DU15NET Cortical Atlases

The DU15NET cortical parcellation atlases provide a15-network parcellation of the cerebral cortex derived from densely sampled fMRI data in 15 participants, each scanned 8–11 sessions. Usable in group-based studies as a significant update to the earlier atlases and in individual-based studies as a prior for model-based network mapping.

Seven atlas variants are available — Consensus (recommended primary atlas), Consensus-Raw, Prior, AgreeProb, Agree53, Agree80, Tripartite — in three spaces: FreeSurfer (fsaverage6/5/fsaverage), HCP (fsLR_32k / fsLR_164k, CIFTI), and MNI152 volume (1/2/3 mm).

Download DU15NET.zip

Precision Network Mapping

Precision network mapping pipeline is the pipeline that estimates individual-specific networks using a 15-network Multi-Session Hierarchical Bayesian Model (MS-HBM), operating on fsaverage6 surface data and using the pre-trained DU15NET group prior directly. This pipeline has been used to study network organization in the cerebral cortex (Du et al., 2024), cerebellum (Saadon-Grosman et al., 2024), striatum (Kosakowski et al., 2024), hippocampus (Angeli et al., 2025), and thalamus (Du et al., 2025), and a within-individual TMS planning study (Sun et al., 2025).

Precision fMRI Dataset

This dataset contains within-individual functional connectivity (FC) correlation matrices computed from densely sampled fMRI data in 11 participants (P2–P4, P6–P9, P12–P15), as reported in Du et al. (2025, Neuron) and Du et al. (2024, J Neurophysiol). Matrices are provided for each participant from both resting-state fixation (Fixation) and task-regressed (Task) data. For each participant, BOLD fMRI data were acquired over 8–11 sessions at the Harvard Center for Brain Science, using a 3T Prisma-fit scanner.

Du et al. (2025) Precision MRI dataset

Within-individual Network Estimates and Task Contrasts

The BALSA scene provides within-individual network parcellations, temporal signal-to-noise maps, and task contrast maps for all 15 participants. It includes individual network parcellations in CIFTI dscalar and border formats, tSNR maps, task contrast Z-maps for Motor, N-Back, Oddball, Visual, Sentence Processing, Theory-of-Mind, and Episodic Projection contrasts, and a Connectome Workbench scene file (`Harvard_Precision_Neuroimaging_Group.scene`). 

Task Protocols and Code

Harvard Dataverse: Detailed protocols and presentation code are available for eight task paradigms used in the precision fMRI session: Episodic Projection, Theory-of-Mind False Belief, Theory-of-Mind Pain, Sentence Processing, N-Back, Motor, Oddball, Visual.