Computer Vision News - December 2021

28 Introduction to Neuroscience Image Processing Blank (24 s) Stim. (15 s) Stim. (15 s) Blank (15 s) Blank (6 s) +----------+ +----------+ +----------+ +----------+ +----------+ | | | Intact | |Scrambled | | | | | | | | Object | | Face | | | ... (repeated 6 times) ... | | +----------+ +----------+ +----------+ +----------+ +----------+ • Stimulus presentation: different images in the category were flashed every 500 ms with blank. fMRI data format in this example The fMRI data are saved as BData , machine-learning analysis oriented data format developed in Kamitani Lab. bdpy is required to read the BData. bdata = bdpy.BData('data/sub-02_task-localizer_bold_preproc_native.h5') # Get fMRI data fmri_data = bdata.select('VoxelData') fmri_data.shape # volumes x voxels SothefMRIdataarecomposedof800volumesand32028voxels. # Get labels labels = bdata.get_label('block_type') np.unique(labels) The data have nine events. Three of them ('PreRest', 'InterRest', and 'PostRest') are 'rest' event in which no visual stimulus was presented. The rest event is not inc luded in the GLM model. In theother six events ('IntactFace', 'IntactObject', 'IntactScene', 'ScrambledFace', and 'ScrambledObject'), a visual stimulus was presented. These 'task' event should be included in the GLMmodel. The labels are used to create task regressors , which model the signal changes caused by the experimental conditions (e.g., stimuli). # Get run numbers runs = bdata.select('Run') runs.shape # samplex x 1 This is an array including run numbers, used for creating run regressors, which explicitly model run-wise fluctuations of the fMRI signals. Wrapping up! Keep looking at the answers and feel free to contact me if you have any questions till next month! Amazing work from Kamitani lab where tutorials and the bData package were written! Thanks a lot Take care for yourself and the people around you! See you next month. To be continued…