\tMcCann!,\tAnna\tBlasi\tRibera$,(,\tSophie\tMoore!,',(\t6 \n\t7 \n(1)\t Department\tof\tWomen\t&\tChildren’s\tHealth,\tKing’s\tCollege\tLondon,\tLondon,\tUK\t8 \n(2)\t Department\tof\tPsychology,\tUniversity\tof\tCambridge,\tCambridge,\tUK\t9 \n(3)\t Centre\tfor\tBrain\tand\tCognitive\tDevelopment,\tBirkbeck,\t University\tof\tLondon,\t10 \nLondon,\tUK\t11 \n(4)\t Department\t of\t Medical\t Physics\t and\t Biomedical\t Engineering,\t University\t12 \nCollege\tLondon,\tLondon,\tUK\t13 \n(5)\t Department\tof\tPsychological\tSciences,\tBirkbeck\tUniversity\tof\tLondon,\tLondon,\t14 \nUK\t15 \n(6)\t Medical\tResearch\tCouncil\tUnit,\tThe\tGambia\tat\tthe\tLondon\tSchool\tof\tHygiene\t16 \nand\tTropical\tMedicine,\tThe\tGambia\t17 \n\t18 \n\t19 \n†\tJoint\tsenior\tauthors\t20 \n\t21 \n\t22 \n\t23 \n\t24 \n\t25 \n\t26 \n\t27 \n\t28 \nCorresponding\tAuthor:\t29 \nSamuel\tBeaton\t30 \nEmail:\tsamuel.1.beaton@kcl.ac.uk\t31 \nAddress:\tDepartment\tof\tWomen\tand\tChildren’s\tHealth\t32 \nSchool\tof\tLife\tCourse\tand\tPopulation\tSciences\t33 \nAddison\tHouse\t34 \nGuy’s\tCampus\t35 \nKing’s\tCollege\tLondon\t36 \nNewcommen\tStreet\t37 \nLondon38 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n2\t\nAbstract\t39 \nSignificance\t40 \nInfant\tfunctional\tnear-infrared\tspectroscopy\t(fNIRS)\tdata\tare\tparticularly\tvulnerable\t41 \nto\t noise;\t participant\t behaviour\t can\t result\t in\t motion\t artefacts\t and\t reduced\t set-up\t42 \ntimes\tcan\tcause\tpoor\toptode\tcoupling.\tAccurate\tchannel\tpruning\tis\ttherefore\tessential\t43 \nbut\tapproaches\tvary\tand\toften\tuse\tadult-derived\tthresholds,\trisking\tunnecessary\tdata\t44 \nloss.\t45 \nAim\t46 \nThis\t work\t systematically\t compared\t pruning\t approaches\t and\t parameter\t choices\t to\t47 \nevaluate\ttheir\teffects\ton\tdata\tquality\tand\tretention\tin\tinfant\tfNIRS.\t48 \nApproach\t49 \nData\t from\t 5–24\t month-old\t infants\t were\t collected\t across\t two\t cohorts,\t using\t two\t50 \nparadigms.\tChannel\tpruning\twas\tperformed\tusing\tthe\tcoefficient\tof\tvariation\t(CV)\tand\t51 \nthe\t Quality\t Testing\t of\t Near\t Infrared\t Scans\t (QT-NIRS)\t tool,\t varying\tkey\t thresholds.\t52 \nMultilevel\tmodels\tassessed\teffects\tof\tpruning\tmethod,\tparameter\tchoice,\tage,\tmotion,\t53 \nand\ttesting\tsite\ton\tsignal-to-noise\tratio\t(SNR)\tand\tchannels\tretained.\t54 \nResults\t55 \nQT-NIRS\t produced\t significantly\t higher\t SNR\t than\t CV\t pruning\t across\t nearly\t all\t age,\t56 \ntask,\t and\t cohort\t combinations,\twhen\t matched\t for\t data\t retention.\t Higher\t QT-NIRS\t57 \nthresholds\t improved\t quality\t but\t reduced\t retention.\t Motion\t prevalence\t strongly\t58 \nreduced\tboth\tSNR\tand\tretention;\ttesting\tsite\tand\tage\thad\tsmaller\tbut\tnotable\teffects.\t59 \nConclusions\t60 \nQT-NIRS\toffers\ta\tbetter\tbalance\tof\tdata\tquality\tand\tretention\tthan\tCV\tpruning.\tLower\t61 \nQT-NIRS\t thresholds\t than\t adult\t defaults\t are\t recommended\t for\t infant\t data.\t These\t62 \nfindings\t provide\t practical\t guidance\t for\t preprocessing\t pipelines\t in\t developmental\t63 \nfNIRS\tresearch.\t64 \n\t65 \nKeywords:\tinfant\tfNIRS,\tchannel\tpruning,\tprocessing66 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n3\t\n1 Introduction\t67 \n1.1 Scalp-optode\tcoupling\t68 \nIn\t a\t typical\t functional\t near-infrared\t spectroscopy\t (fNIRS)\t experiment,\t participants\t69 \nwear\ta\theadband\tor\tcap\tembedded\twith\toptodes\tto\tmonitor\tbrain\tactivity\tby\temitting\t70 \nand\tdetecting\tnear-infrared\tlight\tat\ttwo\t separate\twavelengths.\tWhen\tfitted\tsecurely,\t71 \nthe\tcap\tensures\ta\tmotion-robust\tsignal,\tenabling\tthe\tstudy\tof\ta\twide\trange\tof\tcognitive\t72 \ntasks\tand\tabilities\twith\tless\tstringent\tdemands\tfor\tstillness\tthan\tother\tneuroimaging\t73 \nmodalities\t[1].\tFor\tthis\treason,\tfNIRS\tis\twidely\tused\twith\tdevelopmental\tpopulations\t74 \n[2].\t However,\tsignal\t processing\t is\t usually\t still\t required\t to\t remove\t artefacts\tarising\t75 \nfrom\tmotion,\tpoor\tscalp-optode\tcoupling,\tand\tphysiological\tsignal\tconfounds\t[3],\t[4],\t76 \n[5].\tInfant\tdata\tis\tparticularly\tsusceptible\tto\tmotion\tand\tpoor\t scalp-optode\tcoupling,\t77 \nas\tinfants\ttypically\texhibit\tincreased\tfussiness,\tlimited\tcompliance\twith\tinstructions,\t78 \nand\t shorter\t attention\t spans,\t which\t increase\t the\t likelihood\t of\t participant\t motion,\t79 \nreduced\tcapping\ttime\tand\t difficulties\tin\t handling\tthe\timaging\theadgear\t [6].\tIn\tfact, \t80 \npoor\tcoupling\tallow s\tlight\tfrom\tthe\tsource\toptode(s)\tto\tescape ,\tor\tambient\tlight\tto\t81 \nflood\tdetector\toptode(s)\t [7].\tAffected\tchannels\tcan\texhibit \tsignal\tsaturation\t(easily\t82 \ndetectable\t by\t unrealistically\t high\t raw\t intensity\t values\t caused\t by\t excessive\t light\t83 \nreaching\t the\t detector)\t and\t greater\t variability,\twhich\t impacts\tthe\t estimation\t of\t the\t84 \nhaemodynamic\tresponse\t[8]\twhose\tamplitude\tis\talready\tlower\tand\tmore\tvariable\tin\t85 \ninfants\tthan\tin\tolder\tparticipants\t[9],\t[10].\t\t86 \n1.2 Strategies\tto\tmitigate\tpoor\toptode\tcoupling\t87 \nFitting\t the\t fNIRS\t headgear\t securely\t aids\t scalp-optode\t coupling\t [7]\t but\t is\t time-88 \nconsuming\tand\tassumes\tstable\tcoupling\tthroughout\trecording,\twhich\tis\tchallenging\t89 \nwhen\tworking\twith\tinfant\tparticipants.\tTo\treduce\tthe\timpact\tof\tpoor\tcoupling\ton\tdata\t90 \nquality,\tpost-hoc\tchannel\tpruning\t–\tthe\texclusion\tof\tdata\tfrom\tan\tentire\tchannel \t–\tis\t91 \ntherefore\t often\t required.\t An\t important\t pruning\t consideration\t is \t the\t trade-off\t92 \nbetween\t data\t retention\t and\t quality:\t removing\t poorly-coupled\t channels\t improves\t93 \noverall\tsignal\tquality\tand\tmitigates\tthe\timpact\tof\tsuperfluous\tsignals,\tbut\treduces\tthe\t94 \nnumber\t of\t remaining\t channels\t and\t participants\t available\t for\t analysis\t[6].\t This\t is\t95 \nparticularly\timportant\tin\tinfant\tresearch\tgiven\tthe\talready\thigh\tattrition\trates\tdue\tto\t96 \nlow\tattention\tspan\tand\t susceptibility\tto\t fussiness\t[11].\tPruning\tmethod\tselection\tis\t97 \nimportant,\t and\t manually\t pruning\t channels\t is\t subjective\t and\t time-consuming\t [8],\t98 \nespecially\t for\t high-density\t imaging\t arrays.\t Two\t methods\t are\t often\t used\t to\t prune\t99 \nchannels:\t the\t coefficient\t of\t variation\t (CV)\t or\t the\t‘Quality\t Testing\t of\t Near-InfraRed\t100 \nSpectroscopy’\t(QT-NIRS)\ttool.\t\t101 \nCV\tpruning\tquantifies\tthe\trelative\tsignal\tvariability.\tChannels\tare\tpruned\tif\tthe\t102 \nCV\t for\t either\t wavelength\t falls\t below\t a\t particular\t threshold\t [12],\t [13],\t if\t the\t CV\t103 \ndifference\t between\t wavelengths\t exceeds\t a\t threshold\t[14],\t or\t both\t[15].\t While\t this\t104 \nmethod\tis\tfaster\tand\tless\tsubjective\tthan\tmanual\tpruning, \ta\tparameter\tchoice\tis\tstill\t105 \nrequired\t and\tsome\tlevels\tof\t variability\t-\twhich\tis\texpected\tin\tthe\ttask -based\tfNIRS\t106 \nsignal\tdue\tto\tthe\tevoked\thaemodynamic\tresponse\t[16]\t-\tmay\tbe\tinterpreted\tas\tsignal\t107 \nnoise.\t Additionally,\t CV\t pruning\t only\t examines\t the\t signal\t in\t the\t time\t domain,\t108 \npotentially\toverlooking\timportant\tsignal\tproperties.\t\t109 \n\t QT-NIRS\tutilizes\tobjective\tsignal\tmeasures\tfrom\tboth\tthe\ttime-\tand\tfrequency\t110 \ndomains\t[7],\t[16],\tproviding\ta\tmore\tcomprehensive\tconsideration\tof\tsignal\tquality. \t111 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n4\t\nChannels\tare\tpruned\tvia\t signal\tcharacteristic\tassessments\tin\ttwo\tdomains\tusing\tthe\t112 \nscalp\tcoupling\tindex\t(SCI)\tand\tpeak\tspectral\tpower\t(PSP) .\tThe\t SCI\tis\ta\ttime -domain\t113 \napproach,\t which\t assesses\t correlation\t between\t wavelengths\t within\t the\t cardiac\t114 \nfrequency\t band\t after\t bandpass\t filtering\t the\t signal,\t with\t high\t values\t indicative\t of\t a\t115 \nstrong\tcardiac\tcomponent\t[17].\tThis\tincorporates\ta\ttime-domain\tsignal\tcharacteristic,\t116 \nbut\t also\t risks\t retaining\t channels\t with\t high\t correlation\t due\t to\t motion-induced\t117 \nartefacts.\t To\t address\t this,\t QT-NIRS\t incorporates\t a\t second\t measure\t based\t on\t the\t118 \nfrequency\tdomain,\tPSP,\twhich\tdetects\tstrong,\trecurrent\toscillations\tin\tthe\tsignal.\tHigh\t119 \nPSP\t values\t in\t the\t cardiac\t frequency\t band\t likely\t correspond\t to\tcardiac\t pulsations,\t120 \nwhereas\tcomponents\twith\tinconsistent\tor\tvarying\tfrequencies\tusually\tresult\tin\tlower\t121 \nPSP\t values.\t QT -NIRS\t utili zes\t strengths\t from\t manual\t pruning\t (physiological\t122 \ngrounding,\t consideration\t of\t both\t time-\t and\t frequency\t domains)\t and\t CV\t pruning\t123 \n(objectivity;\t efficiency)\t making\t it\t a\t robust\t tool\t for\t assessing\t data\t quality\t at\t the\t124 \nchannel\tlevel.\t125 \nRecently,\tQT-NIRS\thas\tbeen\tincreasingly\tadopted\tin\tinfant\tfNIRS\tresearch\t[18],\t126 \n[19]\t yet\t independent\t comparisons\t with\t other\t pruning\t approaches\t are\t yet\t to\t be\t127 \nestablished,\tand\tempirical\testimations\tof\t SCI\tand\tPSP\tparameters\tare\tavailable\tonly\t128 \nfor\tadult\tparticipants\t [7],\t[16].\tThis\thighlights\tthe\tneed\tto\trefine\tits\timplementation\t129 \nwith\tinfant\tdata.\t In\taddition\tto\tbehaviour,\tboth\tphysiological\tand\tanatomical\tfactors\t130 \nmay\t affect\t channel\t pruning\t in\t infants:\t they\t have\t thinner\t scalp\t tissue\t and\t higher\t131 \ncardiac\tsignal\tfrequencies\t(~1.3–3.2\tHz\tat\trest,\tcompared\tto\t~1–1.7\tHz\tin\tadults)[20],\t132 \n[21].\tThe\tformer\tmay\tresult\tin\ta\tweak\tsuperficial\tcardiac\tsignal,\twhereas\tthe\tlatter\t133 \nmay\t result\t in\t coarse\t representations\t of\t the\t infant\t cardiac\t signal\t by\t fNIRS\t134 \ninstrumentation\tsampling\trates\toptimized\tfor\tadult\tparticipants\t[22].\tFurther,\tsignal\t135 \nquality\tcan\tbe\tdetrimentally\taffected\tby\tskin\tand\thair\tcolour,\thair\ttype,\tage\tand\teven\t136 \nhead\t size\t with\t darker\t skin\t pigmentation\t and\t thicker\t hair\t corresponding\t to\t poorer\t137 \nsignal\tquality\twhen\tcompared\tto\tother\tskin\tand\thair\ttypes\t [23],\tillustrating\tthe\tneed\t138 \nfor\tinclusion\tof\tless-frequently\tsampled\tpopulations\tin\tfNIRS\tstudies.\t139 \nAgainst\tthis\tbackdrop,\tthe\tobjectives\tof\tthis\twork\tare\tto:\t140 \n(a) Compare\t QT-NIRS\t as\t a\t pruning\t method\t against\t pruning\t using\t CV,\t which\t is\t141 \nused\tfrequently\tby\tfNIRS\tusers\t\t–\tand\tprovides\ta\tbaseline\tfor\tchannel\tpruning\t142 \nvia\ta\tpreviously\temployed\tmethod\tand\tparameters\t143 \n(b) Investigate\t contextual\t and\t data-derived\t measures\t which\t affect\tdata\t quality\t144 \nand\tchannel\tretention\t(with\ta\tparticular\tfocus\ton\tQT-NIRS\tparameter\tchoices,\t145 \nage\tand\tmotion\tincidence)\t146 \n(c) Provide\tguidance\ton\tchannel\tpruning\tand\tQT-NIRS\tuse\tfor\tinfant\tparticipants\t147 \nTo\t achieve\t this,\t fNIRS\t data\t from\t the\t Brain\t Imaging\t for\t Global\t HealTh\t (BRIGHT)\t148 \nProject\t[24],\ta\tlongitudinal\tstudy\tof\tinfant\tdevelopment\tin\tKiang\tWest \t(The\tGambia)\t149 \nand\tCambridge\t(UK),\twere\tanalyzed.\tThe\tanalyses\tin\tthis\twork\tincorporates\tdata\tfrom\t150 \ntwo\tdifferent\texperimental\tparadigms,\tcollected\tfrom\tboth\tthe\tGambian\tand\tUK\tsites\t151 \nand\t pertaining\t to\t participants\t with\t physical\t and\t behavioural\t characteristics\t from\t152 \nboth\t a\t commonly -sampled\t and\t a\t more\t underrepresented\t population\t in\t fNIRS\t153 \nresearch\t[23].\tThe\tlongitudinal\tnature\tof\tthe\tdata\t(with\tfive\ttime\tpoints\tover\tthe\tfirst\t154 \ntwo\tyears\tof\tlife)\t further\tenables\tthe\tinvestigation\tof \tthe\teffects\tof\tage\tacross\tearly\t155 \nchildhood\twhile\taccounting\tfor\tvariability\tin\tcohort\tand\ttask.\t156 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n5\t\n\t Based\ton\tprior\tliterature\tand\texploratory\tfindings,\tthe\tfollowing\thypotheses\t157 \nwere\tformulated:\t158 \n(1) QT-NIRS\tbased\tpruning\twill\tresult\tin\ta\tgreater\tbalance\tof\tdata\tquality\tand\t159 \nretention\t compared\t to\t CV,\t due\t to\t its\t multidomain\t examination\t of\t signal\t160 \ncharacteristics\t161 \n(2) Motion\t will\t be\t negatively\t associated\t with\t signal\t quality\t and\t retention,\t162 \nbecause\tof\t(i)\tincreased\tlikelihood\tof\tlatent,\tundetected\tartefacts\tin\tdata\t163 \nwhere\t more\t motion\t is\t detected,\t (ii)\t displacement\t of\t optodes\t affecting\t164 \nsignal\tquality\tafter\tmotion\tartefacts,\tor\tboth.