Fig. 1

Paradigm & workflow. Panels a–d indicate the workflow for the data available in the Monash rsPET-MR dataset²⁷ and panels e-g indicate the workflow for the results presented under Technical Validation, and in Jamadar et al.²⁰. (a) Participants completed a demographics, safety screening and cognitive assessment an hour prior to MR-PET scanning. (b) Next, participants were prepared for scanning; a cannula was placed in the forearm vein of each arm, and then haemoglobin and blood sugar level was taken. (c) Participants then underwent a 95-minute MR-PET scan using the paradigm shown here. (d) Illustration of the data obtained for each method (left to right): Structural T1 MRI anatomical images for each subject (subj); static PET (sPET) acquired a single image per subject; functional PET (fPET) was binned into 16 sec images, resulting in a timeseries of images for each subject; fMR images were obtained with TR 2.45 sec, resulting in a timeseries for each subject. (e) Structural T1 MRI was registered to MNI space and then segmented into 82 regions of interest (ROI). This parcellation was applied to the sPET, fPET and fMRI images. (f) Illustration of example processing steps for each modality, as used in Jamadar et al. (2020). sPET was demeaned, parcellated into 82 regions (ROIs) and then correlated across subjects. Scatterplot shows an example correlation between two regions across subjects, out of a total of 82 × 82 region-wise correlations. fPET was motion corrected (moco), filtered, parcellated, and then correlated across time-series for each subject. Illustration shows 2 example timeseries of the total 82 × 82 region-by-region correlations conducted. fMRI was preprocessed, parcellated, correlated across time for each subject, then group-averaged. (g) Subject-level matrices for sPET, fPET and fMRI were then group averaged. Matrices are indicative of potential connectivity matrices, and are those that are reported in Jamadar et al.²⁰.