This web site provides files associated with the following paper: > Kay, K., Jamison, K.W., Zhang, R.-Y., Ugurbil, K. A temporal decomposition method for identifying venous effects in task-based fMRI. Nature Methods (2020). The paper is available at https://doi.org/10.1038/s41592-020-0941-6. The files on this site include (1) an archive of the analysis scripts used for the paper, (2) a small example dataset that is convenient for trying out the technique, (3) pre-processed fMRI data for all datasets used in the paper, and (4) FreeSurfer output directories for the subjects used. The code that implements the TDM technique is not provided here but is organized at http://github.com/kendrickkay/TDM/. This repository also includes an example script. The outputs of this script are viewable at https://htmlpreview.github.io/?https://github.com/kendrickkay/TDM/blob/master/html/example1.html. A video walk-through of the example script is viewable at https://www.youtube.com/watch?v=Sz13i-9EtmA. The raw fMRI data (not pre-processed) for all datasets are available at OpenNeuro at https://doi.org/10.18112/openneuro.ds002702.v1.0.1 ## **Updates** ## - 2020/06/20 - Updated the codearchive.zip with the latest versions of all of the code. - 2020/04/11 - Added Fig5additionalfigures.zip. - 2020/04/11 - The previous codearchive.zip file has been renamed to codearchive_oldversion.zip. There is a new version of the codearchive.zip file that corresponds to the upcoming new version of the paper. - 2020/03/30 - We have added a full set of pre-processed data as well as FreeSurfer output directories. - 2020/03/30 - We added a small exampledataset_inputsonly.mat file. - 2020/03/30 - The exampledataset.mat file had an incorrect version of the 'imglookup' and 'extrapmask' variables. Basically, the old version reflected a "low resolution" version of these variables. We have updated the exampledataset.mat file with the correct version. ## **FILES** ## - **codearchive.zip** is an archive of analysis scripts used for the paper. - **Fig5additionalfigures.zip** contains figures related to Figure 5, specifically, raw figure outputs for all hemispheres and all experimental conditions. ## **DATA FILES (example dataset)** ## This example dataset corresponds to Dataset D2 from the paper. To reduce data size, we include here only vertices within left hemisphere V1-V3 as determined from the Wang 2015 atlas; we include data from only Depth 2, Depth 4, and Depth 6; and we include only the first 4 fMRI runs (out of the 9 runs collected). The example dataset consists of two related files. The 'exampledataset.mat' file contains pre-processed data starting with the time-series data. The 'exampledataset_inputsonly.mat' file is derived from this earlier file (using example1.m) and contains only the inputs needed to call extracthrfmanifold.m. Contents of **exampledataset.mat**: - **'design'** is 1 x 4 cell vector with the design matrix for each of the 4 runs. Each element is 368 time points x 6 conditions. Each column is a binary vector indicating the onsets of a given condition (coded as 1s). - **'data'** is 1 x 4 cell vector with pre-processed fMRI time-series data for each of the 4 runs. Each element is 121,695 vertices x 368 time points. Note that pre-processing involved a temporal interpolation step (dealing with slice time differences and temporal upsampling) and a spatial interpolation step (dealing with motion correction, EPI spatial undistortion, and coregistration to anatomy). The units of the data are raw scanner units. Note also that the vertices actually break down into [40,565 vertices * 3 depths] reflecting the fact that data have been prepared onto vertices from 3 different cortical depths (Depth 2, Depth 4, Depth 6). - **'stimdur'** - a scalar with the stimulus duration (3.5 seconds) - **'tr'** - a scalar with the sampling rate after pre-processing (1 second) - **'bc'** - vertices x 1 with bias-corrected EPI intensity values. Values generally range in [0,2]. - **'imglookup'** - rows x columns providing integer indices into the data vertices. The value of each pixel indicates which vertex is the closest to that pixel. This can be used (in conjunction with *extrapmask*) to quickly generate surface visualizations. Note