\t165 \n(3) Data\tquality\tand\tretention\twill\tbe\tdiminished\tin\tthe\tGambian\tcohort,\tgiven\t166 \nthe\t increased\t probability\t of\t darker,\t coarser\t hair\t types\t interfering\t with\t167 \noptode-scalp\tcoupling\t168 \n2 Methods\t169 \n2.1 Data\t170 \n2.1.1 Participants\t171 \nParticipants\t were\t recruited\tinto\t the\t BRIGHT\t project,\t from\t early\t 2016\t to\t February\t172 \n2018,\t and\t fNIRS\t data\twere\t collected\t when\t infants\t were\t\t 1-,\t 5-,\t 8-,\t 12-,\t 18-\t and\t 24\t173 \nmonths\t(hereafter\t𝑥mo\tfor\t𝑥\tmonths\tof\tage),\tplus\ta\tfollow-up\tbetween\t3\tand\t5\tyears\t174 \nof\tage\t[25],\t[26].\tThe\t1mo\tfNIRS\tprotocol\twas\tlimited\tto\tauditory\tstimuli\twith\tsleeping\t175 \nparticipants\t[27],\tlikely\tinducing\tinfant\tmotion\twith\ta\tdifferent\tnoise\tprofile;\tat\t3 -5\t176 \nyears,\ta\tdifferent\tfNIRS\tcap\tfor\tdata\tcollection\twas\tused\tand\tdata\tat\tthis\tage\twere\tnot\t177 \ncollected\tin\tthe\tUK\tsite.\tTo\tenable\tmatched\tdataset\tcomparisons,\tdata\tfrom\tthe\tother\t178 \n5\ttime\tpoints\t(5-,\t8-,\t12-,\t18-\tand\t24mo)\twas\ttherefore\tused\tin\tthis\twork.\tParticipants\t179 \nmet\t the\t inclusion\t criteria\t if\t infants\t were\t born\t at\t 37 –42\t weeks’\t gestation\t (both\t180 \ncohorts)\tand\thad\ta\tminimum\tbirth\tweight\tof\t2.5\tkg\t(UK\tonly).\t181 \nAfter\t applying\t exclusion\t criteria,\t a\t total\t of\t 204\t mother-infant\t dyads\t were\t182 \nincluded\t in\t the\t Gambian\t cohort;\t of\t these,\t 185\t remained\t at\t the\t 24mo\t timepoint.\t183 \nPregnant,\t Mandinka -speaking\t women\t were\t recruited\t during\t routine\t antenatal\t184 \nclinical\t assessments\t at\tMRCG@LSHTM\t Keneba\t Field\t Station\tby\t fieldworkers\t in\t the\t185 \nGambian\t BRIGHT\t Project\t team.\t An\t information\t sheet\t and\t consent\t form\t written\t in\t186 \nEnglish\t were\t provided\t to\t potential\t recruits\t then\t explained\t fully\t in\t Mandinka\t by\t a\t187 \nstudy\tstaff\tmember.\tfNIRS\tdata\tcollection\ttook\tplace\tat \tMRC\tUnit\tThe\tGambia\tat\tthe\t188 \nLondon\t School\t of\t Hygiene\t and\t Tropical\t Medicine\t (‘MRCG@LSHTM’)\t Keneba\t Field\t189 \nStation.\tEthical\tapproval\twas\tgranted\tby\tthe\tjoint\tGambia\tGovernment/MRC\tEthics\t190 \nCommittee\tunder\tthe\ttitle:\t‘Developing\tbrain\tfunction\tfor-age\tcurves\tfrom\tbirth\tusing\t191 \nnovel\tbiomarkers\tof\tneurocognitive\tfunction’,\tSCC\tnumber\t1451v2.\t192 \n\t 61\t mother-infant\t dyads\t were\t enrolled\t in\t the\t UK\t from\t the\t Rosie\t Hospital,\t193 \nCambridge\t University\t Hospitals\t NHS\t Foundation\t Trust.\t Information\t about\t BRIGHT\t194 \nwas\t provided\t during\t antenatal\t appointments,\t with\t families\t expressing\t an\t interest\t195 \ncontacted\t and\t recruited\t subsequently\t via\t email\t or\t phone\t call. \t Data\t collection\t196 \nprimarily\t took\t place\t at\t Evelyn\t Perinatal\t Imaging\t Centre\t at\t Rosie\t Hospital,\t197 \nAddenbrooke’s\tHospital,\tCambridge,\tand\tto\ta\tlesser\textent\tat\tthe\tCentre\tfor\tBrain\tand\t198 \nCognitive\t Development\t in\t Cambridge\t [24],\t [28].\t Ethical\t approval\t was\t given\t by\t the\t199 \nNational\t Research\t Ethics\t Service\t Committee\t East\t of\t England\t (REC\t reference\t200 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n6\t\n13/EE/02000);\tinformed\twritten\tconsent\twas\tobtained\tfrom\tall\tparents/carers\tprior\t201 \nto\tparticipation.\t202 \n2.1.2 fNIRS\tParadigms\t203 \nSocial/Non-Social\tparadigm\t204 \nFull\t details\t of\t the\t Social/Non-Social\t (SNS)\t task\t paradigm\t can\t be\t found\t in\t[15]\t and\t205 \n[29].\t Briefly,\t the\t paradigm\t consisted\t of\t alternating\t visual\t social\t (silent),\t auditory\t206 \nsocial\tand\tauditory\tnon -social\tstimuli.\tStimuli\trepeated\tuntil\tthe\tparticipant\tbecame\t207 \nbored\t or\t fussy\t or\t the\t end\t of\t the\t task\t was\t reached;\t inter-stimulus\t baselines\t varied\t208 \nbetween\t9\tto\t12\tseconds.\t\t209 \n\t Visual\tsocial\tstimuli\tconsisted\tof\tfull\tcolour,\tlife-sized\tvideos\tof\tadults\tfrom\tthe\t210 \nsame\tpopulation\tas\tthe\tparticipant,\ton\ta\t24 -inch\tscreen\t~100cm\taway.\tThroughout,\t211 \nadult\t actors\t in\t the\t video\t either\t moved\t their\t eyes\t or\t played\t ‘hand-games’\t for\t 9-12\t212 \nseconds.\tActors,\ttheir\tactions,\tand\tconcurrent\tspoken\tauditory\tstimuli\twere\tvaried\tto\t213 \nprevent\t anticipatory\t brain\t activity.\t Auditory\t stimuli\t were\t non-synchronized\t to\t the\t214 \nvideo\tin\tterms\tof\tboth\tduration\t(8\tseconds)\tand\tcontent,\twith\tenvironmental\tsounds\t215 \nfor\tnon-social\tstimuli\tand\tnon-vocal\tspeech\tsounds\tfor\tsocial\tstimuli.\tSounds\tin\teach\t216 \ncondition\t(social\tand\tnon-social)\twere\tmatched\tfor\tduration\tand\tsound\tintensity.\t217 \nHabituation\tand\tnovelty\tdetection\t(HaND)\tparadigm\t218 \nThe\t experimental\t paradigm\t included\t 25\t trials,\t each\t consisting\t of\t a\t spoken\t 8sec\t219 \nsentence\tin\tthe\tfamily’s\tfirst\tlanguage\t(i.e.\tEnglish\tor\tMandinka)\tfollowed\tby\t10sec\tof\t220 \nsilence.\tThe\tfirst\ttrial\twas\tpreceded\tby\tat\tleast\t10sec\tof\tsilence,\tacting\tas\ta\tbaseline.\t\t221 \nThe\tsame\trecording,\twith\ta\tfemale\tvoice,\twas\tused\tfor\ttrials\t1 -15;\ta\tdifferent,\t222 \nmale\tvoice\twas\tused\tfor\ttrials\t16-20;\tfinally,\tthe\toriginal,\tfemale\trecording\twas\tagain\t223 \nused\tfor\ttrials\t21 -25.\tThe\tstimulus\tsentence:\t “Hi\tbaby!\tHow\tare\tyou?\tAre\tyou\thaving\t224 \nfun?\tThank\tyou\tfor\tcoming\tto\tsee\tus\ttoday.\tWe're\tvery\thappy\tto\tsee\tyou”\twas\ttranslated\t225 \nto\tMandinka\tto\tmaintain\tthe\tsame\tsemantic\tmeaning.\t226 \n\t Technical\t detail\t on\t the\t recording,\t processing\t and\t playback\t of\t the\t auditory\t227 \nstimulus\tcan\tbe\tfound\tin\tprevious\twork\t[14],\t[26].\t228 \n2.1.3 Data\tcollection\t229 \nCustom-made\t headgear\t was\t fitted\t after\t head\t measurements\t (head\t circumference,\t230 \nand\tear-to-ear\tboth\taround\tthe\tforehead\tand\tover\tthe\ttop\tof\tthe\thead)\thad\tbeen\ttaken\t231 \nto\t aid\t with\t the\t alignment\t of\t fNIRS\t headgear\t with\t the\t 10/20\t system\t anatomical\t232 \nlandmarks.\t Headgear\t consisted\t of\t custom -built\t stretchy\t silicone\t headbands\t to \t233 \nincrease\tfriction\tand\tprevent\tslippage,\twith\tattached\tprobes\tinto\twhich\toptodes\twere\t234 \nclipped.\tOptodes\twere\tdesigned\tto\taccommodate\t glass\toptic\tfibres\tat\tright-angles\tto\t235 \nallow\tthem\tto\tsit\tflush\ton\tthe\tscalp.\tThe\theadband\twas\tfastened\taround\tthe\thead\tto\t236 \nprovide\teven\tpressure\tover\tthe\tbase\tof\tthe\tprobes\t [30],\t[31].\tIn\tthe\tleft\themisphere,\t237 \nheadgear\t was\t placed\t such\t that\t source\t 4\t in\t Figure\t 1\t was\t centered\t abo ve\t the\t238 \npreauricular\t point,\tso\t that\tthe\t channel\t it\t formed\t with\t the\t detector\tlocated\t directly\t239 \nbehind\tit\tsat\tabove\tT3\tin\tthe\t10-20\tsystem;\tthe\tequivalent\tright\themisphere\tchannel\t240 \nwas\tabove\tT4.\tThe\tarray\tangle\twas\tguided\tby\tthe\theadband,\twhich\twas\tplaced\ton\tthe\t241 \nhead\tso\tthat\tit\ttouched\tthe\tjoin\tbetween \tthe\tear\tand\thead\tand,\tfrontally,\tlay\tover\tthe\t242 \ninfant\tbrow\tline\t(through\tFp1\tand\tFp2\tin\tthe\t10-20\tsystem)\t[32].\t243 \nThe\t headgear\t was\t designed\t to\t record\t responses\t bilaterally\t from\t auditory-244 \nassociative\t brain\t regions,\t including\t the\t inferior\t frontal\t gyrus\t (IFG),\t middle\t and\t245 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n7\t\nsuperior\t temporal\t regions,\t and\t the\t temporo-parietal\t junction\t[14],\t [33],\t [34].\t The\t246 \nfNIRS\tarray\tcomprised\t17\tchannels\tin\teach\themisphere\twith\ta\tsource -detector\t(S-D)\t247 \ndistance\tof\t2cm,\tcorresponding\tto\ta\tpenetration\tdepth\tof\t~1cm\tfrom\tthe\tskin\tsurface\t248 \nand\t thus\t permitting\t measurement\t of\t the\t gyri\t and\t superficial\t sulci\t[14].\t fNIRS\t data\t249 \nwere\t collected\t using\t this\t array\t design\t with\t the\t NTS\t optical\t topography\t system\t250 \n(Gowerlabs\tLtd.,\tUK)\twith\ta\tsampling\tfrequency\tof\t10Hz\tand\tsource\twavelengths\tof\t251 \n850\tand\t780nm.π\t252 \n\t Infants\tsat\ton\ta\tcarer’s\tlap\tduring\t data\tacquisition.\tCarers\twere\tdiscouraged\t253 \nfrom\tinteracting\twith\tthe\tinfant\tto\tattempt\tto\tminimize\tconfounding\tstimuli\thowever\t254 \ninfants’\tattention\twas\tengaged ,\tif\tnecessary,\twith\t(non-social,\tnon-auditory)\tbubble-255 \nblowing\t and\t silent\t demonstration\t of\t soft\t toys,\t which\t also\tminimized\t infant\t head\t256 \nmovement.\tThe\tHaND\ttask\twas\tpart\tof\ta\tlarger\tbattery\tof\tfNIRS\tassessments\twith\ta\t257 \ntotal\trecording\ttime\tof\t~\t21min\t30 s\t(6min\tsocial\ttask;\t4min\tfunctional\tconnectivity\t258 \ndata\t acquisition;\t 7min\t 30 s\t HaND;\t 4min\t further\t functional\t connectivity).\t Where\t259 \npossible,\t paradigms\t were\t completed\t uninterrupted;\t sessions\t were\t paused\t and\t260 \nsubsequently\tresumed\tin\tthe\tevent\tof\tinfant\tdiscomfort.\t[35]\t261 \n2.2 Channel\tPruning\t262 \n2.2.1 Pre-pruning\tsteps\t263 \nFirst,\t channels\t were\t excluded\t from\t analyses\t if\t their\t minimum\t light\t intensity\t value\t264 \nwas\t less\t than\t 3e-4,\t based\t on\t previous\t experience\t with\t the\t NTS\t system\t[14];\t these\t265 \nwere\tlabelled\t‘channels\twith\tsignal\textrema’\t(CSE).\t Analysis\tof\tthe\tpruning\tmethods\t266 \nwas\tconducted\ton\tmotion -free\tsegments.\tTo\tfind\tmotion -free\tdata,\tmotion\tartefacts\t267 \nwere\t detected\t using\thmrMotionArtifactByChannel\t function\t from\t Homer2\t[36]\t with\t268 \nestablished\t infant\t fNIRS\t preprocessing\t parameters:\t tMotion\t =\t 1,\t tMask\t =\t 1,\t269 \nSTDEVthresh\t=\t15 \tand\tAMPthresh\t=\t0.4\t [37].\tData\tfor\teach\tchannel\twas\tsplit\tinto\t3\t270 \nsecond\t windows\t per\t channel,\t as\t this\t aligned\t with\t QT-NIRS\t temporal\t segmentation\t271 \nand\tthus\tavoided\tadditional\tprocessing\tcomplexities.\tWindows\twere\texcluded\tfrom\t272 \npruning\tanalyses\tif\tthey\tcontained\tartefacts\tat\tany\tpoint.\tIf\tthe\tdata\tfor\ta\tparticular\t273 \nchannel\tat\tone\twavelength\twas\texcluded,\tdata\tfor\tboth\twavelengths\twas\tremoved.\t\t274 \nFigure\t 1:\t Array\t layout\t during\t data\t collection.\t Red\t dots\t\nindicate\tposition\tof\toptodes\tcentred\tabove\tthe\tpreauricular\t\npoint\tduring\theadgear\tfitting.\t\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n8\t\nThe\ttwo\tchannel\tpruning\tmethods\tassessed\tin\tthis\twork\twere\timplemented\t275 \nusing\tcustom-written\tscripts\tin\tMATLAB \t[26].\tBoth\tpruning\tmethods\tdescribed\tuse\t276 \nraw\tlight\tintensity\tdata\tas\tinput.\t\t\t277 \n2.2.2 CV\tPruning\t278 \nCV\tpruning\twas\tconducted\tusing\tan\tin -house\tscript,\tand\tCV\titself\twas\tcalculated\ton\t279 \nmotion-free\tdata\tfor\teach\twavelength\tand\tchannel\tusing\tthe\tequation:\t280 \n𝐶𝑉 = 𝜎\n|𝜇|\t281 \nwhere\t𝜎\tand\t𝜇\tare\tthe\tstandard\tdeviation\tand\tmean\tof\tthe\tlight\tintensity\tsignal\tin\ta\t282 \nchannel,\trespectively\t[38].\tChannels\twere\tpruned\tif\tthe\tdifference\tbetween\tCV\tvalues\t283 \nfor\teach\twavelength\texceeded\t0.2\tbased\ton\tprevious\texperience\twith\tthe\tsame\tdata\t284 \nand\tsystem\t[14].\t285 \n2.2.3 QT-NIRS\tpruning\t286 \nQT-NIRS\t was\t implemented\t using\t the\t function\t qtnirs\t (available\t at\t287 \nhttps://github.com/lpollonini/qt-nirs\tat\tthe\ttime\tof\twriting)\tfor\tprecise\tcontrol\tover\t288 \nthe\tpruning\tand\tquick,\trepeated\tprocessing\tof\tthe\tlarge\tvolume\tof\tdata.\t\t289 \n\t More\tdetail\ton\tQT -NIRS\tcan\tbe\tfound\tin\tpublications\tdescribing\tthe\tmethods\t290 \n[7],\t[17]\tbut\tan\toutline\tis\tprovided\there.\tFirst,\tbandpass\tfiltering\tis\tconducted\tto\tretain\t291 \nonly\tthose\tfrequencies\tin\tthe\tcardiac\tband ,\t~1.3 − 3.2𝐻𝑧\tin\tthe\tcase\tof\tinfants \t[20].\t292 \nThe\t cross -correlation\t of\t contemporaneous\t (zero -lag)\t wavelength\t signals\t in\t the\t293 \ncardiac\tfrequency\tband,\t(i.e.,\tSCI),\tis\tthen\tcalculated:\t294 \n𝑆𝐶𝐼\t = \t 𝑥̅)! ⊗ \t 𝑥̅)\"\t(0)\t295 \nwhere\t𝑥̅)* \trepresents\tthe\tlight\tintensity\tsignal\tfor\twavelengths\t𝑖 = 1, 2\tin\tmotion-free\t296 \nsignal.\tPSP\tis\tthe\tmaximum\tsignal\t value\tin\tthe\tfrequency\tdomain,\trepresenting\tthe\t297 \ndominant\toscillation\tin\tthe\tbandpass -filtered\tsignal\tand\tpresumed\tto\tcorrespond\tto\t298 \nthe\tcardiac\tfrequency\tin\twell-coupled\tdata.\tThe\trecorded\tsignal\tis\tdivided\tinto\tequal-299 \nlength\twindows,\tand\tboth\tSCI\tand\tPSP\tare\tcalculated\tfor\teach\twindow.\tA\twindow's\t300 \nsignal\tis\tconsidered\tof\tsufficient\tquality\tif\tboth\tthe\tcalculated\tSCI\tand\tPSP\texceed\tthe\t301 \nuser-defined\tthresholds,\tsci_threshold\tand\tpsp_threshold,\trespectively.\t302 \n\t The\t focus\t was\t on\t the\t alteration\t of\t sci_threshold\t and\t psp_threshold\t during\t303 \nanalysis,\t since\t each\t provides\t a\t threshold\t for\t one\t of\t the\t key\t measures\t of\t optode\t304 \ncoupling\tquality\tused\tto\tassess\tdata\tquality\twith\tQT-NIRS,\tand\tadult\treference\tvalues\t305 \nof\tsci_threshold\t= 0.8\tand\tpsp_threshold\t= 0.1\tare\tavailable\tfor\tthese\ttwo\tparameters\t306 \n[16].