Top of project is at: https://osf.io/f6uyt/ (Project: Optimal laterality analysis fMRI) This project uses existing data from Lisa Bruckert’s thesis to explore different methods for estimating language lateralisation. This is a complex project that uses Matlab (SPM) and R for analysis/figure creation, as well as MRIcroGL (python). See **Flowchart for Wiki.png** for organisation of files (described also below). Main file at top level is **optimising measurement.qmd**, a quarto file in R that generates the manuscript and some of the figures. This dataset has previously been used in the following studies: Bruckert, L. 2016. “Is Language Laterality Related to Language Abilities?” PhD thesis. https://ora.ox.ac.uk/objects/uuid:05e80d0d-8d0b-4cb2-8f94-22763603fab5. Bruckert, L., Thompson, P. A., Watkins, K. E., Bishop, D. V. M., & Woodhead, Z. V. J. (2021). Investigating the effects of handedness on the consistency of lateralization for speech production and semantic processing tasks using functional transcranial Doppler sonography. Laterality, 26(6), 680–705. https://doi.org/10.1080/1357650X.2021.1898416 Thompson, P. A., Watkins, K. E., Woodhead, Z. V. J., & Bishop, D. V. M. (2023). Generalized models for quantifying laterality using functional transcranial Doppler ultrasound. Human Brain Mapping, 44(1), 35–48. https://doi.org/10.1002/hbm.26138 NB for related projects see: https://osf.io/tkpm2/ (Project: CANDICE A2: Assessing reliability of language laterality measures using fTCD) https://osf.io/gw4en/ (Project: A General Statistical Analysis for quantifying laterality using functional Transcranial Doppler Ultrasound) **fMRI Data files** **fMRI_tmaps** https://osf.io/tgp4k/ This component on OSF contains one folder for each of 5 contrasts (WG1, PP1, PP3, PP5, AN5), with all participants within the task folder. The participant ID is a 7 character code. The first 3 characters are a unique ID number that matches the number in the fTCD data files. The last letter is A or T, to denote whether this person was originally categorised by Bruckert as atypical or typical. The number after tstat in the file name denotes the contrast used for the t-map. For the main analysis we use: contrast 1 for Word Generation (words vs rest) contrast 5 for Auditory Naming (auditory sentences vs backward speech) contrast 1 for Semantic Matching (Pyramids and Palm Trees picture condition vs rest). In supplementary material we also consider two other contrasts from Pyramids and Palm Trees: condition 3, which is control perceptual condition (lines) vs rest; and condition 5, which is pictures vs perceptual control). These files were created from Bruckert’s original fMRI files that had been processed by FEAT (held by Oxford University). They were reprocessed to standard brain to be suitable for processing with LI-Toolbox. The conversion was done with script: **batch_FLIRT_Lisa_b.sh**. **fMRI_masks** https://osf.io/f6uyt/ These are taken from SPM’s LI_toolbox, except for the LI_mca_mask, which is formed by combining the frontal, temporal and parietal masks. To use this mask as a ‘custom mask’ with the LI toolbox or mirror method (see below), this mask should be added to the other masks in SPM12/Toolbox/LI/Data **Analysis scripts** **LI_DB.m**: Matlab script that is identical to LI.m from LI toolbox, except that it calls function LI_boot_DB.m rather than LI_boot.m. If these files are added to LI toolbox in SPM12, then they should run just like the original LI toolbox scripts. To obtain results described in write-up, can select all the .nii files required, and then select options: - Thresholding method: bootstrap - Inclusive mask: select frontal/parietal/temporal/cerebellar or ‘custom’ to select the MCA mask. - Exclusive mask: Midline +/- 5 mm - Optional steps: None - Use bootstrap defaults: Yes **LI_boot_DB.m**: Function called from LI_DB.m. Same as the original toolbox LI_boot.m, except for minor changes that are described in the comments at top of script, and also commented within script. Have added 95% CI for weighted mean – by taking 2.5th and 97.5th centiles from weighted distribution. Weighted distribution is bx. The trimmed mean min and max have been overwritten by the 95% CI for weighted mean in the output. The maximum threshold is now also saved in output to ‘topthresh’. This makes it possible to explore how threshold range affects LI (not included in write-up). In addition, minor formatting change for graphic output file so it fits on paper. **LI_boot_DB** calls **LI_boot_hf** as usual – i.e. helper function is unchanged from toolbox version. **mirror.m**: Matlab script that computes the mirror LI for individual .nii files. Uses SPM functions to select files and masks. Script calls **mirror_boot.m** which does the mirror computations. Creates output file with default name of *mirrorsummary.csv*. N.B. Although this is based on LI.m from toolbox, it is not so user-friendly. It will ask the user to select files to analyse, and then present options for inclusive and exclusive masks. It does not give any graphical output. **make_diff_ni.m** creates the individual difference files that are used for averaged difference maps in Figures 7 and 14. The whole t-statistic map for each individual is divided into L and R and the R side flipped and subtracted from the L. **Compute_GAM_LI.rmd**: FCTD analysis. Creates LIs from raw FTCD .eps files using GAM model. This is based on script used by Thompson et al, (https://osf.io/gw4en/) modified to analyse the data from Semantic Matching (aka PPT, Pyramids and Palm trees). The raw .eps data are on https://osf.io/gw4en/. There is also a copy here: https://osf.io/bcxus/files/osfstorage. This script also reads data from *Bruckert_WordGen_results.csv* (original from Bruckert with some demographic data), and from *WordGen_trial_inclusion_x.csv* and *PPTT_trial_inclusion_x.csv*, which