NeoRS: a neonatal resting state fMRI data preprocessing pipeline

Resting state fMRI (rsfMRI) has been shown to be a promising tool to study intrinsic functional connectivity and assess its integrity in cerebral development. In neonates, where fMRI is limited to few paradigms, rsfMRI was shown to be a relevant tool to explore regional interactions of brain networks. However, to identify the resting state networks, data needs to be carefully processed. Because of the non-collaborative nature of the neonates, the differences in brain size and the reversed contrast compared to adults, neonates can't be processed with the existing adult pipelines. Therefore, we developed NeoRS. The main processing steps include atlas registration, skull tripping, segmentation, slice timing and head motion correction and confounds regression. To address the specificity of neonatal brain imaging, particular attention was given to registration including neonatal atlas type and parameters, such as brain size variations, and contrast differences compared to adults. Furthermore, head motion was scrutinized and optimized, as it is a major issue when processing neonatal data. The pipeline includes visual quality control assessment checkpoints. To assess its effectiveness, we used the data from the Baby Connectome Project including 10 neonates. NeoRS was designed to work on both multi-band and single-band acquisitions and is applicable on smaller datasets. It also includes popular functional connectivity analysis features such as seed based correlations. Language, default mode, dorsal attention, visual, ventral attention, motor and fronto parietal networks were evaluated. The different analyzed networks were in agreement with previously published studies in the neonate. NeoRS is coded in Matlab, it is open-source and available on https://github.com/venguix/NeoRS. NeoRS allows robust image processing of the neonatal rsfMRI data that can be readily customized to different datasets.