\t Default\t parameters\t were\t used\t for\t window\t size\t (3\t seconds)\t and\t quality\t307 \nthreshold,\t or\t q_threshold\t (0.75),\t which\t prunes\t channels\t with\t less\t than\t 75%\t of\t308 \nwindows\tmeeting\tboth\tSCI\tand\tPSP\tthreshold\tvalues.\t309 \n2.3 Statistical\tanalyses\t310 \nMulti-level\tmodels\t(MLMs),\ta\tform\tof\tlinear\tregression\tthat\testimates\tvariance\tat\t311 \nmultiple\tlevels,\twere\tused\tfor\tstatistical\tanalyses,\tto\teffectively\taccount\tfor\trepeated\t312 \nmeasures,\thierarchical\tdata\tstructures\t(including\tparticipant-\tand\tchannel-level\t313 \nmeasures),\tand\tmissing\tdata\t[39].\tAll\tmodels\twere\tfitted\tin\tR\t4.4.1\t[40]\tusing\tthe\t314 \nlme4\tpackage\t[41].\tAll\tmodels\tincluded\trandom\tintercepts\tfor\teach\tparticipant\tto\t315 \naccount\tfor\tindividual\tvariation.\tFinal\tmodels\tused\tfor\tanalysis\tare\tdescribed\tin\t316 \n‘Models\tand\tAnalyses’.\t317 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n9\t\n2.3.1 Measures,\tOutcomes\tand\tEffects\t318 \nConsidering\tthe\tanticipated\tdata\tquality/inclusion\ttrade-off,\twhich\tis\tcentral\tto\t319 \ndecision\tmaking\taround\tpreprocessing\tof\tfNIRS\tdata\t[6],\tthe\tperformance\tof\tpruning\t320 \nmethods\tand\tparameters\twere\tassessed\tusing\ttwo\tmetrics:\t(i)\tsignal\t-to-noise\tratio\t321 \n(SNR)\tand\t(ii)\tchannel\tinclusion/exclusion\tpercentage.\t\t322 \nThe\teffects\tof\tother\tfactors\tsuch\tas\tage,\tmotion,\tand\tsignal\textrema,\twere\t323 \nincluded\tas\tpredictors.\tOther\tmeasures\twere\tmore\tparticular\tto\tthis\tdataset\t-\tsuch\t324 \nas\ttask,\tcohort\tand\toptode\tposition.\tThey\twere\tincluded\tas\tcovariates\tto\taccount\tfor\t325 \nthe\tvariability\tthey\tmay\tcause.\t326 \nThe\tvariables\tused\tare\tlisted\tbelow,\twith\tletters\tcontained\tin\tbrackets\t327 \nindicating\twhether\tthey\twere\tused\tas\toutcome\tvariables\t(O),\tpredictors\t(P),\tor\t328 \ncovariates\t(C).\tFull\trationale\tcan\tbe\tfound\tin\tSupplementary\tMaterials\t1.\t329 \n2.3.1.1 Task-relevant\tchannel\tsignal-to-noise\tratio\t(O)\t330 \nSignal\tquality\tafter\tpruning\twas\tmeasured\ton\tthe\tincluded\tchannels\tusing\tthe\tsignal-331 \nto-noise\tratio\t(SNR):\t332 \n𝑆𝑁𝑅 = 20 log!+\n𝜇\n𝜎\t\t\t333 \nwith\t𝜇\tand\t𝜎\tthe\tmean\tand\tstandard\tdeviation\tof\tthe\tsignal,\trespectively\t[5].\tSNR\twas\t334 \ncalculated\tin\t‘task-relevant\tchannels’\t(TRCs)\t–\tchannels\twhere\ta\ttrue\thaemodynamic\t335 \nsignal\t was\t observed.\t Age-specific\t TRCs\t were\t determined\t using\t prior\t analyses:\t the\t336 \nSNS\tTRCs\twere\ttaken\tfrom\twork\tby\tBenerradi\tand\tcolleagues\t[25]\twhereas\tthe\tHaND\t337 \nTRCs\twere\ttaken\tfrom\twork\tby\tBlasi\tand\tcolleagues\t [26].\tTRCs\tfor\teach\tage\tand\ttask\t338 \nwere\t those\t which\t exhibited\t a\t haemodynamic\t response\t in\t both\t chromophores\t for\t339 \neither\t cohort,\t except\t the\t SNS\t task\t at\t 18mo:\t only\t 1\t channel\t met\t this\t criterion,\t so\t340 \nchannels\twere\tadded\tfor\tthis\tage/task\tcombination\tif\tthey\twere\tin\tthe\tset\tof\tTRCs\tfor\t341 \nat\tleast\ttwo\tother\tages\tfor\tthe\tSNS\ttask.\tThis\toutcome\twas\tnamed\tthe\ttask -relevant\t342 \nchannel\tsignal-to-noise\tratio\t(TRC\tSNR).\t343 \n2.3.1.2 Channels\tretained\t(O)\t344 \nThe\tnumber\tof\tchannels\tper\tparticipant\tincluded\tafter\tpruning\tusing\tthe\tdescribed\t345 \nmethod\tand\tparameter(s)\twas\tsummed\t–\tthis\twas\tlabelled\t‘Channels\tRetained’\t(CR).\t346 \n2.3.1.3 SCI\tThreshold\t(P)\t347 \nsci_threshold\tvalues\tranging\tfrom\t0.05\tto\t0.9\twere\tused,\twith\tincrements\tof\t0.05.\tThe\t348 \nupper\t threshold\t of\t 0.9\t sits\t between\t the\t recommended\t value\t for\t adult\t participants\t349 \n(0.8)\t and\t the\t theoretical\t maximum\t of\t SCI\t =\t 1.\t Initial\t exploratory\t analyses\t (see \t350 \nSupplementary\t Materials\t 2)\t for\t each\t age/task/cohort\t combination\t indicated\t that\t351 \neven\tvery\tlow\tSCI\tvalues\tcontinued\tto\talter\tsignal\tquality,\tso\tthe\tentire\trange\tof\tlower\t352 \nvalues\twas\tused.\t353 \n2.3.1.4 PSP\tThreshold\t(P)\t354 \nThe\t psp_threshold\t value\t was\t varied,\t ranging\t from\t 0.005\t to\t 0.1,\t with\t increments\t of\t355 \n0.005.\tThe\tupper\tvalue\tof\t0.1\tis\tthe\trecommended\tthreshold\tfor\tadult\tparticipants.\t356 \nValues\t which\t spanned\t the\t entire\t possible\t range\t lower\t than\t this\t–\t based\t on\t initial\t357 \nresults\tfrom\tsimpler\tMLM\tanalyses\t–\twere\tused;\tthis\tis\tsupported\tby\tthe\tconsiderable\t358 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n10\t\nproportion\tof\tinfant\tPSP\tvalues\tlower\tthan\tthe\tadult\tthreshold\tapparent\tduring\tpost-359 \nanalysis\tdata\texamination\tin\tthe\tUK\tcohort\t(see\tFigure\t2).\t360 \n2.3.1.5 Percentage\tof\tmotion\t(P)\t361 \nFor\t each\t participant/age/task\t combination,\t the\t cross -channel\t mean\t of\t the\t362 \npercentage\t of\t windows\t per\t channel\t containing\t motion \t as\t identified\t by\t363 \nhmrMotionArtifactByChannel\t and\t subsequently\t excluded\t from\t analyses\t (see\t364 \n‘Pruning’)\twas\tlabelled\tthe\t‘Percentage\tof\tMotion’\t(PoM)\tproviding\ta\tmeasure\tof\tthe\t365 \nprevalence\tof\tmotion.\t366 \n2.3.1.6 Age\t(P)\t367 \nAge\t was\t included\t as\t a\t five-level\t predictor\t (5-,\t 8-,\t 12-,\t 18-,\t 24mo)\t to\t assess\t change\t368 \nwith\tage.\t369 \n2.3.1.7 Cohort\t(C)\t370 \nCohort\twas\tincluded\tas\ta\ttwo-level\tcovariate\t(Gambia\tand\tUK).\t371 \n2.3.1.8 Task\t(C)\t372 \nTask\t was\t included\t as\t a\t two -level\t covariate\t (HaND\t and\t SNS)\t to\t account\t for\t the\t373 \ncontribution\tto\tvariance\tof\tthe\tdifferent\tparadigms.\t374 \n2.3.1.9 Channels\tpruned\tdue\tto\tsignal\textrema\t(C)\t375 \nThe\tnumber\tof\tCSE\tper\tparticipant,\tfor\teach\tage\tand\ttask,\twas\tused\tas\ta\tparticipant -376 \nlevel\tcovariate\tand\tproxy\tmeasure\tof\tpoor\toptode\tcoupling.\t377 \n\t378 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n11\t\n2.3.2 Models\tand\tanalyses\t379 \n2.3.2.1 Comparison\tof\tQT-NIRS\tand\tCV\tpruning\t380 \nIn\tline\twith\tobjective\t(a),\teach\tfNIRS\trecording\tunderwent\tchannel\tpruning\tusing:\t381 \n1. CV,\tby\tpruning\tchannels\twhere\tthe\tCV\tvalues\tfor\tdifferent\twavelength\tsignals\t382 \ndiffered\tby\t20%\tor\tmore\t(‘CV’)\t383 \n2. sci_threshold\tonly\tat\tevery\tparameter\tvalue,\tby\tsetting\tpsp_threshold\tto\t0,\t(‘SCI\t384 \nOnly’),\tand\t385 \n3. every\t combination\t of\t sci_threshold\t and\t psp_threshold\t parameters\t listed\t in\t386 \n‘Measures,\tOutcomes\tand\tEffects’\t\t(‘Full\tQT-NIRS’)\t387 \nTop:\t bar\t charts\t showing\t the\t proportion\t low\t channel\t Average\t SCI\t and\t Average\t PSP\t\nmeasures\tin\trelation\tto\tthe\tadult\trecommended\tparameters\tof\t0.8\t(SCI)\tand\t0.1\t(PSP),\t\nand\thalf\tof\tthese\tthreshold\tvalues\t(0.4\tand\t0.05,\trespectively).\tTop -left:\tGambia\tcohort.\t\nTop-right:\tUK\tcohort.\tBottom:\tnumber\tof\tparticipants,\tby \tage\tand\tTask,\twith\tless\tthan\t\n60%\t of\t acceptable\t channels\t exhibiting\t mean\t SCI\t and\t PSP\t values\t compared\t to\t adult\t\nthreshold\t values\t of\t 0.8\t and\t 0.1,\t respectively.\tHighest\t exclusion\t rate\t was\t ~33%\t for\t UK\t\ninfants\tat\t18mo\tduring\tthe\tSNS\ttask.\t\nFigure\t2:\tData\tcharacteristics\tin\trelation\tto\tadult\tQT-NIRS\tthresholds.\t\t\nSNS \nHaND (5)\n \n \nSNS (5)\n \n \n HaND (\n8) \n \nSNS (\n8) \n \n HaND \n(12) \n \nSNS (12)\n \nHaND (18)\n \nSNS (18)\n \nHaND (\n24) \n \nSNS (24)\n \nHaND (5)\n \n \nSNS (5)\n \n \n HaND (8)\n \n \nSNS (8)\n \n \n HaND \n(12)\n \n \nSNS (12)\n \nHaND (18)\n \nSNS (18)\n \nHaND (24)\n \n \nSNS (24)\n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n12\t\nPruning\tmethod\t1 \tprovides\ta\tbaseline\tfor\tthe\tchannel\tpruning,\tusing\ta\tmethod\tand\t388 \npruning\t criteria\t which\t has\t previously\t been\t used\t for\t the\t HaND\t task\t[14].\t Pruning\t389 \napproach\t2\tpermits\tcomparison\tof\ttwo \ttime-domain\tmethods\t(methods\t1\tand\t2)\tto\t390 \nassess\tthe\timpact\tof\tincorporating\ttemporally\tspecific\tmeasures.\tPruning\tapproach\t3\t391 \nprovides\t insight\t into\t the\t benefit\t of\t additionally\t using\t a\tPSP\t threshold\t and\t pruning\t392 \nusing\tfrequency\tcharacteristics\tof\tthe\tsignal.\t\t393 \nFor\tevery\tage/task/cohort\tcombination,\teach\t sci_threshold\tparameter\tchoice\t394 \n(SCI)\t or\t combination\t of\t sci-\t and\t psp_threshold\t values\t (QT-NIRS)\t were\t given\t two\t395 \nseparate\trankings\taccording\tto\ttheir\tsimilarity\tto\tCV\tpruning,\tin\tterms\tof\tthe\tmean\t396 \nnumber\t of\t channels\t retained\t across\t all\t participants\t and\t the\t total\t number\t of\t397 \nparticipants\texcluded.\tThese\ttwo\trankings\twere\tthen\tcombined\tto\tfind\tthe\tparameter\t398 \n(SCI)\tor\tparameters\t(QT-NIRS)\tproducing\tthe\tclosest\tapproximation\tto\tdata\tretention\t399 \nprovided\tby\tCV\tpruning.\t400 \n\t For\t each\t of\t the\t 20\t age\t (5\t levels)/task\t (2\t levels)/cohort\t (2\t levels)\t401 \ncombinations,\tthe\tfollowing\tmodel\twas\tthen\tfitted\tto\tparticipant-level\tdata:\t402 \n𝑇𝑅𝐶\t𝑆𝑁𝑅\t~\t𝑃𝑟𝑢𝑛𝑖𝑛𝑔\t𝑀𝑒𝑡ℎ𝑜𝑑\t + \t (1\t|\t𝐼𝐷)\t403 \nEquation\t1\t404 \nto\t compare\t the\t effect\t of\t pruning\t methods\t (1)-(3)\t on\t signal\t quality\t whilst\t using\t a\t405 \nrandom\t intercept\t for\t each\t infant\t to\t account\t for\t inter -participant\t variability.\t To\t406 \ncorrect\tfor\tmultiple\tcomparisons,\tp-values\twere\tBonferroni-corrected.\t407 \n2.3.2.2 Effect\tof\tSCI\tand\tPSP\tthreshold\tchoice\ton\tsignal\tquality\tand\tretention\t408 \nModels\twere\tdesigned\tto\tinvestigate\tthe\teffect\tof\tsci_threshold,\tpsp_threshold,\tage,\tand\t409 \nmotion\ton\tTRC\tSNR\tand\tChannels\tRetained\t(Objectives\t(b)\tand\t(c)).\t\t410 \nTo\t examine\t potential\t combinations\t of\t theoretically\t viable\t predictors,\t411 \ncovariates,\t and\t their\t interactions\t used\t for\t each\t outcome,\ta\t systematic\t approach\t to\t412 \nmodel\t building\t was\t used\t as\t a\t first\t step,\t combining\t predictors\t in\t various\t model\t413 \nformulae\tusing\tcombinatorial\tlogic\tbefore\tassessing\tmodel\tfit.\tModel\tfit\twas\tassessed\t414 \nusing\t the\t Akaike\t Information\t Criterion\t (AIC)\t[42],\t a\t model\t selection\t criterion\t that\t415 \nbalances\tgoodness\tof\tfit\twith\tmodel\tcomplexity\t[43,\tp.\t824].\tModel\tvariables\twere\talso\t416 \nincluded\t based\t on\t the\t outcomes\t of\t the\t subsidiary\t investigations\t described\t in\t417 \nSupplementary\tMaterials\t3\twhich\tinvestigated\tthe\tfactors\taffecting\tmotion\tincidence\t418 \nand\taverage\tSCI\tand\tPSP\tmeasures\tat\tthe\tchannel\tlevel.\tThese\tadditional\tvariables\tare\t419 \nincluded\tin\tthe\tbottom\tline\tof\tEquation\t2.\t420 \nMindful\t of\t model\t convergence\t issues\t and\t overfitting,\t the\t priority\t when\t421 \nconstructing\t models\t was\t to\t include\t predictors\t of\t interest,\t plus\t interaction\t terms,\t422 \nrandom\tslopes\tand\trandom\tintercepts\twhich\tincorporated\tthem.\tThis\tresulted\tin\tthe\t423 \nfollowing\tmodel\tfor\tboth\toutcomes\tTRC\tSNR\tand\tChannels\tRetained:\t424 \n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t𝑂𝑢𝑡𝑐𝑜𝑚𝑒\t~\t\t𝑆𝐶𝐼\t ∗ \t𝑃𝑆𝑃\t + \t𝐴𝑔𝑒\t ∗ \t𝑆𝐶𝐼\t + \t𝐴𝑔𝑒\t ∗ \t𝑃𝑆𝑃\t +\t\t425 \n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t𝑆𝐶𝐼 ∗ \t𝑃𝑜𝑀\t\t + \t𝑃𝑆𝑃 ∗ \t𝑃𝑜𝑀\t\t + \t𝐴𝑔𝑒\t ∗ 𝑃𝑜𝑀\t +\t426 \n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t𝑇𝑎𝑠𝑘\t + \t𝐶𝑜ℎ𝑜𝑟𝑡 + 𝐶𝑆𝐸 + \t (1\t|\t𝐼𝐷) +\t427 \n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t(1\t|\t𝑆𝐶𝐼: 𝐶𝑜ℎ𝑜𝑟𝑡) \t + \t (1\t|\t𝑃𝑆𝑃: 𝐶𝑜ℎ𝑜𝑟𝑡)\t\t428 \nEquation\t2\t429 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n13\t\nwhere\tSCI\t=\tsci_threshold,\tPSP\t=\tpsp_threshold,\tand\tsymbol\t‘∗’\tdenotes\tan\tinteraction\t430 \nterm\t(itself\tdenoted\tusing\t‘:’)\tplus\tall\tindividual\tterms\tused\tin\tthe\tinteraction,\tas\tis\t431 \nconsistent\twith\tnotation\tused\tin\tthe\t lme4\tpackage.\tRandom\tintercept\tterms\t(bottom\t432 \nrow)\t were\t included\t due\t to\t quantitative\t support\t based\t on\t subsidiary\t MLM\t433 \ninvestigations\tinto\taverage\tSCI\tand\tPSP\tmeasures.\t\t434 \nAssessments\tof\tthe\tmodel\tresiduals\tshowed\tthat\tthey\twere\tnon-normally\t435 \ndistributed,\tso\tto\tcalculate\teffects\ta\tbootstrapping\tapproach\twas\tused.\tBootstrap\t436 \ndatasets\twere\tgenerated\tby\tsampling\trows\tfrom\tthe\toriginal\tdataset\twith\t437 \nreplacement,\tusing\ta\tfixed\trandom\tseed\tfor\treproducibility.\tThe\trelevant\tmodel\tfor\t438 \neach\tbootstrap\tsample\twas\tfitted\tusing\tthe\tlmer\tfunction\tfrom\tthe\tlme4\tpackage\tand\t439 \nextracted\tfixed\teffect\testimates.\tTo\tmitigate\tpotential\tbiases\tcaused\tby\tfitting\tmodels\t440 \nto\tdata\twith\tdifferent\tscales,\tvariables\t𝑥*\twere\tscaled\tand\tcentred:\t441 \n𝑥*′ = \t 𝑥* − 𝑥̅\n𝑠,\n\t442 \nEquation\t3\t443 \nwhere\t 𝑥*′\t is\t the\t scaled\t value,\t and\t 𝑥̅\tand\t 𝑠,\t are\t the\t sample\t mean\t and\t standard\t444 \ndeviation,\trespectively,\tof\tvariable\t𝑥.