have 0 or 1 denoting if individual trials are included. Most people have all trials, but occasional trials marked for exclusion if signal dropout or other interruption to recording. **Process_LI_mirror_ftcd.rmd**: This script reads in LI data from fTCD and two methods of fMRI (toolbox and mirror) and creates a long-form file called *bigdf.csv* which can be used when generating plots etc. It also reads *bruckert_demog.csv* which has data on handedness and original laterality classification. It reads files called *toolboxsummary.csv* and *mirrorsummary.csv*, which are created by **LI_DB.m** and **mirror.m**, respectively. These should have data from all tasks, participants and masks. For fTCD, it reads files *Doppler_GAM_PP1.csv* and *Doppler_GAM_WG1.csv*, which are generated by **Compute_GAM_LI.rmd**. Script also generates the heatmap for correlations across ROIs/tasks. **Images** **Figure 1**: Averaged blood flow velocity over course of a word generation trial. *Images/Doppler_demo.eps* (.png version also saved) Figure created within **Optimising measurement.qmd** Reads data: *demoDop_RH_WG.csv*, which is saved on Data_processed **Figure 2**: LI estimates (based on mean activation) at different thresholds in a sample individual fMRI dataset using a word generation task for activation. *Images/LIplot.eps* Created from **plot histos from toolbox.R** (saved on Analysis Scripts) Reads data: *threshvals.csv* – this is output from sample case from LI toolbox (modified by DB so can save it). Threshvals is saved on **Data_processed** **Figure 3**: Distributions of bootstrapped LI estimates at different thresholds in a sample individual fMRI dataset using a word generation task for activation. *Images/densitythreshplot.tiff* Created from **plot histos from toolbox.R** (saved on Analysis Scripts) NB. Various formats saved, including high resolution .tiff. The .eps version doesn’t save the shading. Reads data: *jn_all.csv* – this is output from LI toolbox (modified by DB so can save it). It is a very large matrix, with one row for each threshold, and one column for each LI estimate at that threshold. *jn_all* is saved on Data_processed **Figure 4**: Weighted and unweighted distributions of bootstrapped LI estimates from the same dataset as Figure 2. *Images/histoplot.tiff* Created from **plot histos from toolbox.R** (saved on Analysis Scripts) NB. Various formats saved, including high resolution .tiff. Reads data: *jn_all.csv* – this is output from LI toolbox (modified by DB so can save it). **Figure 5**: Time-course of trials in expressive (word generation) and receptive (semantic matching) tasks *Images/timecourse.png* (created from powerpoint – original pptx saved as well as pdf) **Figure 6**: Heatmaps of t-statistic maps for the three tasks superimposed on SPM152 brain; averages for typical and atypical groups. *Images/bigbrains.tiff* Created from MRIcroGL – python script is on Analysis scripts The output from the python script is assembled into composite figure using **magick_image_manipulation.rmd** (see Analysis scripts) **Figure 7**: Heatmaps of difference t-statistic maps used in the mirror method for the three tasks superimposed on SPM152 brain; averages for typical and atypical groups. *Images/mirrorbrains.tiff* Created from MRIcroGL – python script (python script for mirror maps) is on Analysis scripts The output from the python script is assembled into composite figure using **magick_image_mirror.rmd** (see Analysis scripts) **Figure 8**: Scatterplots of relationship between fTCD and two methods of fMRI laterality for MCA region. *Images/Doppler_fmri2meth_MCA_2task.eps* Figure created within **Optimising measurement.qmd** Uses data *bigdf.csv*, which is in Data_Processed (and was created using **process_LI_mirror_ftcd.rmd**) **Figure 9**: Scatterplots of relationship for Word Generation between fTCD and two methods of fMRI laterality for 4 ROIs *Images/Doppler_fmri2meth_4ROI_WG1.eps* Figure created within ‘Optimising measurement.qmd’ Uses data ‘bigdf.csv’, which is in Data_Processed (and was created using **process_LI_mirror_ftcd.rmd**) **Figure 10**: Scatterplots of relationship for Semantic Matching between fTCD and two methods of fMRI laterality for 4 ROIs *Images/Doppler_fmri2meth_4ROI_PP1.eps* Figure created within ‘Optimising measurement.qmd’ Uses *data bigdf.csv,* which is in Data: Processed (and was created using **process_LI_mirror_ftcd.rmd**) **Figure 11**: Heatmap of Spearman correlations between LIs for three tasks and four ROIs. Mirror method is above the diagonal, and toolbox method below the diagonal. *Images/heatcorrel.eps* Created using **process_LI_mirror_ftcd.rmd** Uses data *bigdf.csv*, which is in Data_Processed **Figure 12**: Stimuli for semantic (left) and perceptual (right) matching conditions in the ‘pyramids and palm trees’ task *Images/PPT stimuli_orig.tiff* **Figure 13**: Patterns of brain activation on the semantic and perceptual ‘pyramids and palm trees’ conditions, and the contrast between them for typical and atypical individuals (classified on fTCD Word Generation) *Images/bigbrainsPP.tiff* Created from MRIcroGL – python script is on Analysis scripts The output from the python script is assembled into composite figure using **magick_image_manipulation.rmd** (see Analysis scripts) **Figure 14**: Left minus right t-statistic maps for the three contrasts in semantic matching task; averages for typical and atypical groups *Images/mirrorbrainsPP.tiff* Created from MRIcroGL – python script (python script for mirror maps) is on Analysis scripts The output from the python script is assembled into composite figure using **magick_image_mirror.rmd** (see Analysis scripts) **Figure 15**: Mean L and R activations by task and ROI *Images/LR_3tasks_4mask.eps* Figure created within ‘Optimising measurement.qmd’ Uses data *bigdf.csv*, which is in Data_Processed