fMRI replicability reporting based on Poldrack et al., 2008 https://www.sciencedirect.com/science/article/pii/S1053811907011020?via%3Dihub ***Experimental design*** **Design specification** Paradigm 1: - 12 blocks (3 face, 2 body, 2 object, 3 scene, 2 baseline) - 6 (3s) movies per block - each block 18s Paradigm 2: - 13 blocks (3 face, 3 body, 3 object, 3 scene, 1 baseline) - 6 (2.7s) movies per block - 6 (300ms) still images per block - each block 18s **Task Specification:** - passive viewing - videos and images of faces, bodies, objects, scenes, and abstract, curvy baseline (posted in OSF) - blocks repeated (in pseudoranodm order; see stimuli presentation code) after each block played once **Planned comparisons:** - faces > objects ***Human Subjects*** **Details on subject sample** - 86 subjects recruited - 2.1-11.9 months - 41 female; 45 male - subjects had to have 96 low-motion volumes (defined in methods) to be included in group random effects analysis (49 subjects met this criteria) - subjects had to have 2 runs with 96 volumes to be included in fROI analysis (30 subjects met this criteria) **Ethics approval** MIT IRB # 0809002909 **Behavioral performance** an adult watched the infant for the duration of the scan to record when the infant was attending to the stimuli ***Data acquisition*** **Image properties--as acquired** First group: - Siemens Trio 3T - custom infant coil - variable number of images acquired depending on subject - sinusoidal pulse sequence - 22 near-axial slices - repetition time, TR = 3s - echo time, TE = 43 ms - flip angle = 90 degrees - field of view, FOV = 192 mm - matrix = 64x64 - slice thickness = 3 mm - slice gap = 20.6 mm Second group: - Siemens Prisma 3T - custom infant coil - variable number of images acquired depending on subject - 44 near-axial slices - repetition time, TR = 3s - echo time, TE = 30ms - flip angle = 90 degrees - field of view, FOV = 160 mm - matrix = 80x80 - slice thickness = 2 mm - slice gap = 0 mm Data preprocessing - custom data processing pipeline using matlab, freesurfer, and fsl Pre-processing: general 1. subrun selection based on motion criteria (>2mm or degrees of motion spliced into subruns) 2. low-motion volumes re-configured as a single nifti file 3. individual functional image was extracted from the middle of the subrun to be used for registering the subruns to one another 4. each subrun motion corrected using FSL MCFLIRT 5. If more than 3 consecutive images had more than 0.5 mm or 0.5 degrees of motion, there had to be at least 7 consecutive low-motion volumes following the last high-motion volume in order for those volumes to be included in the analysis 6. each subrun had to have at least 24 volumes after accounting for motion and sleep TRs 7. skull-stripped using FSL BET2 8. intensity normalized **Within- and Inter-subject registrations** 1. middle image of each subrun was extracted and used as an example image for registration. If the middle image was corrupted by motion or distortion, a better image was selected to be the example image. The example image from the middle subrun of the first visit was used as the target image 2. All other subruns from each subject were registered to that subject’s target image 3. The target image for each subject was registered to a template image. 4. For data collected with Coil 1, the template image was the same one used previously (2). For data collected with Coil 2, a template image was selected from a single subject from whom we had a high-resolution anatomical image on which to display functional data. 5. Subrun and target image registrations were concatenated so that each subrun was individually registered to template space. 6. Given the distortion of the images due to the sinusoidal EPI (quiet sequence), regular registration tools do not effectively register functional data. As such, we attempted to register each image using a rigid, an affine, and a partial affine registration with FSL FLIRT. The best image was selected by eye from the three registration options and manually tuned for the best possible data alignment. Each image took between 2 and 8 hours of human labor to register. 