\t445 \n3 Results\t446 \nCV\tpruning\tyielded\tan\taverage\tTRC\tSNR\tof\t23.7 ± 1.66\tacross\tall\t20\tage/task/cohort\t447 \ncombinations.\tPruning\twith\tQT-NIRS\tusing\tjust\tthe\tSCI\tthreshold\tresulted\tin\ta\tmean\t448 \nincrease\tof\t2.13 ± 0.644\tin\tTRC\tSNR\tbeyond\tthat\tobtained\tusing\tCV\tpruning.\tSimilarly,\t449 \nusing\t both\t SCI-\t and\t PSP\t thresholding\t in\t combination\t resulted\t in\t a\t mean\t TRC\t SNR\t450 \nincrease\t of\t 2.16 ± 0.627\tin\t comparison\t to\t CV\t pruning\t (see\t Figure\t 3,\t and\t451 \nSupplementary\tMaterials\t4\tfor\tthe\tfull\tcomparison\tof\tTRC\tSNR\tvalues).\t452 \nFigure\t 3:\t Example\t violin\t plot\t demonstrating\t typical\t differences\t\nbetween\tobtained\tmean\tTRC\tSNR\tvalues\tusing\tCV\tpruning,\tQT-NIRS\t\nusing\t SCI\t threshold\t only,\t and\t QT-NIRS\t utilising\t both\t parameters\t\nfor\tGambian\tparticipants\tat\t12mo\tduring\tthe\tSNS\ttask.\t\n TRC SNR for 12 month infants. Task: SNS; Cohort: Gambia \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n14\t\n3.1 Comparison\tof\tQT-NIRS\tand\tCV\tpruning\t453 \n3.1.1 Pruning\tusing\tCV\tand\tQT-NIRS\t454 \nThe\t effect\t size\t and\t Bonferroni -corrected\t significance\t values\t of\t the\t 3\t contrast\t455 \nconditions\twere\tcalculated\tusing\tEquation\t1;\tresults\tare\tdisplayed\tin\t\t456 \nTable\t1.\tFor\t19\tof\tthe\t20\tage/task/cohort\tcombinations,\tsignificant\t(p\t<\t0.01)\tpositive\t457 \neffects\twere\tfound\ton\tthe\tTRC \tSNR\twhen\tcontrolling\tfor\tdata\trejection\twhen\tusing\t458 \nQT-NIRS,\t using\t either\t the\t SCI\t Only\t or\t Full\t QT -NIRS\t approach.\t In\t 17\t of\t 20\t459 \ncombinations,\t the\t differences\t were\t found\t to\t be\t statistically\t significant\t at\t the\t p\t <\t460 \n0.0001\t threshold.\t Though\t the\t effect\t was\t positive\t for\t the\t single\t remaining\t461 \ncombination\tof\t20,\tit\tdid\tnot\treach\tstatistical\tsignificance.\t462 \nTable\t1:\tCondition\tcontrasts\tbetween\teach\tof\tthe\tthree\tpruning\tmethods\t463 \n   CV vs SCI Only CV vs Both Parameters  SCI Only vs Both Parameters  \nAge \n(months) Task Cohort Significance Effect \nSize Significance  Effect \nSize Significance Effect Size \n5 HaND Gambia <0.001 20.85 <0.001 20.83 1.0 0.05 \n8 HaND Gambia <0.001 12.71 <0.001 12.75 1.0 <0.01 \n12 HaND Gambia <0.001 19.32 <0.001 19.70 1.0 0.27 \n18 HaND Gambia <0.001 17.04 <0.001 17.10 1.0 0.04 \n24 HaND Gambia <0.001 12.36 <0.001 12.12 1.0 0.17 \n5 SNS Gambia <0.001 16.97 <0.001 16.93 1.0 0.02 \n8 SNS Gambia <0.001 14.69 <0.001 14.70 1.0 0 \n12 SNS Gambia <0.001 16.95 <0.001 17.65 1.0 0.49 \n18 SNS Gambia <0.001 15.10 <0.001 15.09 1.0 <0.01 \n24 SNS Gambia <0.001 13.38 <0.001 13.38 1.0 0 \n5 HaND UK <0.001 9.40 <0.001 9.40 1.0 <0.01 \n8 HaND UK <0.001 12.90 <0.001 12.90 1.0 <0.01 \n12 HaND UK <0.001 8.70 <0.001 8.56 1.0 0.10 \n18 HaND UK <0.001 10.92 <0.001 10.92 1.0 <0.01 \n24 HaND UK 0.012 4.68 0.012 4.68 1.0 0 \n5 SNS UK <0.001 6.93 <0.001 7.80 1.0 0.62 \n8 SNS UK <0.001 7.18 <0.001 7.28 1.0 0.07 \n12 SNS UK 0.924 3.23 0.288 3.87 1.0 0.45 \n18 SNS UK 1.0 2.84 1.0 2.84 1.0 0 \n24 SNS UK 1.0 0.76 1.0 0.76 1.0 <0.01 \nMethods\tare:\tCV,\tQT -NIRS\tusing\tthe\tsci_threshold\tparameter\tonly,\tand\tQT -NIRS\tusing\t464 \nboth\t the\t sci_threshold\t and\t psp_threshold\t parameters.\tSignificance\t values\t calculated\t465 \nduring\tbootstrapping\tand\treported\tafter\tcorrection.\tEffect\tsizes\trounded\tto\t2d.p.\t466 \n3.1.2 Pruning\tusing\tSCI\tOnly\tcompared\twith\tboth\tparameters\t467 \nNo\t significant\t statistical\t differences\t were\t found\t between\t TRC\t SNR\t values\t when\t468 \npruning\t using\t SCI\t only\t and\t Full\t QT-NIRS\t approaches.\t In\t every\t case,\t however,\t the\t469 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n15\t\nmean\t TRC\t SNR\t across\t participants\t was\t higher\t when\t using\t both\t parameters\t than\t470 \nwhen\tusing\tsci_threshold\talone.\t\t471 \n\t A\trepresentative\tcomparison\tof\tthe\t three\tdifferent\tpruning\tmethods\tis\tgiven\t472 \nin\t Figure\t 3,\t with\t higher\t average\t TRC\t SNR\t values\t obtained\t using\t both\t QT -NIRS\t473 \napproaches\twhen\tcompared\tto\tCV\tpruning.\tA\thigher\tmean\tTRC\tSNR\tis\tobtained\tusing\t474 \nboth\tparameters\twhen\tpruning\twith\tFull\tQT -NIRS\tcompared\tto\tusing\tSCI\tOnly,\tbut\t475 \nwith\ta\tmore\terratic\tdistribution\tof\tvalues.\t476 \n3.2 Predictors\tof\tTRC\tSNR\tand\tChannels\tRetained\t477 \nThe\tfocus\there\tis\tprimarily\ton\treporting\tresults\tfor\tthe\toutcomes\tof\tinterest\tand\tfixed\t478 \neffects\twith\twider\tgeneralisability\tbut\tfull\tresults\tare\tincluded\tin\tFigure\t4.\tEffect\tsizes\t479 \nwere\tclassified\tas\tsmall,\tmedium\tor\tlarge\tif\tthe\tabsolute\tvalue\tof\tthe\tEstimate \twas\t<\t480 \n0.15,\t <\t 0.35,\t or\t ≥\t 0.35,\t respectively,\t using\t Cohen’s\t 𝑓\"\tthresholds\t (Cohen,\t 2013,\t481 \nchap.9).\t\t482 \n3.2.1 Predictors\tfor\tTRC\tSNR\t483 \nBoth\t SCI\t Threshold\t (𝛽 = 0.0107,\t 95%\t CI\t [0.0084, 0.0131],\t 𝑆𝐸 = 0.0012)\t and\t PSP\t484 \nThreshold\t (𝛽 = 0.0530,\t 95%\t CI\t[0.0505, 0.0556],\t 𝑆𝐸 = 0.0013)\t had\t small,\t positive\t485 \neffects\t representing\t an\t increase\t in\t signal\t quality\t for\t higher\t threshold\t values.\t The\t486 \ninteraction\t effect\t between\t PSP\t Threshold\t and\t SCI\t Threshold\t had\t a\t small,\t negative\t487 \neffect\t(𝛽 = −0.0019,\t95%\tCI\t[−0.0037, −0.0001],\t𝑆𝐸 = 0.0009).\t\t488 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n16\t\n\t489 \nFigure\t 4:\t Forest\t plots\t showing\t the\t results\t for\t the\t effect\t of\t each\t fixed\t effect\t in\t\nEquation\t2\ton\tTRC\tSNR\tand\tChannels\tRetained,\tobtained\tvia\tbootstrapping\tscaled\t\nvalues.\t\t\nTop:\tForest\tplot\tshowing\tthe\ttype\tand\trelative\teffect\tsize\tof\teach\tfixed\teffect\ton\tTRC\tSNR.\t\nBottom:\t Forest\t plot\t showing\t the\t type\t and\t relative\teffect\t size\t of\t each\t fixed\t effect\t on\t\nChannels\tRetained.\t\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n17\t\nPoM\t had\t the\t largest\t impact\t on\t TRC\t SNR,\t with\t a\t large\t negative\t effect\t (𝛽 =490 \n−0.5004,\t95%\tCI\t [−0.5028, −0.4979],\t𝑆𝐸 = 0.0012)\tcorresponding\tto\ta\tdecrease\tin\t491 \nsignal\tquality\tin\tdata\twith\tmore\tfrequent\tincidences\tof\tmotion.\tThis\tdecrease\tin\tsignal\t492 \nquality\twith\tmotion\twas\toffset\tslightly\tas\tparticipants\taged,\tdriven\tmostly\tby\ta\tmore\t493 \nmoderate\tdecrease\tin\tthe\t18mo\tdata,\tillustrated\tby\ta\tsmall\tnegative\tinteraction\teffect\t494 \nbetween\t and\t PoM\t (𝛽 = −0.0485,\t 95%\t CI\t[−0.0506, −0.4062],\t 𝑆𝐸 = 0.0011).\t Small\t495 \ninteraction\t effects\t between\t PoM\t and\t both\t SCI\t Threshold\t ( 𝛽 = 0.0027,\t 95%\t CI\t496 \n[0.0008, 0.0047],\t 𝑆𝐸 = 0.0010)\t and\t PSP\t Threshold\t ( 𝛽 = 0.0253,\t 95%\t CI\t497 \n[0.0234, 0.0272],\t 𝑆𝐸 = 0.0010)\t exhibited\t trends\t which\t saw\t high\t threshold\t values\t498 \nmitigate\t the\t detrimental\t effect\t on\t TRC\t SNR;\t the\t slightly\t larger\t main\t (PSP)\t and\t499 \ninteraction\t(PSP:PoM)\teffect\tin\tthe\tcase\tof\tPSP\tThreshold\tled\tto\tgreater\tmitigation\tof\t500 \nthe\tTRC\tSNR\tdecline\tdue\tto\tPoM\tthan\tin\tthe\tcase\tof\tSCI\t(see\tFigure\t5a).\t\t501 \nThe\t small,\t negative\t effect\t of\t Cohort\t ( 𝛽 = −0.8783,\t 95%\t CI\t502 \n[−0.8832, −0.8733],\t𝑆𝐸 = 0.0025)\tis\tnotable\tsince\tthis\tfixed\teffect\thad\tlarger\teffects\t503 \non\tother\toutcomes,\tincluding\tChannels\tRetained.\tThe\tnon-significant\teffect\tof\tPSP:Age\t504 \n(Figure\t 6b)\t is\t notable\t given\t the\t size\t of\t the\t dataset\t and\t rarity\t of\t non-significant\t505 \npredictors\tthroughout\tthis\twork.\t506 \nFigure\t5:\tThe\trelationship\tbetween\tmotion,\tand\tdata\tquality\tand\tretention.\t\nBlue,\titalicised\taxes\tvalues\trepresent\tthe\tapproximate\toriginal\tvalues\tbefore\tscaling\tduring\t\nanalysis.\t(a)\tPredicted\tTRC\tSNR\ttrend\tby\tPercentage\tof\tMotion,\tgrouped\tby\tPSP\tThreshold.\t\n(b)\tPredicted\tChannels\tRetained\ttrend\tby\tPercentage\tof\tMotion,\tgrouped\tby\t PSP\tThreshold.\t\n(c)\t Predicted\t TRC\t SNR\t trend\t by\t Percentage\t of\t Motion,\t grouped\t by\t Age.\t (d)\t Predicted\t\nChannels\tRetained\ttrend\tby\tPercentage\tof\tMotion,\tgrouped\tby\tAge.\t\n28 \n24 \n20 \n0% \n 4% \n 8% \n 12% \n70 \n60\n50 \n40\n30\n20\n0% \n 4% \n 8% \n 12% \n30 \n25 \n20 \n15 \n0.0% \n 2.5% \n 5.0% \n 7.5% \n 10.0% \n 12.5% \n65 \n60\n55 \n50\n45\n40\n0.0% \n 2.5% \n 5.0% \n 7.5% \n 10.0% \n 12.5% \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n18\t\t\t507 \n(a)\t The\t interaction\t between\t and\t SCI\t Threshold,\t showing\t the\t\nmitigating\timpact\tof\thigh\tSCI\tvalues\tfor\tyounger\tparticipants.\t(b)\tThe\t\ninteraction\tbetween\tand\tPSP\tThreshold,\tin\twhich\ta\tpattern\twith\tage\t\nis\tharder\tto\tascertain.\t\nFigure\t6:\t Effect\t of\t the\t interaction\t between\t QT-NIRS\t thresholds\t\nand\tage\ton\tchannel\tretention.\t\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n19\t\n3.2.2 Predictors\tfor\tChannels\tRetained\t508 \nBoth\t SCI\t Threshold\t (𝛽 = −0.1381,\t 95%\t CI\t[−0.1401, −0.1361],\t 𝑆𝐸 = 0.0010)\t and\t509 \nPSP\tThreshold\t( 𝛽 = −0.4067,\t95%\tCI\t [−0.4087, −0.4048],\t𝑆𝐸 = 0.0009)\thad\tsmall\t510 \nand\t large\t negative\t effects\t on\t Channels\t Retained,\t respectively\t (see\tFigure\t 7).\t This\t511 \nindicates\tan\tincrease\tin\tthe\tnumber\tof\tchannels\tpruned\twhen\tusing\thigher\tthreshold\t512 \nvalues,\t particularly\t for\t the\t PSP\t Threshold.\t The\t interaction\t effect\t between\t SCI\t513 \nThreshold\t and\t PSP\t Threshold\t was\t small\t and\t negative\t ( 𝛽 = −0.8783,\t 95%\t CI\t514 \n[−0.8832, −0.8733],\t𝑆𝐸 = 0.0025),\twith\teach\tthreshold\talleviating\tthe\tnegative\teffect\t515 \nof\tthe\tother\tat\thigh\tparameter\tvalues.\t516 \n\t PoM\t had\t a\t significant,\t medium\t negative\t impact\t ( 𝛽 = −0.2013,\t 95%\t CI\t517 \n[−0.2030, −0.1995],\t 𝑆𝐸 = 0.0009),\t with\t higher\t amounts\t of\t motion\t associated\t with\t518 \ndecreased\t channel\t retention.\t At\t the\t two\t oldest\t ages\t (i.e.\t 18-\t and\t 24mo),\t greater\t519 \nproportions\tof\tmotion\tin\tthe\tdata\thad\ta\tless\tdrastic\tnegative\timpact\ton\tthe\tnumber\tof\t520 \nchannels\tpruned\t(see\tFigure\t5d)\twhich\twas\tcaptured\tby\tthe\tsmall\tpositive\tinteraction\t521 \neffect\t Age:PoM\t (𝛽 = 0.0610,\t 95%\t CI\t [0.0593, 0.6264],\t 𝑆𝐸 = 0.0009).\t As\t with\t TRC\t522 \nSNR,\t interaction\t effects\t of\t PoM\t with\t both\t SCI\t Threshold\t ( 𝛽 = −0.0034,\t 95%\t CI\t523 \n[−0.0050, −0.0019],\t 𝑆𝐸 = 0.0008)\t and\t PSP\t Threshold\t ( 𝛽 = −0.1962,\t 95%\t CI\t524 \n[−0.1977, −0.1945],\t 𝑆𝐸 = 0.0008)\t moderated\t this\t channel\t reduction\t for\t higher\t525 \npercentages\t of\t motion.\t Though\t both\t interactions\t were\t significant,\t PSP:PoM\t was\t of\t526 \nmedium\teffect\tsize\tand\talso\tacted\ton\ttwo\t(negative)\tmedium\tmain\teffects,\tleading\tto\t527 \na\tnear\tnegation\tof\tthe\tdetrimental\timpact\tof\tmotion\ton\tpredicted\tChannels\tRetained\t528 \nfor\tlow\tPSP\tThreshold\tvalues\t(see\tFigure\t5b).\t529 \n\t Task\t (𝛽 = −0.2038,\t 95%\t CI\t [−0.2069, −0.2007],\t 𝑆𝐸 = 0.0015)\t and\t Cohort\t530 \n(𝛽 = −0.8783,\t 95%\t CI\t [−0.8832, −0.8733],\t 𝑆𝐸 = 0.0025)\t had\t medium\t and\t large\t531 \neffects\t on\t Channels\t Retained,\t respectively,\t resulting\t in\t lower\t channel\t retention\t for\t532 \nthe\tSNS\ttask\tand\tUK\tcohort.\t\t533 \n\t534 \n\t535 \n\t536 \n\t537 \n\t538 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n20\t\n\t539 \n(a)\tThe\tinteraction\tbetween\tthresholds\tacting\ton\tTRC\tSNR,\tshowing\tan\tinconsistent,\t\nslight\tdecrease\tin\tTRC\tSNR\twhen\tcompared\tto\tthe\toverall\ttrend\tfor\thigh\tSCI-\tand\tPSP\t\nThreshold\t values.\t (b)\t The\t interaction\t between\t thresholds\t acting\t on\t Channel\t\nRetained,\tshowing\ta\tmore\tconsistent\ttrend\tacross\tthreshold\tlevels\tand\tthe\tdampening\t\nof\tthe\treduction\tin\tchannels\twhen\tboth\tSCI-\tand\tPSP\tThreshold\tvalues\tare\thigh.\t\nFigure\t7:\tThe\teffect\tof\tthe\tinteraction\tbetween\tSCI\tThreshold\tand\tPSP\tThreshold\t\non\tsignal\tquality\tand\tretention.\t\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n21\t\n4 Discussion\t540 \nChannel\tpruning\tin\tinfant\tfNIRS\tresearch\tis\t a\tcrucial\tstep\tin\tdata -analysis,\tyet\toften\t541 \nperformed\t with\t subjective \t parameters\t or\t adult\t derived\t threshold s.\t Moreover,\t542 \napproaches\t to\t channel\t pruning,\t including\t QT-NIRS\t and\t CV\t pruning,\tin\t infant\t fNIRS\t543 \ndata\t have\t often\t relied\t on\t adult -derived\t thresholds\t or\t single -domain\t measures,\t544 \nwithout\tsystematic\tevaluation\tof\ttheir\tsuitability\tfor\tinfant\tdata. \tThe\twork\tdescribed\t545 \nhere\t took\t a\t quantitative\t approach\t to\t directly\t compare\t pruning\t strategies\t and\t546 \nparameter\t settings\t for\t infant\t sparse\t array\t data.