Fig. 1: A computationally efficient and scalable neuroimaging pipeline is empowered by deep learning algorithms and the workflow manager. | Nature Methods

Fig. 1: A computationally efficient and scalable neuroimaging pipeline is empowered by deep learning algorithms and the workflow manager.

From: DeepPrep: an accelerated, scalable and robust pipeline for neuroimaging preprocessing empowered by deep learning

Fig. 1: A computationally efficient and scalable neuroimaging pipeline is empowered by deep learning algorithms and the workflow manager.

a, DeepPrep workflow. DeepPrep accepts both anatomical and fMRI data for preprocessing. White boxes highlight deep learning algorithms for brain tissue segmentation, CSR, cortical surface registration and volumetric spatial normalization, which replace the most time-intensive modules in conventional pipelines. T1w, T1 weighted. BOLD, blood oxygenation level dependent. b, The preprocessing pipeline is organized into multiple relatively independent yet integrated processes. Imaging data adhering to the BIDS format are preprocessed through a structured workflow managed by Nextflow14. Nextflow schedules task processes and allocates computational resources across diverse infrastructures, including local computers, HPC clusters and cloud computing environments. The pipeline yields standardized preprocessed imaging and derivative files, quality control reports and runtime reports as its outputs.

Back to article page