that the vertex indices "repeat" for each cortical depth. Also, note that we include only a crop of the full imglookup. - **'extrapmask'** - rows x columns indicating which pixels are invalid (e.g. outside the surface). Contents of **exampledataset_inputsonly.mat**: - **'data'** is vertices x 31 time points x 6 conditions with hemodynamic timecourses in units of percent signal change. The first time point is coincident with condition onset. - **'intensity'** is vertices x 1 with bias-corrected EPI intensity values. - **'tr'** - a scalar with the sampling rate ## **DATA FILES (full datasets)** ## There are a total of 16 datasets (D1-D16) used in the paper. The example dataset described above is just a small subset of one of the datasets (D2). Here, we provide pre-processed fMRI time-series data corresponding to all datasets from the paper. Note that unlike the small example dataset, we provide all vertices from both hemispheres, all 6 cortical depths, and all fMRI runs. Contents of **datasetDD/information.mat**: - **'design'** is a 1 x runs cell vector with the design matrix for each run. Each element is time points x conditions. Each column is a binary vector indicating the onsets of a given condition (coded as 1s). - **'stimdur'** - a scalar with the stimulus duration in seconds - **'tr'** - a scalar with the sampling rate (after pre-processing) - **'bc'** - vertices x 1 with bias-corrected EPI intensity values. Values generally range in [0,2]. Note that the vertices actually break down into [N vertices * 6 depths] reflecting the fact that data have been prepared onto vertices from 6 different cortical depths (Depth 1–6). - **'imglookup'** - 1 x 3 cell vector. The three elements correspond to 3 different views (gEVC (referring to a flattened patch in early visual cortex), gVTC (referring to a flattened patch in ventral temporal cortex (note that in this case, the right hemisphere appears on the left)), sphere-occip (referring to a occipital view of a sphere)). Each element is rows x columns providing integer indices into the data vertices. The value of each pixel indicates which vertex is the closest to that pixel. This can be used (in conjunction with *extrapmask*) to quickly generate surface visualizations. Note that the vertex indices "repeat" for each cortical depth. - **'extrapmask'** - 1 x 3 cell vector. The three elements correspond to 3 different views (just like 'imglookup'). Each element is rows x columns indicating which pixels are invalid (e.g. outside the surface). - **'numlh'** is the number of vertices in the left hemisphere - **'numrh'** is the number of vertices in the right hemisphere Contents of **'datasetDD/runRR.mat'**: - **'data'** is vertices x 6 depths x time points with the pre-processed fMRI time-series data. The format is int16 (to save disk space). Note that pre-processing involved a temporal interpolation step (dealing with slice time differences and temporal upsampling) and a spatial interpolation step (dealing with motion correction, EPI spatial undistortion, and coregistration to anatomy). The units of the data are raw scanner units. (Note that in the course of pre-processing, a few runs were omitted due to poor subject performance. Specifically, from Dataset D6 we omitted runs 4 and 6, and from Dataset D9 we omitted run 4. Thus, the number of runs in the pre-processed data does not necessarily match the number of runs in the raw acquired data.) Contents of **freesurfer/subjSS.tgz**: This is a gzipped tar of the FreeSurfer output directory for subjSS. Beyond the standard FreeSurfer outputs, specific files of interest are noted below: - **'surf/[lh,rh].[white,inflated,layerA1-A6,pial,sphere,sphere.reg]DENSETRUNCpt'** - These contain the dense and truncated cortical surface reconstructions. Layer A1-A6 refer to the six depth-varying surfaces from superficial (1) to deep (6). In the course of pre-processing, the fMRI data are interpolated onto the locations of the vertices in the layer A1-A6 surfaces. The number of vertices in 'datasetDD/runRR.mat' is equal to the number of vertices in these cortical surface files. - **'label/[lh,rh]DENSETRUNCpt.Kastner2015Labels.mgz'** - These files contain the Kastner 2015 atlas as prepared for the DENSETRUNCpt surfaces.