\t It\t was\t found\t that\t QT-NIRS,\t in\t its\t547 \nconsideration\tof\tboth\ttime-\tand\tfrequency-domain\tsignal\tquality\tmeasures,\tachieved\t548 \na\t better\t balance\t of\t data\t quality\t and\t retention\t than\t CV\t pruning .\t Additionally,\t549 \nparameter\t values\t for\t QT-NIRS\t were\t investigated\t alongside\t contextual\t factors,\t with\t550 \nlower\t values\t than \t those\t used\t for\t adults\t likely \t being\t preferable.\t These\t findings\t551 \nsupplement\twork\tin\tthe\twider\tliterature\taimed\tat\timproving\tinfant\tNIR\timaging\tdata\t552 \npipelines\tmore\tbroadly\t[6],\t[37],\t[44].\t553 \n4.1 QT-NIRS\tas\ta\tchannel\tpruning\tmethod\t554 \nQT-NIRS\tproduced\tinfant\tfNIRS\tdata\tof\tsignificantly \tgreater\tSNR\tthan\tpruning\tin\tall\t555 \nbut\tone\tage/task/cohort-specific\tcomparison\t(SNS\ttask\tin\tUK\tinfants\tat\t24mo),\twhilst\t556 \ncontrolling\tfor\tdata\tretention.\tHigher\tsignal\tquality\tremained\tstatistically\tsignificant\t557 \nwhen\t channels\t were\t pruned\t using\t SCI\t only,\t suggesting\t that\t evaluating\t the\t signal\t558 \nusing\t a\t method\t which\t accounts\t for\t a\t temporal\t data\t characteristic\t (correlation)\t is\t559 \nmore\t effective\t than\t assessment\t using\t time -independent\t measures.\t It\t may\t also\t560 \nindicate\t that\t evaluation\t of\t the\t signal\t in\t subsampled\t windows\t provides\t a\t more\t561 \naccurate\treflection\tof\tsignal\tquality\tthan\tholistic\tsignal\tassessment,\tas\twas\tthe\tcase\t562 \nwith\tCV\tpruning.\tThis\tdifference\tin\tsignal\tquality,\teven\twithout\tusing\tPSP\tis\tobserved\t563 \ndespite\tevidence\tsuggesting\tthat\tSCI\tmay\tbe\tbiased\tby\tlatent,\tundetected\tmotion\tin\t564 \nthe\tsignal\tin\tat\tleast\tsome\tparticipants\t(see\tEffects\tof\tMotion).\t565 \nUsing\tboth\tSCI-\tand\tPSP\tthresholds\tfurther\timproved\tsignal\tquality,\tresulting\t566 \nin\thigher\tmean\tTRC\tSNR\tvalues\tin\tall\tcases.\tThough\tnon -significant,\tthe\thigher\tTRC\t567 \nSNR\t values\t reinforce\t the\t advantage\t of\t incorporating\t both\t frequency -\t and\t time-568 \ndomain\tmetrics\tand\tprovide\tfurther\tjustification\tfor\tthe\tuse\tof\tQT-NIRS\twhen\tchannel\t569 \npruning\tof\tinfant\tfNIRS\tdata.\t570 \n4.2 Effects\tof\tQT-NIRS\tparameter\tchoice\t571 \nHigher\tSCI-\tand\tPSP\tthresholds\treduced\tdata\tretention\tby\tpruning\tmore\tchannels,\tbut\t572 \nimproved\tsignal\tquality\tin\tTRCs.\tThe\teffect\tof\tparameter\tchanges\twas\tsmaller\tthan\t573 \nanticipated,\tparticularly\ton\tsignal\tquality.\tPSP\tThreshold\thad\tmore\tinfluence\tthan\tSCI\t574 \nThreshold\ton\tboth\toutcomes\tbut\tparticularly\ton\tChannels\tRetained,\tillustrating\tthat\t575 \ncaution\tis\tneeded\twhen\taltering\tthis\tthreshold\tto\tavoid\tunnecessary\tdata\tloss.\t\t576 \n\t \tA\t positive\t SCI:PSP\t interaction\t reduced\t the\t channel\t pruning\t rate\t relative\t to\t577 \nthat\t expected\t when\t considering\t both\t thresholds\t in\t isolation\t (see\t Figure\t 7b),\t578 \nsuggesting\t that\t the\t potential\t cost\t to\t data\t retention\t from\t increasing\t one\t threshold\t579 \nvalue\t is\t mitigated\t when\t the\t other\t value\t is\t also\t high.\t This\t aligns\t with\t the\t use\t of\t580 \ncomplementary\tsignal\tmetrics\tin\tQT-NIRS\twhich\tmust\tboth\tbe\tof\tsufficient\tquality\tto\t581 \nretain\tchannels:\teach\tthreshold\twill\texclude\tchannels\twhich\tmay\tbe\tincluded\tby\tthe\t582 \nother,\twith\tthe\toverlap\tin\texcluded\tchannels\tincreasing\twith\tparameter\tvalues.\t583 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n22\t\n\t\t The\tslight\tnegative\tSCI:PSP\tinteraction\teffect\ton\tTRC\tSNR\tmay\tbe\tdriven\tby\tthe\t584 \nhighest\t values\t for\t both\t parameters\t (Figure\t 7a).\t Higher\t thresholds\t increased\t the\t585 \nlikelihood\t of\t pruning\t channels\t with\t neuronally -evoked\t cortical\t haemodynamic\t586 \nresponses\tand,\tconsequently,\thigh\tSNR;\ttheir\tremoval\tis\tlikely\tto\treduce\tthe\tmean\tTRC\t587 \nSNR\tmore\tthan\twould\tbe\texpected\tfor\tlower\tthreshold\tvalues.\tChange\tin\tTRC\tSNR\tis\t588 \nalso\tless\tuniform\tacross\tSCI -\tand\tPSP\tThreshold\tvalues\tthan\tfor\tChannels\tRetained.\t589 \nUnlike\t channel\t retention,\t which\t reflects\t the\t whole\t array,\t TRC\t SNR\t is\t localized\t to\t590 \nselect\t channels.\t Single\t channels\t will\t therefore\t likely\t contribute\t a\t much\t greater\t591 \nproportional\t value\t to\t the\t average\t TRC\t SNR\t and\t the\t consequence\t of\t pruning\t these\t592 \nchannels\t is\t likely\t to\t have\t a\t larger\t relative\t impact\t on\t the\t TRC\t SNR.\t Additionally,\t593 \npruning\ta\tchannel\tconsistently\treduces\tChannels\tRetained\twhereas\tits\timpact\ton\tTRC\t594 \nSNR\tdepends\ton\tthe\tpruned\tchannel’s\tSNR,\tfurther\tcontributing\tto\tthe\tinconsistency\t595 \nin\tchange.\t596 \n4.3 Effects\tof\tMotion\t597 \nSubstantial\t negative\t effects\t of\t PoM\t on\t both\t signal\t quality\t and\t channel\t retention\t598 \nsuggest\t that\t motion\t reduces\t signal\t quality\t in\t at\t least\t some\t channels\t even\t when\t599 \nartefacts\tare\tnot\tconsidered\tin\tthe\tpruning\tanalysis\tprocess,\tas\twas\tthe\tcase\tin\tthis\t600 \nwork.\tMotion\tartefacts\tmay\tcause\toptodes\tto\tmove,\tdislodge,\tor\tuncouple\tcompletely,\t601 \ncausing\t a\t poorer\t quality\t signal\t reflected\t in\t decreased\t TRC\t SNR.\t In\t turn,\t channels\t602 \naffected\t by\t motion\t may\t be\t pruned\t to\t a\t greater\t extent\t during\t QT-NIRS\t processing,\t603 \nleading\tto\tlower\tChannels\tRetained\tvalues.\t604 \nInteraction\t effects\t show\t PSP\t Threshold\t had\t the\t greater\t impact\t on\t motion-605 \naffected\t data\t of\t the\t two\t thresholds,\t improving\t signal\t quality\t but\t reducing\t channel\t606 \nretention,\t especially\t at\t high\t threshold\t values.\t In\t contrast,\t a\t comparatively\t modest\t607 \neffect\tof\tSCI\tThreshold\ton\tsignal\tquality\tand\tretention\twas\tfound.\tMotion\texhibited\ta\t608 \nsignificant\tadverse\teffect\ton\tAverage\tPSP,\tlikely\tdue\tto\toptode\tdisplacement\tsevere\t609 \nenough\tto\tdisrupt\tcoupling,\tand\ta\tpositive\teffect\tof\tmotion\tincidences\ton\tAverage\tSCI,\t610 \nsuggesting\t that\t SCI\t measures\t may\t be\t capturing\t correlation\t induced\t by\t latent,\t611 \nundetected\t motion\t artefacts\t in\t the\t data.\t Future\t work\t may\t examine\t the\t effect\t of\t612 \ndifferent\t motion\t detection\t parameters,\t or\t alternative\t motion\t correction\t methods\t613 \naltogether\tsuch\tas\tthe\tSobel\tfilter\t[45],\tacceptance\trate\tadaptive\talgorithm\t[46],\tglobal\t614 \nvariance\tof\ttemporal\tderivatives\t[47],\t[48]\tor\tentropy-based\tmethods\t[49].\t615 \n\t The\t potential\t (interaction)\t effect\t of\t high\t PSP\t Threshold\t values\t on\t channel\t616 \nretention,\t especially\t for\t channels\t with\t a\t greater\t proportion\t of\t motion,\t warrants\t617 \ncaution\t for\t users\t when\t looking\t to\t employ\t high\t PSP\t Threshold\t values.\t This\t is\t618 \nparticularly\t true\t given\t the\t relatively\t small\t beneficial\t impact\t on\t signal\t quality\t of\t619 \nincreasing\t the\t PSP\t threshold,\t indicated\t by\t \t its\t main\t effect\t on\t TRC\t SNR.\t This\t is\t620 \nconsistent\t with\t the\t argument\t for\t using\t lower\t values\t discussed\t in\t the\t621 \nRecommendations\tsection\tof\tthe\tDiscussion.\tConversely,\tmotion\tincidences\thave\tthe\t622 \nlargest\tnegative\timpact\ton\tTRC\tSNR;\tlow\tPSP\tthresholds\tmay\texacerbate\tthis\teffect\tif\t623 \nmotion\tis\tnot\tappropriately\taddressed.\t624 \n4.4 Effects\tof\tage\t625 \nAge\thad\tsmall\teffects\ton\tTRC\tSNR\tand\tChannels\tRetained,\tsuggesting\tlimited\tchanges\t626 \nin\tsignal\tquality\tand\tretention\t in\tinfants\tbetween\tthe\tage\tof \t5\tand\t24 mo.\tReflecting\t627 \nthis,\t associated\t trends\t were\t less\t commonly\t observed\t with\t age\t than\t for\t other\t628 \npredictors;\thowever,\tit\twas\tnotable\tthat\tchannel\tloss\twas\tless\tsevere\twhen\tincreasing\t629 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n23\t\nthe\tSCI\tThreshold\tin\tyounger\tparticipants\t(Figure\t6).\tAdditionally,\twhile\thigher\tsignal\t630 \nPoM\t is\t associated\t with\t reduced\t channel\t retention,\t older\t infants\t retained\t more\t631 \nchannels\tthan\tyounger\tones\tfor\tdata\twith\thigh\tmotion\tprevalence\t( Figure\t5d).\tThis\t632 \nmay\tbe\tdue\tto\tyounger\tinfants\tmanually\ttouching\tor\tgrabbing\tthe\tcap\tmore,\tcausing\t633 \nmore\t severe\t artefacts\t that\t permanently\t displace\t optodes\t and\t affect\t subsequent\t634 \ncoupling,\t an\t assertion\t supported\t by\t post-hoc\t data\t examination\t which\t suggested\t a\t635 \ndecrease\tin\tmotion\tseverity\twith\tage\t(see\tSupplementary\tMaterials,\tFigure\tS\t7)\t636 \nInteractions\twith\tage\tfor\tTRC\tSNR\tshowed\tno\tmeaningful\tpatterns.\tThis\tmay\t637 \nreflect\t a\t lack\t of\t consistent\t changes\t within\tthe\tage\t range\t sampled\t here.\t It\t may\t also\t638 \nreflect\tthe\tmultifaceted\tmix\tof\tconcepts\twhich\t‘age’\trepresents:\tthe\tinterplay\tof\tthe\t639 \nphysiological\tand\tbehavioural\tchanges\twith\tage\tmay\tbe\ttoo\tcomplex\tfor\ta\tsimple\tfixed\t640 \neffect\t to\t capture.\t Interactions\t between\t age\t and\t other\t model\t terms\t (e.g.\t CSE,\t CL)\t641 \ncaused\t convergence\t issues\t and\t were\t omitted.\t Future\t research\t may\t priorit ize\t642 \ninvestigating\tage-related\tchange\tand\tassociated\tfactors\taffecting\tsignal\tquality\t–\tsuch\t643 \nas\thair\tcharacteristics\t[23],\thair\tstyle\t[31],\tor\thair\ttype\tchanges\twith\tage\t[50].\t644 \n4.5 Other\tpredictors\t\t645 \n4.5.1 Cohort\t646 \nCohort\t had\t substantial\t effects\t on\t signal\t quality\t and\t retention,\t likely\t due\t to\t factors\t647 \nsuch\t as\t testing\t environment,\t tester\t experience,\tparent\t and\t infant\t behaviours,\tand\t648 \nsample\t size.\t Since\t cohort\t was\t not\t a\t primary\t predictor,\t its\t main\t effect \t was\t not\t649 \ninvestigated\t in\t depth\tand\t interaction\t terms\t were\t not\t included.\t Nevertheless,\t data\t650 \nfrom\tthe\tGambian\tcohort \t–\twhose\tskin\tand\thair\tcharacteristics\t have\tbeen\tfound \tto\t651 \npose\t challenges\t for\t fNIRS\t signal\t quality\t [23]\t –\t exhibited\t better\t signal\t quality\t and\t652 \nretention,\t to\t the\t extent\t that\t data\t exclusion\t was\t far\t lower\t for\t Gambian\t infants\tin\t653 \ngeneral\t and\t almost\t non-existent\t at\t the\t infant\t level\t (see\t Figure\t 2).\t There\t may\t be\t654 \nseveral\treasons\tfor\tthis.\tFirstly,\tUK\tinfants\tgenerally\thad\tfiner\thair\twhich\twas\tlonger\t655 \nat\tlater\tassessment\tages,\tpossibly\tcausing\tcap\tslippage\tand\tpoorer\tsubsequent\tsignal\t656 \nquality,\t as\t these\t values\t appear\t to\t decrease\t with\t age.\t Lower\t motion\t incidence\t in\t657 \nGambian\tinfants\tmay\talso\tplay\ta\trole,\tpossibly\tdue\tto\ta\trelative\tlack\tof\tfamiliarity\twith\t658 \ndigital\tscreens\tin\tdaily\tlife\tleading\tto\tan\tincreased\tfocus\ton\ta\tnovel,\tunfamiliar\tobject.\t659 \nWhile\t physical\t characteristics\t of\t participants\t undoubtedly\t affect\t the\t fNIRS\t signal,\t660 \ncohort\tdifferences\tin\tthis\tstudy\temphas ize\tthe\tneed\tto\tconsider\tother\tfactors\twhich\t661 \naffect\tsignal\tquality\tduring\tdata\tcollection.\t\t662 \n4.5.2 Weak\tor\tsaturated\tchannel\tsignals\t\t663 \nA\tsmall\tand\tsignificant\tnegative\teffect\tof\tCSE\ton\tchannel\tretention\twas\tanticipated,\t664 \nconsidering\t it\t is\t itself\t a\t measure\t of\t channel\t removal.\t CSE\t was\t also\t significantly\t665 \nnegatively\t associated\t with\t data\t quality,\t however,\t suggesting\t poor\t coupling\t in\t666 \nchannels\twith\textremely\tlow\tquality\tsignal\tcould\tbe\taffected\tby\t –\tor\tcausing\t–\tsignal\t667 \nquality\t issues\t elsewhere\t in\t the\t array.\t It\t may\t be\t of\t interest\t to\t investigate\t whether\t668 \nsignal\t quality\t was\t worse\t in\t channels\t located\t closely\t to\t the\t CSE\t channels\t in\t future\t669 \nwork,\tas\thas\tbeen\tthe\tfocus\tof\tprior\twork\tinto\tmotion\tartefact\tdetection\t[49].\t670 \n4.6 Strengths\tand\tlimitations\t671 \nA\tkey\tstrength\tof\tthis\twork\tis\tthe\tdataset\tused.\tData\tencompassed\tfive\t testing\ttime\t672 \npoints\tacross\tthe\tage\trange \t5-\tto\t24mo,\tallowing\tassessment\tof\tQT -NIRS\tand\tits\tkey\t673 \nparameters\tfor\tinfants\twhilst\taccounting\tfor\tage -related\tstructural\tchanges\tin\tskull\t674 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n24\t\nthickness,\tcardiac\tsignal\tproperties,\tand\tsurface\tvasculature\t[51].