7. To display group results, images collected with Coil 1 were transformed to an anatomical image collected with the same coil. Images collected with Coil 2 were transformed to the anatomical space of an image collected with Coil 2. **Smoothing** spatially smoothed with a 3mm FWHM Gaussian kernel using FSL SUSAN ***Statistical modeling*** **General Issues** All code provided on OSF Instrasubject fMRI modeling info - whole-brain voxel-wise general linear model (GLM) using custom MATLAB scripts - block-based - 4 condition regressors (faces, scenes, bodies, and objects) - 6 motion regressors - linear trend regressor - 5 PCA noise regressors - Condition regressors were defined as a boxcar function for the duration of the stimulus presentation - Infant inattention or sleep was accounted for using a single impulse nuisance (‘sleep’) regressor. The sleep regressor was defined as a boxcar function with a 1 for each TR the infant was not looking at the stimuli, and the corresponding TR was set to 0 for all condition regressors. - Boxcar condition and sleep regressors were convolved with an infant hemodynamic response function (HRF) that is characterized by a longer time to peak and deeper undershoot compared to the standard adult HRF. - Data and regressors were demeaned for each subrun - demeaned data and regressors were concatenated across subruns - beta values were computed for each condition in a whole-brain voxel-wise GLM - Subject-level contrast maps were computed as the difference between the face beta and the object beta for each voxel using in-house MATLAB code. - subject-level contrast maps were computed voxel-wise as the difference between the face beta and the object beta (face-object) **Group modeling info** a group random effects analysis was run using Freesurfer mri_concat and Freesurfer mri_glmfit ***Statistical inference*** **Inference on statistic image** - We only interpreted significance on the group random effects maps - a cluster of voxels was considered significant the voxel-level significance was P<0.05 and within a parcel of interest (lateral, ventral, STS, or MPFC) **ROI analysis** Search Space Selection: Desikan areas were selected such that each anatomical ROI would contain known areas of functional specialization in adults: - The lateral-occipital parcel includes the Desikan lateral occipital area – cortical areas that include the occipital place area (OPA), occipital face area (OFA), extrastriate body area (EBA), and lateral occipital object area (LOC) - The ventral-temporal parcel includes parahippocampal, fusiform, and entorhinal areas – cortical areas that include the fusiform face area (FFA), anterior temporal lobe (ATL), and parahippocampal place area (PPA) - The STS parcel includes transverse temporal, middle temporal, bank STS, superior temporal, supramarginal, and inferior temporal areas and MPFC includes the superior frontal area - we created a hand-drawn parcel to include all infant subcortical areas excluding cerebellum. - The lateral-occipital, ventral-temporal, and STS parcels were validated to produce face-selective results corresponding to the OFA, FFA, and fSTS regions in a group of four adults with Paradigm 1. Search Space Creation: 1. combined Freesurfer surface labels from the Desikan atlas 2. converted labels to MNI volume space 3. inflated and filled the volume parcels and removed any overlap from all parcels 4. Adult parcels were transformed to infant space by registering the standard infant anatomical to the MNI brain and applying an inverse transformation 5. All infant parcels were edited to remove any remaining overlap Subruns for ROI analysis Due to the variable amount of data in each subrun for each subject and the impact this could have on reliable parameter estimates from the GLM, we first combined or split subruns to approximately equate the amount of data across subruns within subjects. For example, if a subject had three subruns and the first was 35 volumes, the second was 57 volumes and third was 220 volumes, then we concatenated the first two subruns to create one subrun and we split the third subrun into 2 resulting in a total of three subruns with approximately 100 volumes per subrun. Voxel selection & condition response extraction - anatomically defined parcels were transformed to subject native space - We used an iterative leave-one-subrun-out procedure such that data were concatenated across all subruns except one prior to the whole-brain voxel-wise GLM and contrast were computed (described above) - The top 5% most significant voxels for faces>objects within an anatomical constraint parcel were selected as the fROI for that subject - parameter estimates were extracted from a GLM on the left-out subrun - amounts of data from each subrun are accounted for using a weighted average (w=(cT(XTX)-1c)-1; where c is the contrast vector and X is the design matrix - Beta values were averaged across participants and experiments. Figures & tables - group RFX statistical map is -log(p) - threshold is 1.3-2 (equates to P=0.05-0.01) - underlying image is a representative functional image transformed to template space