Skip to content

DLSurfer — curated deep-learning morphometry orchestrator

DLSurfer turns a T1w (± lesion mask) into a FreeSurfer-shaped SUBJECTS_DIR using validated FastSurfer / FreeSurfer DL tools. With a lesion mask and profile core-lesion, the same plan can inpaint T1w (neuroLIT) then run ordinary FastSurfer so surfaces and labels are estimated on a filled anatomy. It is study-agnostic: lesion steps stay off unless you choose a lesion profile and pass a mask.

Docs host: dlsurfer.readthedocs.io. Sibling connectomics pipeline: DKT Connectome (BIDS DWI → DKT connectome). DLSurfer is isolated morphometry orchestration — it does not import or modify the DKT workflow; handoff is optional print-only notes.

Linux + Apptainer (FreeSurfer license required). Release: v0.1.0.

Next: What DLSurfer does · Installation · Tutorial · Outputs

This package orchestrates FastSurfer, FreeSurfer, and neuroLIT (and optional specialty tools). Cite those papers — Methods / engines.

Quick start

License, SIFs, and dlsurfer doctor are on Installation. Then the Tutorial (dry-run + EG evidence).

source env/cluster_containers.sh
dlsurfer doctor
dlsurfer run --profile core-lesion \
  --t1 T1.nii.gz --lesion-mask lesion.nii.gz \
  --sd ./subjects --sid sub-01 --fs-license $FS_LICENSE
# add --execute when the plan looks right

HPC examples: hpc/sbatch_*.sh. Flags: Usage · full CLI: Command-line reference.

Short of a BIDS App

Unlike DKT Connectome (./run bids out participant), DLSurfer’s production API is still explicit paths. A stub exists — BIDS App status.

Contact