\tData\tfrom\ttwo\tsites\t675 \nwere\talso\tassessed,\tone\tof\twhich\twas\ta\trural\tsetting\tin\ta\tsub-Saharan\tAfrican\tcountry,\t676 \naddressing\ta\tcommon\tbias\twhich\toften\texists\tin\tinfant\tneuroimaging\tstudies\twhen\t677 \nparticipant\trecruitment\tis\tlimited\tto\tpredominantly\tWhite\tinfants\tfrom\thigh -income\t678 \ncontexts\t[52].\tAs\tsuch,\tthe\tfindings\tare\tlikely\tto\tbe\tmore\trobust\tand\tgeneralisable\tthan\t679 \nthose\tderived\tfrom\tsmaller\tor\t single-site\tsamples.\tThe\tinclusion\tof\tlongitudinal\tdata\t680 \nallows\t for\t the\t assessment\t of\t within-\t and\t between-participant\t changes\t over\t time,\t681 \nproviding\tricher\tinsights\tinto\tthe\teffect\tof\tprocessing\tmethods\t\tthan\twould\tbe\tpossible\t682 \nfrom\tcross-sectional\tanalyses\talone.\t683 \n\t Another\tstrength\tis\tthe\tMLM\tapproach\twhich\taccounted\tfor\tthe\thierarchical\t684 \nvariance\tstructure\tof\tlongitudinal\tdata\tgrouped\tby\ttask\tand\tcohort,\treducing\tbias\tand\t685 \nenabling\t random\t intercepts\t and\t slopes\t during\t regression\t analysis\t to\t capture\t686 \nindividual\t differences\t [39].\t This\t approach\t also\t handled\t missing\t data\t introduced\t687 \nthrough\tprior\tchannel\tremoval\t(CSE\tchannels),\tmissed\tvisits\tor\tincomplete\ttesting,\t688 \nwhich\twould\tnot\thave\tbeen\tpossible\twith\tmany\tother\tcommon\tanalysis\tprocedures\t689 \n[53].\tDomain\tknowledge\tand\tdata -driven\tinsight\twere\tcombined\tto\tpriorit ize\tmodel\t690 \npredictors\t and\t validate\t model\t assumptions\t and\t additionally\t fitted\t and\t assessed\t691 \nsubsidiary\t MLMs\t to\t ensure\t the\t final\t model\t (Equation\t 2)\t was\t as\t comprehensive\t as\t692 \npossible.\t693 \nQT-NIRS\t was\t compared\t with\t CV\t pruning,\t using\t a\t maximum\t wavelength\t CV\t694 \ndifference\t of\t 0.2,\t as\t previously\t applied\t with\t data\t from\t this\t study\t[14].\t Significant\t695 \ndifferences\tin\tsignal\tquality\tbetween\tQT -NIRS\tand\tCV\tpruning\twere\tfound\tin\tall\tbut\t696 \none\tage/task/cohort\tcombination.\tHowever,\tanother\tcommon\tCV\tpruning\tapproach\t697 \nuses\ta\tsingle-channel\tthreshold\tinstead,\tpruning\tboth\tchannel\twavelength\tsignals\tat\t698 \nleast\tone\tof\tthem\texceeds\tit\t [12],\t[32],\t[54] .\tFuture\twork\tcould\tinvestigate\twhether\t699 \nthe\tsignificant\tdifferences\tfound\there\tpersist\twhen\tusing\tthis\talternative\tthresholding\t700 \nmethod.\t\t701 \nThe\t focus\t of\t this\t work\t was\t on\t optimizing\t thresholds\t of\t the\t two\t QT-NIRS\t702 \nparameters\t which\t are\t most\t pivotal\t to\t performance,\t for\t which\t only\t reported\t703 \nrecommended\tvalues\tfor\tadult\tparticipants\twere\tfound\tin\tthe\tliterature.\tFuture\twork\t704 \nmay\tfocus\ton\tother\tparameters,\tsuch\tas\tchanging\tthe\tquality\tthreshold\t(q_threshold),\t705 \nbalancing\t channel\t quality\t discernment\t with\t the\t risk\t of\t losing\t nuance\t in\t signal\t706 \ncharacteristics.\tFuture\tstudies\tcould\talso\tinvestigate\tthe\timpact\tof\taltering\twindow\t707 \nsize\tor\toverlap:\tthe\tformer\twill\tlikely\tbalance\tmeasurement\taccuracy\twithin\twindows\t708 \nagainst\toverall\ttemporal\tsensitivity;\tthe\tlatter\tmay\tprovide\tmore\twindow\ttemporal\t709 \nsensitivity\t at\t the\t expense\t of\t computational\t efficiency.\t Given\t the\t dominance\t of\t PSP\t710 \nThreshold\tchange\ton\toutcomes\treported\there,\tit\tmay\talso\tbe\tof\tinterest\tto\texplore\t711 \npruning\tusing\tonly\tthe\tPSP\tthreshold\t[55].\t712 \nThis\t work\t must\t also\t be\t placed\t in\t the\t context\t of\t the\t increasing\t impact\t of\t713 \nmachine\tlearning\ton\tinfant\tNIR\timaging\tdata\tprocessing,\twith\tfuture\twork\tin\tthe\tfield\t714 \nlikely\t to\t assess\t the\t efficacy\t and\t generalisability\t of\t such\t methods.\t Deep\t learning\t715 \napproaches\t are\t frequently\t being\t added\t to\t the\t literature:\t a\t machine\t learning\t based\t716 \ndetector\thas\tbeen\tdeveloped\tto\tidentify\t‘bad’\tchannels\tto\tbe\tpruned,\tfor\texample\t[55].\t717 \nThis\t approach\t was\t shown\t to\t be\t more\t adaptive,\t interpretable,\t and\t effective\t across\t718 \ndiverse\t noise\t types\t than\t QT-NIRS\t and\t other\t methods,\t so\t future\t work\t may\t seek\t to\t719 \nassess\tthe\tefficacy\tof\tthis\tindependently.\tIn\tmitigation,\tcare\tshould\tbe\ttaken\tto\tensure\t720 \nlimitations\t common\t to\t many\t deep\t learning\t approaches\t (e.g.\t overfitting)\t are\t721 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n25\t\naddressed\t through\t appropriate\t validation\t strategies,\t including\t the\t use\t of\t722 \nindependent\ttest\tsets\tand\tcross-validation.\t723 \nModel\tdesign\twas\tlimited\tby\tconvergence\tissues,\tparticularly\twhen\tincluding\t724 \nrandom\t slopes\t and\t intercepts.\t The\t most\t likely\t and\t important\t sources\t of\t variation\t725 \nwere\tprioritized,\thowever\tit\twas\tnot\tpossible\tto\tfully\tinclude\tall\trelevant\tinteraction\t726 \nterms\t or\t random\t effects.\t Additionally,\t linear\t terms\t were\t used\t in\t the\t models\t to\t727 \nsimplify\t interpretability.\t Future\t work\t may\t incorporate\t non -linear\t terms,\t as\t an\t728 \nalternative\t to\t the\t bootstrapping\t approach\t for\t dealing\t with\t non-normal\t residuals.\t729 \nOther\t alternative\t approaches\t could\t include\t sensitivity\t analysis\t of\t the\t influential\t730 \npoints\t and\t outliers;\t variance\t modelling\t for\t specific\t predictors;\t and\t utilisation\t of\t731 \nrobust\tstandard\terrors.\t732 \n\t While\tage\tis\tlikely\tto\treflect\tmore\tthan\tbehaviour\tand\tmotion,\tthe\tlongitudinal\t733 \nstudy\tdesign\tmay\talso\tresult\tin\tparticipant\tfamiliarity\twith\tthe\ttesting\tprocedures\tand\t734 \nthe\t paradigms,\t in\t turn\t possibly\t affecting\t behaviour,\t attention,\t stress\t levels,\t735 \nengagement,\tand\t–\tconsequently\t–\tthe\trecorded\tdata.\tThus,\tcaution\tshould\tbe\ttaken\t736 \nwhen\texamining\tthe\teffects\tof\tthe\t age\tmodel\tterm\tand \tits\tinteractions,\trecognising\t737 \nthis\tas\ta\tpotential\tconfounding\tfactor\tin\tthis\twork.\t\t738 \nFurther\t evaluation\t of\t QT-NIRS\t as\t a\t channel\t pruning\t method\t for\t infant\t fNIRS\t739 \ndata\tis\tstill\tnecessary\tto\taddress\tthe\tlimitations\tof\tthis\twork,\textend\tit\tto\thigh-density\t740 \nsystems,\t and\t compare\t it\t to\t alternative\t approaches\t including\t those\t incorporating\t741 \nmachine\tlearning\t[55].\t742 \n4.7 Recommendations\t743 \nBased\ton\tthis\twork,\tthe\tfollowing\tguidelines\tfor\tchannel\tpruning\tinfant\tfNIRS\tdata\tare\t744 \nrecommended:\t745 \n(1) QT-NIRS\tas\ta\tpruning\tmethod\tis\tpreferable\tto\tpruning\tusing\tCV\t(when\tusing\ta\t746 \nminimum\tthreshold\tdifference\tbetween\twavelengths)\t747 \n(2) Users\tshould\tconduct\tchannel\tpruning\ton\tmotion-free\tdata,\twith\tconsiderable\t748 \nemphasis\tduring\tinitial\tpreprocessing\tplaced\ton\tadequately\tidentifying\tmotion\t749 \nartefacts\t\t750 \n(3) Tuning\tof\tthe\tPSP\tThreshold\tshould\tbe\tpriorit ized\tover\tthe\tSCI\tthreshold\tin\t751 \ndata\t\t752 \n(4) To\t provide\t a\t good\t trade -off\t between\t data\t quality\t and\t retention,\t lower\t753 \nthresholds\t(with\tminima\tof\tpsp_threshold\t≈\t0.04–0.05\tand\tsci_threshold\t≈\t0.6)\t754 \ncan\tbe\tused \tfor\tinfant\tdata\tthan\tth ose\trecommended\tfor\tadults,\tespecially\tin\t755 \nolder\tinfants\twith\tfine/slippery\thair,\tsince\t(amongst\tother\tfactors):\t756 \n(i) There\tare\ta\tlarge\tproportion\tof\tdata\tshowing\tSCI-\tand\tPSP\tvalues\tlower\t757 \nthan\tthe\tadult\tthresholds\tof\t0.8\tand\t0.1,\trespectively\t(Figure\t2)\t758 \n(ii) The\trisk\tof\tremoving\tdata\twith\thigher\tthresholds\tis\tlikely\tgreater\tthan\t759 \nthe\tpotential\tgain\tin\timproved\tsignal\tquality\t\t760 \n(iii) There\t is\t a\tplateau\t in\t both\t average\t TRC\t SNR\t and\t Channels\t Retained\t761 \nvalues\twhen\tusing\tlower\tthreshold\tvalues\tthan\tthe\tadvised\tminima\t(see\t762 \nSupplementary\tMaterials,\tFigure\tS\t6)\t763 \nA\t pictorial\tguide\t to\t the\t effect\t of\t the\t most\t common\t predictors\t on\t data\t quality\t and\t764 \nretention\tis\talso\tincluded\tin\tFigure\t8.\t765 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n26\t\nIt\tis\tstill\tstrongly\trecommended\t that\tfNIRS\tusers\t \ttake\tthe\t appropriate\ttime\tto\t766 \nunderstand\tthe\tdataset\tbeing\tanalysed\tsince\tno\tapproach\tcan\tpossibly\tbe\tuniversal.\t767 \nTo\t aid\t final\t parameter\t selection,\t a\t tool\t which\t other\t users\t may\t find\t useful\t to\t guide\t768 \nparameter\t selection\t was\t developed .\t Alongside\t code\t used\t during\t processing\t and\t769 \nanalysis\t reported\t in\t this\t study\t can\t be\t found\t at:\t https://github.com/sam-770 \nbeaton/pruningComparisons/.\t The\t tool \t is\t designed\t to\t examine\t the\t effects\t of\t771 \nparameter\t threshold\t changes\t within\t a\t group\t (e.g.\t age),\t by\t establishing\t a\t trend\t772 \ncapturing\t the\t trade-off\t between\t data\t quality\t and\t retention,\t then\t assessing\t which\t773 \nspecific\tSCI-\tand\tPSP\tThreshold\tparameter\tcombinations\tperform\tbest\tin\trelation\tto\t774 \nthis\ttrend.\t\t775 \n5 Conclusion\t776 \nRecent\tadvances\tin\tfNIRS\tchannel\tpruning\tapproaches\tshow\tpromise\tfor\timproving\t777 \nthe\t accuracy\t of\t preprocessing\t by\t evaluating\t both\t the\t time\t and\t frequency\t domains.\t778 \nThe\t work\t described\t here\t compared\t QT-NIRS\t with\t an\t established\t pruning\t method\t779 \nutilising\tCV,\tacross\t5\tinfant\tages,\ttwo\tparadigms,\tand\ttwo\tsites.\tIt\twas\tfound\tthat\tQT-780 \nNIRS\tprovides\tdata\twith\tgreater\tsignal\tquality\twhen\tcontrolling\tfor\tdata\tretention.\t781 \nFigure\t8:\tGuide\tto\tthe\teffects\tof\tthe\tparameters\tand\tdata\tcharacteristics\ton\tdata\t\nquality\tand\tretention.\t\nFirst\tfive\trows\trepresent\tpositive\t(green)\tor\tnegative(red)\tassociations\twith\tincreasing\t\nnumerical\tparameters\t(SCI\tThreshold\tand\tPSP\tThreshold)\tor\tdata\tcharacteristics\t(Age,\t\nPoM,\tCSE).\tAll\tpositive\teffects\tare\tsmall\tin\tsize.\tBottom\ttwo\trows\trepresent\tcateg orical\t\nvariables,\t with\t greyscale\t shading\t indicative\t of\t the\t impact\t on\t outcome\t changing\t the\t\ncategorical\tvariable\tmay\thave.\t\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n27\t\nThe\t consequences\t of\t different\t parameter\t choices\t for\t QT -NIRS\t were\t also\t782 \ndemonstrated,\thighlighting\tthe\timportance\tof\tthe\tPSP\tThreshold,\tplus\tthe\tinfluence\tof\t783 \nmotion.\t Evidence-based\t recommendations\t for\t QT -NIRS\t pruning,\t and\t parameter\t784 \nchoice\tfor\tinfant\tdata\twith\tdifferent\tcharacteristics,\tare\tprovided.\t\t785 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n28\t\nDisclosures\t786 \nThe\tauthors\thave\tno\tconflict\tof\tinterest\tto\tdeclare.\t787 \n\t788 \nCode\tand\tData\tAvailability\t\t789 \nAll\t data\t used\t in\t the\t analyses\t presented\t can\t be\t made\t available\t following\t relevant\t790 \napprovals.\tThe\tcode\tused\tto\tconduct\tthe\tanalyses\tand\tgenerate\tthe\tfigures\tpresented\t791 \nin\tthis\tpaper\tare\tavailable\tat\t https://github.com/sam-beaton/pruningComparisons.\t792 \nCode\trelies\ton\tthe\tfollowing\topen\tsource\tR\tlibraries:\tarules\t[56],\tbroom\t[57],\tcar\t[58],\t793 \ndata\t[59],\tdoParallel\t[60],\tdplyr\t[61],\teffectsize\t[62],\teffects\t[58],\te1071\t[63],\tforeach\t794 \n[64],\tggplot2\t[65],\tggpubr\t[66],\tgridExtra\t[67],\tlme4\t[41],\tlmerTest\t[68],\tMASS\t[69],\t795 \nmatrix\t[70],\tMuMIn\t[71],\tScales\t[72],\ttidyr\t[73],\tviridis\t[74].\t796 \n\t797 \nAcknowledgments\t798 \nWe\t would\t like\t to\t place\t on\t record\t our\t thanks\t to\t the\t children,\t mothers\t and\t wider\t799 \nfamilies\t who\t took\t part\t in\t this\t study.\t In\t addition,\t we\t would\t like\t to\t thank\t the\t data\t800 \ncollection\tteams\tin\tboth\tThe\tGambia\tand\tthe\tUK .\tFinally,\twe\tthank\tLuca\tPollonini\tfor\t801 \nhis\tcontribution\tto\tearly\tdiscussions\tconcerning\tthe\tframing\tof\tthe\twork,\tand\tJohann\t802 \nBenerradi,\tfor\tproviding\ttask-relevant\tchannels\tfor\tthe\tsocial/non-social\ttask\tprior\tto\t803 \ntheir\tpublication.\t804 \n\t805 \nFunding\t806 \nSam\tBeaton,\tEbrima\tMbye,\tSamantha\tMcCann\t(to\tJuly\t2024)\tand\tSophie\tMoore\tare\t807 \nsupported\tby\ta\tWellcome\tTrust\tSenior\tResearch\tFellowship\t(220225/Z/20/Z)\theld\t808 \nby\t Sophie\t Moore.\t The\t BRIGHT\t Study\t was\t funded\t by\t the\t Gates\t Foundation\t809 \n(OPP1127625)\t and\t core\t funding\t MC -A760-5QX00\t to\t the\t International\t Nutrition\t810 \nGroup\tby\tthe\tMedical\tResearch\tCouncil\tUK\tand\tthe\tUK\tDepartment\tfor\tInternational\t811 \nDevelopment\t(DfID)\tunder\tthe\tMRC/DfID\tConcordat\tagreement.\tFurther\tsupport\twas\t812 \nprovided\tthrough\ta\tUKRI\tFuture\tLeaders\tFellowship\t(MR/S018425/1)\theld\tby\tSarah\t813 \nLloyd-Fox.\t Borja\t Blanco\t was\t supported\t by\t a\t Medical\t Research\t Council\t Programme\t814 \nGrant\t(MR/T003057/1)\tand\ta\tUKRI\tFuture\tLeaders\tfellowship\t(MR/S018425/1). 815 \n\t816 \nAuthor\tContributions\t(CRediT\ttaxonomy)\t817 \nSB:\t conceptualisation,\t formal\t analysis,\t methodology,\t software,\t validation,\t818 \nvisualisation,\t writing\t –\t original\t draft,\t writing\t –\t review\t and\t editing;\t BB:\t819 \nconceptualisation,\t methodology,\t supervision,\t writing\t –\t review\t and\t editing;\t CB:\t820 \nconceptualisation,\t methodology,\t writing\t –\t review\t and\t editing;\t CE:\t funding\t821 \nacquisition,\t project\t administration,\t resources,\t writing\t –\t review\t and\t editing;\t SLF:\t822 \nfunding\tacquisition,\tproject\tadministration,\tresources,\twriting\t –\treview\tand\tediting;\t823 \nEM:\t data\t curation,\t investigation,\t project\t administration;\t SMc:\t data\t curation,\t824 \ninvestigation,\t project\t administration,\t supervision;\t ABR:\t conceptualisation,\t data\t825 \ncuration,\t supervision,\t visualisation,\t writing\t –\t review\t and\t editing;\t SM:\t funding\t826 \nacquisition,\t project\t administration,\t resources,\t supervision,\t writing\t –\t review\t and\t827 \nediting.\t828 \n\t829 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n29\t\nBiographies\t\t830 \nSamuel\t Beaton\t is\t a\t doctoral\t researcher\t and\t research\t assistant\t at\t King’s\t College\t831 \nLondon,\t developing\t analytical\t pipelines\t for\t functional\t near-infrared\t spectroscopy\t832 \n(fNIRS)\tand\thigh-density\tdiffuse\toptical\ttomography\t(HD-DOT)\tdata.\t833 \nChiara\t Bulgarelli \t is\t a\t Senior\t Lecturer\t at\t the\t Centre\t for\t Brain\t and\t Cognitive\t834 \nDevelopment\tat\tBirkbeck,\tUniversity\tof\tLondon.\tShe\thas\tover\ta\tdecade\tof\texperience\t835 \nusing\t fNIRS\t with\t infants\t and\t toddlers\t to\t study\t the\t neural\t mechanisms\t underlying\t836 \nsocial\t interactions\t and\t the\t development\t of\t brain\t connectivity.\t Recently,\t she\t837 \npioneered\t the\t use\t of\t fNIRS\t in\t non-traditional\t lab\t settings,\t such\t as\t with\t toddlers\t838 \ninteracting\tin\ta\tvirtual\treality\tenvironment.\t839 \nBorja\tBlanco\tis\ta\tpostdoctoral\tresearch\tassociate\tin\tthe\tDepartment\tof\tPsychology\tat\t840 \nthe\t University\t of\t Cambridge.\t His\t work\t focuses\t on\t developing\t data\t processing\t and\t841 \nanalysis\tmethods\tfor\toptical\tneuroimaging\tin\tdevelopmental\tpopulations.\tHe\tapplies\t842 \nthese\tmethods\tto\tinvestigate\tinfant\tfunctional\tbrain\tdevelopment\tand\tthe\tcontextual\t843 \nfactors\tthat\tinfluence\tit.\t844 \nClare\tElwell\tis\ta\tprofessor\tof\tmedical\tphysics\tleading\tprojects\tto\tinvestigate\tbrain\t845 \nfunction\t using\t functional\t near\t infrared\t spectroscopy\t in\t high\t and\t low\t resource\t846 \nsettings\tin\tadult\tand\tinfants.\t847 \nSarah\tLloyd-Fox\tis\ta\tPrincipal\tResearch\tAssociate\tin\tthe\tDepartment\tof\tPsychology,\t848 \nUniversity\t of\t Cambridge.\t She\t leads\t several\t multi-disciplinary\t projects\t focusing\t on\t849 \ndevelopmental\t trajectories\t of\t early\t cognitive\t and\t brain\t development\t during\t850 \npregnancy,\tinfancy\tand\tearly\tchildhood.\tHer\tresearch\tfocuses\ton\tunderstanding\thow\t851 \nfamily\tand\tenvironmental\tcontext\t -\ti.e.\tcontextual\tfactors\tsuch\tas\tpoverty\tassociated\t852 \nchallenges\tand\tenriched\tmultigenerational\tfamily\tsupport\t\t-\tshape\tearly\tlife.\t853 \nEbrima\t Mbye\t is\t a\tField\t Coordinator\t at\tMRCG@LSHTM,\t formerly\t employed\t on\t the\t854 \nBRIGHT\tProject\tand\tcurrently\tengaged\ton\tthe\tINDiGO\tTrial\tworking\tunder\tProfessor\t855 \nSophie\tMoore.\t856 \nSamantha\tMcCann\tis\ta\tPublic\tHealth\tRegistrar\tand\tformerly\t Postdoctoral\tResearch\t857 \nAssociate\t in\t the\t Department\t of\t Women\t and\t Children’s\t Health\t at\t King’s\t College\t858 \nLondon.\t Her\tmain\t research\t interest\t is\t supporting\t early\t child\t development,\t with\t a\t859 \nstrong\t focus\t on\t the\t impact\t of\t undernutrition\t in\t infancy\t on\t long -term\t860 \nneurodevelopmental\toutcomes.\t861 \nAnna\t Blasi\t is\t a\t Postdoctoral\t Research\t Fellow\t at\t UCL.\t Her\t research\t interests\t are\t862 \ncentered\ton\tfunctional\taspects\tof\thuman\tphysiology.\tHer\tresearch\tcareer\tstarted\twith\t863 \nmodels\tof\tthe\tcardiovascular\tsystem\tand\tthe\teffects\tof\tdisease.\tThrough\ther\twork\tat\t864 \nUCL,\t KCL,\t and\t Birkbeck,\t her\t research\t interests\t have\t shifted\t toward\t the\t use\t of\t865 \nfunctional\t imaging\t (fNIRS,\t fMRI)\t to\t study\t brain\t function\t and\t neurocognitive\t866 \ndevelopment\tin\tearly\tinfancy.\t867 \nSophie\tMoore\tis\tProfessor\tof\tGlobal\tWomen\tand\tChildren's\tHealth\tin\tthe\tDepartment\t868 \nof\tWomen\t&\tChildren's\tHealth \tat\tKing’s\tCollege\tLondon \tand\tan\tHonorary\tAssociate\t869 \nProfessor\t at\t the\t London\t School\t of\t Hygiene\t and\t Tropical\t Medicine\t (LSHTM).\t Her\t870 \nresearch\t focuses\t on\t the\t nutritional\t regulation\t of\t ‘healthy’\t fetal\t and\t infant\t growth,\t871 \nincorporating\t infant\t immune\t and\t brain\t development\t as\t outcomes,\t and\t on\t the\t872 \nmechanisms\tthrough\twhich\tmaternal,\tinfant\tand\tchildhood\tnutrition\tmay\tinfluence\t873 \ndevelopment\tand\tlater\thealth.\t874 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n30\t\nReferences\t875 \n\t[1]\t L.\tL.\tEmberson,\tB.\tD.\tZinszer,\tR.\tD.\tS.\tRaizada,\tand\tR.\tN.\tAslin,\t‘Decoding\tthe\tinfant\t876 \nmind:\tMultivariate\tpattern\tanalysis\t(MVPA)\tusing\tfNIRS’,\t PLOS\tONE,\tvol.\t12,\tno.\t877 \n4,\tp.\te0172500,\tApr.\t2017,\tdoi:\t10.1371/journal.pone.0172500.\t878 \n[2]\t J.\tGervain,\tY.\tMinagawa,\tL.\tEmberson,\tand\tS.\tLloyd -Fox,\t‘Using\tfunctional\tnear -879 \ninfrared\tspectroscopy\tto\tstudy\tthe\tearly\tdeveloping\tbrain:\tfuture\tdirections\tand\t880 \nnew\t challenges’,\t Neurophotonics,\t vol.\t 10,\t no.\t 2,\t p.\t 023519,\t Apr.\t 2023,\t doi:\t881 \n10.1117/1.NPh.10.2.023519.\t882 \n[3]\t D.\t A.\t Boas,\t K.\t Chen,\t D.\t Grebert,\t and\t M.\t A.\t Franceschini,\t ‘Improving\t the\t diffuse\t883 \noptical\t imaging\t spatial\t resolution\t of\t the\t cerebral\t hemodynamic\t response\t to\t884 \nbrain\t activation\t in\t humans’,\tOpt.\t Lett.,\t vol.\t 29,\t no.\t 13,\t p.\t 1506,\t Jul.\t 2004,\t doi:\t885 \n10.1364/OL.29.001506.\t886 \n[4]\t H.\t Santosa,\t X.\t Zhai,\t F.\t Fishburn,\t P.\t J.\t Sparto,\t and\t T.\t J.\t Huppert,\t ‘Quantitative\t887 \ncomparison\tof\tcorrection\ttechniques\tfor\tremoving\tsystemic\tphysiological\tsignal\t888 \nin\tfunctional\tnear-infrared\tspectroscopy\tstudies’,\tNeurophotonics,\tvol.\t7,\tno.\t3,\tp.\t889 \n035009,\tSep.\t2020,\tdoi:\t10.1117/1.NPh.7.3.035009.\t890 \n[5]\t M.\tA.\tYücel\t et\tal.,\t‘Best\tpractices\tfor\tfNIRS\tpublications’,\t Neurophotonics,\tvol.\t8,\t891 \nno.\t01,\tJan.\t2021,\tdoi:\t10.1117/1.NPh.8.1.012101.\t892 \n[6]\t J.\t Gemignani\t and\t J.\t Gervain,\t ‘Comparing\t different\t pre-processing\t routines\t for\t893 \ninfant\t fNIRS\t data’,\t Dev.\t Cogn.\t Neurosci. ,\t vol.\t 48,\t p.\t 100943,\t Apr.\t 2021,\t doi:\t894 \n10.1016/j.dcn.2021.100943.\t895 \n[7]\t L.\t Pollonini,\t H.\t Bortfeld,\t and\t J.\t S.\t Oghalai,\t ‘PHOEBE:\t a\t method\t for\t real\t time\t896 \nmapping\t of\t optodes-scalp\t coupling\t in\t functional\t near-infrared\t spectroscopy’,\t897 \nBiomed.\t Opt.\t Express ,\t vol.\t 7,\t no.\t 12,\t pp.\t 5104 –5119,\t Dec.\t 2016,\t doi:\t898 \n10.1364/BOE.7.005104.\t899 \n[8]\t L.\tM.\tHocke,\tI.\tK.\tOni,\tC.\tC.\tDuszynski,\tA.\tV.\tCorrigan,\tB.\tD.\tFrederick,\tand\tJ.\tF.\tDunn,\t900 \n‘Automated\t Processing\t of\t fNIRS\t Data —A\t Visual\t Guide\t to\t the\t Pitfalls\t and\t901 \nConsequences’,\t Algorithms,\t vol.\t 11,\t no.\t 5,\t Art.\t no.\t 5,\t May\t 2018,\t doi:\t902 \n10.3390/a11050067.\t903 \n[9]\t T.\t Arichi\t et\t al.,\t ‘Development\t of\t BOLD\t signal\t hemodynamic\t responses\t in\t the\t904 \nhuman\t brain’,\t NeuroImage,\t vol.\t 63,\t no.\t 2,\t pp.\t 663 –673,\t Nov.\t 2012,\t doi:\t905 \n10.1016/j.neuroimage.2012.06.054.\t906 \n[10]\tC.\t Issard\t and\t J.\t Gervain,\t ‘Variability\t of\t the\t hemodynamic\t response\t in\t infants:\t907 \nInfluence\tof\texperimental\tdesign\tand\tstimulus\tcomplexity’,\t Dev.\tCogn.\tNeurosci.,\t908 \nvol.\t33,\tpp.\t182–193,\tFeb.\t2018,\tdoi:\t10.1016/j.dcn.2018.01.009.\t909 \n[11]\tS.\tBaek,\tS.\tMarques,\tK.\tCasey,\tM.\tTesterman,\tF.\tMcGill,\tand\tL.\tEmberson,\t‘Attrition\t910 \nrate\tin\tinfant\tfNIRS\tresearch:\tA\tmeta -analysis’,\tInfancy,\tvol.\t28,\tno.\t3,\tpp.\t507 –911 \n531,\t2023,\tdoi:\t10.1111/infa.12521.\t912 \n[12]\tE.\tM.\tFrijia\tet\tal.,\t‘Functional\timaging\tof\tthe\tdeveloping\tbrain\twith\twearable\thigh-913 \ndensity\tdiffuse\toptical\ttomography:\tA\tnew\tbenchmark\tfor\tinfant\tneuroimaging\t914 \noutside\tthe\tscanner\tenvironment’,\t NeuroImage,\tvol.\t225,\tp.\t117490,\tJan.\t2021,\t915 \ndoi:\t10.1016/j.neuroimage.2020.117490.\t916 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n31\t\n[13]\tL.\tH.\tCollins-Jones\tet\tal.,\t‘Whole-head\thigh-density\tdiffuse\toptical\ttomography\tto\t917 \nmap\t infant\t audio-visual\t responses\t to\t social\t and\t non-social\t stimuli’,\t Imaging\t918 \nNeurosci.,\tvol.\t2,\tpp.\t1–19,\tSep.\t2024,\tdoi:\t10.1162/imag_a_00244.\t919 \n[14]\tS.\tLloyd-Fox\t et\tal.,\t‘Habituation\tand\tnovelty\tdetection\tfNIRS\tbrain\tresponses\tin\t920 \n5-\t and\t 8-month-old\t infants:\t The\t Gambia\t and\t UK’,\tDev.\t Sci.,\t vol.\t 22,\t no.\t 5,\t Sep.\t921 \n2019,\tdoi:\t10.1111/desc.12817.\t922 \n[15]\tS.\tLloyd-Fox,\tA.\tBlasi,\tA.\tVolein,\tN.\tEverdell,\tC.\tE.\tElwell,\tand\tM.\tH.\tJohnson,\t‘Social\t923 \nPerception\tin\tInfancy:\tA\tNear\tInfrared\tSpectroscopy\tStudy’,\t Child\tDev.,\tvol.\t80,\t924 \nno.\t4,\tpp.\t986–999,\tJul.\t2009,\tdoi:\t10.1111/j.1467-8624.2009.01312.x.\t925 \n[16]\tS.\t M.\t Hernandez\t and\t L.\t Pollonini,\t ‘NIRSplot:\t A\t Tool\t for\t Quality\t Assessment\t of\t926 \nfNIRS\t Scans’,\t in\tBiophotonics\t Congress:\t Biomedical\t Optics\t 2020\t (Translational,\t927 \nMicroscopy,\tOCT,\tOTS,\tBRAIN) ,\tWashington,\tDC:\tOptica\tPublishing\tGroup,\t2020,\t928 \np.\tBM2C.5.\tdoi:\t10.1364/BRAIN.2020.BM2C.5.\t929 \n[17]\tL.\t Pollonini,\t C.\t Olds,\t H.\t Abaya,\t H.\t Bortfeld,\t M.\t S.\t Beauchamp,\t and\t J.\t S.\t Oghalai,\t930 \n‘Auditory\t cortex\t activation\t to\t natural\t speech\t and\t simulated\t cochlear\t implant\t931 \nspeech\t measured\t with\t functional\t near-infrared\t spectroscopy’,\t Hear.\t Res.,\t vol.\t932 \n309,\tpp.\t84–93,\tMar.\t2014,\tdoi:\t10.1016/j.heares.2013.11.007.\t933 \n[18]\tB.\t Blanco\t et\t al.,\t ‘Cortical\t Responses\t to\t Social\t Stimuli\t in\t Infants\t at\t Elevated\t934 \nLikelihood\t of\t ASD\t and/or\t ADHD:\t a\t Prospective\t Cross-Condition\t fNIRS\t Study’,\t935 \nCortex,\tSep.\t2023,\tdoi:\t10.1016/j.cortex.2023.07.010.\t936 \n[19]\tL.\tK.\tGossé,\tP.\tPinti,\tF.\tWiesemann,\tC.\tE.\tE.\tElwell,\tand\tE.\tJ.\tH.\tJones,\t‘Developing\t937 \ncustomized\tNIRS-EEG\tfor\tinfant\tsleep\tresearch:\tmethodological\tconsiderations’,\t938 \nNeurophotonics,\t vol.\t 10,\t no.\t 3,\t p.\t 035010,\t Sep.\t 2023,\t doi:\t939 \n10.1117/1.NPh.10.3.035010.\t940 \n[20]\tS.\t Fleming\tet\t al.,\t ‘Normal\t ranges\t of\t heart\t rate\t and\t respiratory\t rate\t in\t children\t941 \nfrom\tbirth\tto\t18\tyears:\ta\tsystematic\treview\tof\tobservational\tstudies’,\tLancet,\tvol.\t942 \n377,\tno.\t9770,\tpp.\t1011–1018,\tMar.\t2011,\tdoi:\t10.1016/S0140-6736(10)62226-943 \nX.\t944 \n[21]\tMedlinePlus\tMedical\tEncyclopedia,\tNational\tLibrary\tof\tMedicine,\t‘Pulse’.\t2025.\t945 \n[Online].\tAvailable:\thttps://medlineplus.gov/ency/article/003399.htm\t946 \n[22]\tK.\tL.\tPerdue,\tA.\tWesterlund,\tS.\tA.\tMcCormick,\tand\tC.\tA.\tN.\tIii,\t‘Extraction\tof\theart\t947 \nrate\tfrom\tfunctional\tnear -infrared\tspectroscopy\tin\tinfants’,\t J.\tBiomed.\tOpt. ,\tvol.\t948 \n19,\tno.\t6,\tp.\t067010,\tJun.\t2014,\tdoi:\t10.1117/1.JBO.19.6.067010.\t949 \n[23]\tM.\tA.\tYücel\tet\tal.,\t‘Inclusivity\tin\tfNIRS\tStudies:\tQuantifying\tthe\tImpact\tof\tHair\tand\t950 \nSkin\t Characteristics\t on\t Signal\t Quality\t with\t Practical\t Recommendations\t for\t951 \nImprovement’,\tOct.\t28,\t2024,\tbioRxiv.\tdoi:\t10.1101/2024.10.28.620644.\t952 \n[24]\tS.\t Lloyd-Fox\t et\t al.,\t ‘The\t Brain\t Imaging\t for\t Global\t Health\t (BRIGHT)\t Project:\t953 \nLongitudinal\tcohort\tstudy\tprotocol’,\tGates\tOpen\tRes.,\tvol.\t7,\tp.\t126,\tOct.\t2023,\tdoi:\t954 \n10.12688/gatesopenres.14795.1.\t955 \n[25]\tJ.\tBenerradi\t et\tal.,\t‘Functional\tspecialisation\tacross\tthe\tfirst\tfive\tyears\tof\tlife:\ta\t956 \nlongitudinal\t characterisation\t of\t social\t perception\t with\t fNIRS’,\t Jul.\t 29,\t 2025,\t957 \nbioRxiv.\tdoi:\t10.1101/2025.07.28.667189.\t958 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n32\t\n[26]\tA.\t Blasi\t Ribera\t et\t al.,\t ‘Longitudinal\t habituation\t and\t novelty\t detection\t neural\t959 \nresponses\t from\t infancy\t to\t early\t childhood\t in\t The\t Gambia\t and\t UK’,\tDev.\t Cogn.\t960 \nNeurosci.,\tvol.\t76,\tp.\t101619,\tDec.\t2025,\tdoi:\t10.1016/j.dcn.2025.101619.\t961 \n[27]\tM.\t Rozhko\t et\t al.,\t ‘Effect\t of\t sleep\t stages\t on\t patterns\t of\t fNIRS\t hemodynamic\t962 \nresponse\tto\tauditory\tparadigms\tin\tone -month-old\tGambian\tand\tUK\tinfants’,\tJul.\t963 \n31,\t2025,\tbioRxiv.\tdoi:\t10.1101/2025.07.28.667238.\t964 \n[28]\tL.\t Kischkel,\t ‘Assessing\t infant\t neurocognitive\t development\t in\t resource -poor\t965 \nsettings:\t the\t example\t of\t memory\t development\t in\t the\t UK\t and\t The\t Gambia’,\t966 \nDoctoral\t thesis,\t University\t College\t London,\t 2019.\t [Online].\t Available:\t967 \nhttps://discovery.ucl.ac.uk/id/eprint/10073018/\t968 \n[29]\tS.\tLloyd-Fox\tet\tal.,\t‘Cortical\tspecialisation\tto\tsocial\tstimuli\tfrom\tthe\tfirst\tdays\tto\t969 \nthe\tsecond\tyear\tof\tlife:\tA\trural\tGambian\tcohort’,\tDev.\tCogn.\tNeurosci.,\tvol.\t25,\tpp.\t970 \n92–104,\tJun.\t2017,\tdoi:\t10.1016/j.dcn.2016.11.005.\t971 \n[30]\tS.\tLloyd-Fox,\tA.\tBlasi,\tand\tC.\tE.\tElwell,\t‘Illuminating\tthe\tdeveloping\tbrain:\tthe\tpast,\t972 \npresent\tand\tfuture\tof\tfunctional\tnear\tinfrared\tspectroscopy’,\tNeurosci.\tBiobehav.\t973 \nRev.,\t vol.\t 34,\t no.\t 3,\t pp.\t 269 –284,\t Mar.\t 2010,\t doi:\t974 \n10.1016/j.neubiorev.2009.07.008.\t975 \n[31]\tL.\tKatus\tet\tal.,\t‘Implementing\tneuroimaging\tand\teye\ttracking\tmethods\tto\tassess\t976 \nneurocognitive\t development\t of\t young\t infants\t in\t low -\t and\t middle -income\t977 \ncountries’,\t Gates\t Open\t Res. ,\t vol.\t 3,\t p.\t 1113,\t Apr.\t 2019,\t doi:\t978 \n10.12688/gatesopenres.12951.1.\t979 \n[32]\tA.\tBlasi,\tS.\tLloyd -Fox,\tMark.\tH.\tJohnson,\tand\tC.\tElwell,\t‘Test –retest\treliability\tof\t980 \nfunctional\t near\t infrared\t spectroscopy\t in\t infants’,\tNeurophotonics,\t vol.\t 1,\t no.\t 2,\t981 \nOct.\t2014,\tdoi:\t10.1117/1.NPh.1.2.025005.\t982 \n[33]\tS.\t Lloyd-Fox,\t J.\t E.\t Richards,\t A.\t Blasi,\t D.\t G.\t M.\t Murphy,\t C.\t E.\t Elwell,\t and\t M.\t H.\t983 \nJohnson,\t ‘Coregistering\t functional\t near-infrared\t spectroscopy\t with\t underlying\t984 \ncortical\tareas\tin\tinfants’,\tNeurophotonics,\tvol.\t1,\tno.\t2,\tp.\t025006,\tOct.\t2014,\tdoi:\t985 \n10.1117/1.NPh.1.2.025006.\t986 \n[34]\tL.\tH.\tCollins -Jones\tet\tal.,\t‘Longitudinal\tinfant\tfNIRS\tchannel -space\tanalyses\tare\t987 \nrobust\t to\t variability\t parameters\t at\t the\t group-level:\t An\t image\t reconstruction\t988 \ninvestigation’,\t NeuroImage,\t vol.\t 237,\t p.\t 118068,\t Aug.\t 2021,\t doi:\t989 \n10.1016/j.neuroimage.2021.118068.\t990 \n[35]\tS.\tMcCann,\t‘Nutrition,\tGrowth\tand\tNeurodevelopment\tAcross\tInfancy\tand\tEarly\t991 \nChildhood\t in\t Rural\t Gambia.’,\t PhD\t Thesis,\t King’s\t College\t London,\t London,\t UK,\t992 \n2020.\t993 \n[36]\tT.\tJ.\tHuppert,\tS.\tG.\tDiamond,\tM.\tA.\tFranceschini,\tand\tD.\tA.\tBoas,\t‘HomER:\ta\treview\t994 \nof\t time-series\t analysis\t methods\t for\t near-infrared\t spectroscopy\t of\t the\t brain’,\t995 \nAppl.\t Opt. ,\t vol.\t 48,\t no.\t 10,\t pp.\t D280 –D298,\t Apr.\t 2009,\t doi:\t996 \n10.1364/AO.48.00D280.\t997 \n[37]\tR.\tDi\tLorenzo\tet\tal.,\t‘Recommendations\tfor\tmotion\tcorrection\tof\tinfant\tfNIRS\tdata\t998 \napplicable\tto\tmultiple\tdata\tsets\tand\tacquisition\tsystems’,\t NeuroImage,\tvol.\t200,\t999 \npp.\t 511–527,\t Oct.\t 2019,\t doi:\t1000 \nhttps://doi.org/10.1016/j.neuroimage.2019.06.056.\t1001 \n[38]\tG.\tA.\tZimeo\tMorais\tet\tal.,\t‘Non-neuronal\tevoked\tand\tspontaneous\themodynamic\t1002 \nchanges\t in\t the\t anterior\t temporal\t region\t of\t the\t human\t head\t may\t lead\t to\t1003 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n33\t\nmisinterpretations\t of\t functional\t near -infrared\t spectroscopy\t signals’,\t1004 \nNeurophotonics,\t vol.\t 5,\t no.\t 1,\t p.\t 011002,\t Aug.\t 2017,\t doi:\t1005 \n10.1117/1.NPh.5.1.011002.\t1006 \n[39]\tA.\tGelman\tand\tJ.\tHill,\t Data\tanalysis\tusing\tregression\tand\tmultilevel/hierarchical\t1007 \nmodels.\tCambridge\tuniversity\tpress,\t2006.\t1008 \n[40]\tR\tCore\tTeam,\t R:\tA\tLanguage\tand\tEnvironment\tfor\tStatistical\tComputing .\tVienna,\t1009 \nAustria:\t R\t Foundation\t for\t Statistical\t Computing,\t 2024.\t [Online].\t Available:\t1010 \nhttps://www.R-project.org/\t1011 \n[41]\tD.\t Bates,\t ‘Fitting\t linear\t mixed -effects\t models\t using\t lme4’,\t ArXiv\t Prepr.\t1012 \nArXiv14065823,\t2014.\t1013 \n[42]\tH.\tAkaike,\t‘A\tnew\tlook\tat\tthe\tstatistical\tmodel\tidentification’,\tIEEE\tTrans.\tAutom.\t1014 \nControl,\tvol.\t19,\tno.\t6,\tpp.\t716–723,\t1974.\t1015 \n[43]\tJ.\tW.\tHarris\tand\tH.\tStöcker,\tHandbook\tof\tmathematics\tand\tcomputational\tscience.\t1016 \nSpringer\tScience\t&\tBusiness\tMedia,\t1998.\t1017 \n[44]\tH.\tF.\tBehrendt,\tC.\tFirk,\tC.\tA.\tN.\tIii,\tand\tK.\tL.\tPerdue,\t‘Motion\tcorrection\tfor\tinfant\t1018 \nfunctional\t near-infrared\t spectroscopy\t with\t an\t application\t to\t live\t interaction\t1019 \ndata’,\t Neurophotonics,\t vol.\t 5,\t no.\t 1,\t p.\t 015004,\t Feb.\t 2018,\t doi:\t1020 \n10.1117/1.NPh.5.1.015004.\t1021 \n[45]\tS.\tJahani,\tS.\tK.\tSetarehdan,\tD.\tA.\tBoas,\tand\tM.\tA.\tYücel,\t‘Motion\tartifact\tdetection\t1022 \nand\tcorrection\tin\tfunctional\tnear -infrared\tspectroscopy:\ta\tnew\thybrid\tmethod\t1023 \nbased\t on\t spline\t interpolation\t method\t and\t Savitzky –Golay\t filtering’,\t1024 \nNeurophotonics,\t vol.\t 5,\t no.\t 1,\t p.\t 015003,\t Feb.\t 2018,\t doi:\t1025 \n10.1117/1.NPh.5.1.015003.\t1026 \n[46]\tS.\t Sutoko\t et\t al.,\t ‘Adaptive\t algorithm\t utilizing\t acceptance\t rate\t for\t eliminating\t1027 \nnoisy\t epochs\t in\t block -design\t functional\t near -infrared\t spectroscopy\t data:\t1028 \napplication\t to\t study\t in\t attention\t deficit/hyperactivity\t disorder\t children’,\t1029 \nNeurophotonics,\t vol.\t 5,\t no.\t 4,\t p.\t 045001,\t Oct.\t 2018,\t doi:\t1030 \n10.1117/1.NPh.5.4.045001.\t1031 \n[47]\tA.\tSherafati,\tA.\tT.\tEggebrecht,\tK.\tM.\tBergonzi,\tT.\tM.\tBurns-Yocum,\tand\tJ.\tP.\tCulver,\t1032 \n‘Improvements\tin\tfunctional\tdiffuse\toptical\ttomography\tmaps\tby\tglobal\tmotion\t1033 \ncensoring\t techniques’,\t in\t Microscopy\t Histopathology\t and\t Analytics ,\t Optica\t1034 \nPublishing\tGroup,\t2018,\tpp.\tJW3A-51.\t1035 \n[48]\tA.\tSherafati\tet\tal.,\t‘Global\tmotion\tdetection\tand\tcensoring\tin\thigh-density\tdiffuse\t1036 \noptical\ttomography’,\tHum.\tBrain\tMapp.,\tvol.\t41,\tno.\t14,\tpp.\t4093–4112,\t2020,\tdoi:\t1037 \n10.1002/hbm.25111.\t1038 \n[49]\tG.\tLee,\tS.\tH.\tJin,\tS.\tH.\tLee,\tB.\tAbibullaev,\tand\tJ.\tAn,\t‘fNIRS\tmotion\tartifact\tcorrection\t1039 \nfor\t overground\t walking\t using\t entropy\t based\t unbalanced\t optode\t decision\t and\t1040 \nwavelet\t regression\t neural\t network’,\t in\t2017\t IEEE\t International\t Conference\t on\t1041 \nMultisensor\tFusion\tand\tIntegration\tfor\tIntelligent\tSystems\t(MFI) ,\tNov.\t2017,\tpp.\t1042 \n186–193.\tdoi:\t10.1109/MFI.2017.8170427.\t1043 \n[50]\tS.\tA.\tFurdon\tand\tD.\tA.\tClark,\t‘SCALP\tHAIR\tCHARACTERISTICS\tIN\tTHE\tNEWBORN\t1044 \nINFANT’,\t Adv.\t Neonatal\t Care ,\t vol.\t 3,\t no.\t 6,\t p.\t 286,\t Dec.\t 2003,\t doi:\t1045 \n10.1016/j.adnc.2003.09.005.\t1046 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n34\t\n[51]\tL.\t L.\t Emberson,\t S.\t L.\t Crosswhite,\t J.\t R.\t Goodwin,\t A.\t J.\t Berger,\t and\t R.\t N.\t Aslin,\t1047 \n‘Isolating\tthe\teffects\tof\tsurface\tvasculature\tin\tinfant\tneuroimaging\tusing\tshort -1048 \ndistance\t optical\t channels:\t a\t combination\t of\t local\t and\t global\t effects’,\t1049 \nNeurophotonics,\t vol.\t 3,\t no.\t 3,\t p.\t 031406,\t Apr.\t 2016,\t doi:\t1050 \n10.1117/1.NPh.3.3.031406.\t1051 \n[52]\tE.\t T.\t Margolis\t et\t al. ,\t ‘Modality -level\t obstacles\t and\t initiatives\t to\t improve\t1052 \nrepresentation\tin\tfetal,\tinfant,\tand\ttoddler\tneuroimaging\tresearch\tsamples’,\tDev.\t1053 \nCogn.\tNeurosci.,\tvol.\t72,\tp.\t101505,\tApr.\t2025,\tdoi:\t10.1016/j.dcn.2024.101505.\t1054 \n[53]\tA.\tP.\tField\tand\tD.\tB.\tWright,\t‘A\tPrimer\ton\tUsing\tMultilevel\tModels\tin\tClinical\tand\t1055 \nExperimental\tPsychopathology\tResearch’,\t J.\tExp.\tPsychopathol. ,\tvol.\t2,\tno.\t2,\tpp.\t1056 \n271–293,\tMay\t2011,\tdoi:\t10.5127/jep.013711.\t1057 \n[54]\tE.\tMercure\tet\tal.,\t‘Language\tExperience\tImpacts\tBrain\tActivation\tfor\tSpoken\tand\t1058 \nSigned\t Language\t in\t Infancy:\t Insights\t From\t Unimodal\t and\t Bimodal\t Bilinguals’,\t1059 \nNeurobiol.\tLang.,\tvol.\t1,\tno.\t1,\tpp.\t9–32,\tMar.\t2020,\tdoi:\t10.1162/nol_a_00001.\t1060 \n[55]\tC.\t Gerloff,\t M.\t A.\t Yücel,\t L.\t Mehlem,\t K.\t Konrad,\t and\t V.\t Reindl,\t ‘NiReject:\t toward\t1061 \nautomated\t bad\t channel\t detection\t in\t functional\t near -infrared\t spectroscopy’,\t1062 \nNeurophotonics,\t vol.\t 11,\t no.\t 4,\t p.\t 045008,\t Nov.\t 2024,\t doi:\t1063 \n10.1117/1.NPh.11.4.045008.\t1064 \n[56]\tM.\tHahsler,\tC.\tBuchta,\tB.\tGruen,\tand\tK.\tHornik,\t arules:\tMining\tAssociation\tRules\t1065 \nand\t Frequent\t Itemsets .\t 2025.\t [Online].\t Available:\t https://CRAN.R -1066 \nproject.org/package=arules\t1067 \n[57]\tD.\tRobinson,\tA.\tHayes,\tand\tS.\tCouch,\t broom:\tConvert\tStatistical\tObjects\tinto\tTidy\t1068 \nTibbles.\t2025.\t[Online].\tAvailable:\thttps://CRAN.R-project.org/package=broom\t1069 \n[58]\tJ.\tFox\tand\tS.\tWeisberg,\t An\tR\tCompanion\tto\tApplied\tRegression ,\t3rd\ted.\tThousand\t1070 \nOaks,\tCA:\tSage,\t2019.\t1071 \n[59]\tT.\tBarrett\t et\tal.,\tdata.table:\tExtension\tof\t`data.frame` .\t2024.\t[Online].\tAvailable:\t1072 \nhttps://CRAN.R-project.org/package=data.table\t1073 \n[60]\tM.\t Corporation\t and\t S.\t Weston,\t doParallel:\t Foreach\t Parallel\t Adaptor\t for\t the\t1074 \n‘parallel’\t Package .\t 2022.\t [Online].\t Available:\t https://CRAN.R -1075 \nproject.org/package=doParallel\t1076 \n[61]\tH.\tWickham,\tR.\tFrançois,\tL.\tHenry,\tK.\tMüller,\tand\tD.\tVaughan,\t dplyr:\tA\tGrammar\t1077 \nof\t Data\t Manipulation .\t 2025.\t [Online].\t Available:\t https://CRAN.R -1078 \nproject.org/package=dplyr\t1079 \n[62]\tD.\t Lüdecke,\t M.\t S.\t Ben-Shachar,\t D.\t Makowski,\t and\t I.\t Patil,\teffectsize:\t Indices\t of\t1080 \nEffect\t Size\t and\t Standardized\t Parameters .\t 2024.\t [Online].\t Available:\t1081 \nhttps://CRAN.R-project.org/package=effectsize\t1082 \n[63]\tD.\t Meyer,\t E.\t Dimitriadou,\t K.\t Hornik,\t A.\t Weingessel,\t and\t F.\t Leisch,\te1071:\t Misc\t1083 \nFunctions\t of\t the\t Department\t of\t Statistics,\t Probability\t Theory\t Group\t (Formerly:\t1084 \nE1071),\t TU\t Wien .\t 2024.\t [Online].\t Available:\t https://CRAN.R -1085 \nproject.org/package=e1071\t1086 \n[64]\tMicrosoft\t and\t S.\t Weston,\t foreach:\t Provides\t Foreach\t Looping\t Construct.\t 2022.\t1087 \n[Online].\tAvailable:\thttps://CRAN.R-project.org/package=foreach\t1088 \n[65]\tH.\t Wickham,\tggplot2:\t Elegant\t Graphics\t for\t Data\t Analysis.\t Springer-Verlag\t New\t1089 \nYork,\t2016.\t[Online].\tAvailable:\thttps://ggplot2.tidyverse.org\t1090 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint \n\n\t \t \t\n\t\n\t \t \t\n\t \t\n35\t\n[66]\tA.\tKassambara,\t ggpubr:\t‘ggplot2’\tBased\tPublication\tReady\tPlots .\t2023.\t[Online].\t1091 \nAvailable:\thttps://CRAN.R-project.org/package=ggpubr\t1092 \n[67]\tB.\tAuguie,\t gridExtra:\tMiscellaneous\tFunctions\tfor\t‘Grid’\tGraphics .\t2017.\t[Online].\t1093 \nAvailable:\thttps://CRAN.R-project.org/package=gridExtra\t1094 \n[68]\tA.\tKuznetsova,\tP.\tB.\tBrockhoff,\tand\tR.\tH.\tB.\tChristensen,\t‘lmerTest\tPackage:\tTests\t1095 \nin\tLinear\tMixed\tEffects\tModels’,\tJ.\tStat.\tSoftw.,\tvol.\t82,\tno.\t13,\tpp.\t1–26,\t2017,\tdoi:\t1096 \n10.18637/jss.v082.i13.\t1097 \n[69]\tW.\t N.\t Venables\t and\t B.\t D.\t Ripley,\tModern\t Applied\t Statistics\t with\t S,\t 4th\t ed.\t New\t1098 \nYork:\tSpringer,\t2002.\t1099 \n[70]\tD.\tBates\tand\tM.\tMaechler,\t Matrix:\tSparse\tand\tDense\tMatrix\tClasses\tand\tMethods .\t1100 \n2024.\t[Online].\tAvailable:\thttps://CRAN.R-project.org/package=Matrix\t1101 \n[71]\tK.\t Bartoń,\t MuMIn:\t Multi -Model\t Inference .\t 2024.\t [Online].\t Available:\t1102 \nhttps://CRAN.R-project.org/package=MuMIn\t1103 \n[72]\tH.\tWickham\tand\tD.\tSeidel,\tscales:\tScale\tFunctions\tfor\tVisualization.\t2023.\t[Online].\t1104 \nAvailable:\thttps://CRAN.R-project.org/package=scales\t1105 \n[73]\tH.\tWickham,\tD.\tVaughan,\tand\tM.\tGirlich,\t tidyr:\tTidy\tMessy\tData .\t2024.\t[Online].\t1106 \nAvailable:\thttps://CRAN.R-project.org/package=tidyr\t1107 \n[74]\tS.\t Garnier,\t N.\t Ross,\t R.\t Rudis,\t A.\t Filippi,\t A.\t Kaplan,\t and\t A.\t Hackett,\t viridis:\t1108 \nColorblind-Friendly\tColor\tMaps\tfor\tR .\t2024.\t[Online].\tAvailable:\thttps://CRAN.R -1109 \nproject.org/package=viridis\t1110 \n1111 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted October 7, 2025. ; https://doi.org/10.1101/2025.10.06.680288doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}