Skip to content

Methods — resources & citations

Auto-generated from dlsurfer/engines.yaml (python scripts/gen_methods_docs.py). Cite these engines, not DLSurfer as morphometry science.

Pinned stack

Pin Value
fastsurfer_image deepmi/fastsurfer:cu132-v2.6.0
fastsurfer_sif_lab Documents/others/containers/fastsurfer_cu126_v2.6.0.sif
freesurfer_min 8.0.0
freesurfer_sif_lab Documents/others/containers/freesurfer_7.4.1.sif
neurolit_min 0.6.0
neurolit_sif_lab Documents/others/containers/lit_0.6.0.sif

Chosen engines

Engine CLI / resource Paper / docs
SynthStrip (synthstrip) · production mri_synthstrip Hoopes A, et al. SynthStrip: skull-stripping for any brain image. NeuroImage 2022.
https://synthstrip.io
DeepN4 (deepn4) · ready deepn4 / dipy_bias_field_correction --method n4 Kanakaraj S, et al. DeepN4: Learning N4ITK Bias Field Correction for T1-weighted Images. 2024. PMC10680935.
https://github.com/MASILab/DeepN4
FastSurferVINN (asegdkt) (fastsurfer_vinn) · production run_fastsurfer.sh (asegdkt module) Henschel L, et al. FastSurfer. NeuroImage 2020;219:117012.
Henschel L, Kuegler D, Reuter M. FastSurferVINN. NeuroImage 2022;251:118933.
SynthSeg (synthseg) · production mri_synthseg Billot B, et al. SynthSeg. Medical Image Analysis / FreeSurfer wiki SynthSeg.
https://surfer.nmr.mgh.harvard.edu/fswiki/SynthSeg
FastSurfer-CC (fastsurfer_cc) · production run_fastsurfer.sh (cc; disable with --no_cc) Pollak C, et al. FastSurfer-CC. Imaging Neuroscience 2026. doi:10.1162/IMAG.a.1221
CerebNet (cerebnet) · production run_fastsurfer.sh (cereb; --no_cereb to disable) Faber J, et al. CerebNet. NeuroImage 2022;264:119703.
HypVINN (hypvinn) · production run_fastsurfer.sh (hypothal; optional --t2) Estrada S, et al. FastSurfer-HypVINN. Imaging Neuroscience 2023.
neuroLIT / FastSurfer-LIT (neurolit) · production lit-inpainting ; lit-postprocessing ; or run_fastsurfer.sh --lesion_mask Pollak C, et al. FastSurfer-LIT. Imaging Neuroscience 2025. doi:10.1162/imag_a_00446
https://github.com/Deep-MI/neurolit
FastSurfer LIT docs: https://deep-mi.org/FastSurfer/stable/overview/modules/LIT.html
FastSurfer recon-surf (classical geometry) (recon_surf) · classical run_fastsurfer.sh (surf) / recon-surf.sh Henschel L, et al. FastSurfer. NeuroImage 2020 (recon-surf description).
Fischl B. FreeSurfer. NeuroImage 2012 (underlying surface methods).
SynthMorph (synthmorph) · production mri_synthmorph Hoffmann M, et al. SynthMorph. IEEE TMI 2022; Imaging Neuroscience 2024.
https://synthmorph.io
EasyReg (easyreg) · production mri_easyreg Iglesias JE, et al. EasyReg. Scientific Reports 2023. doi:10.1038/s41598-023-33781-0
https://surfer.nmr.mgh.harvard.edu/fswiki/EasyReg
FreeSurfer 8 recon-all (freesurfer8) · production recon-all FreeSurfer 8.0.0 release notes (SynthSeg, SynthStrip, SynthMorph in recon-all).
Fischl B. FreeSurfer. NeuroImage 2012.
TopoFit (topofit) · production mris_estimate_wm Hoopes A, et al. TopoFit. MIDL 2022.
https://github.com/ahoopes/topofit
V2C-Flow (Vox2Cortex-Flow) (v2c_flow) · ready vox2organ / V2C-Flow pretrained Bongratz F, et al. Vox2Cortex. CVPR 2022.
Bongratz F, et al. V2C-Flow. arXiv:2401.12938 (2024).
https://github.com/ai-med/Vox2Cortex
SUGAR (sugar) · ready SUGAR (see upstream README) Ren J, et al. SUGAR. arXiv:2307.00511 (2023).
https://github.com/pBFSLab/SUGAR
DL-CTE / CortexMorph family (dl_cte) · ready dl-cte / cortexmorph Rebsamen M, et al. DL+DiReCT. Hum Brain Mapp 2020.
CortexMorph arXiv:2307.11567 (2023).
Wu et al. DL-CTE 2025. https://github.com/wujiong-hub/DL-CTE
FreeSurfer longitudinal (-base / -long) (freesurfer_long) · production recon-all -base … ; recon-all -long … Reuter et al. NeuroImage 2012 (longitudinal FreeSurfer)
FreeSurfer Longitudinal Processing docs
FreeSurfer hippocampal/amygdala subfields (HA) (segment_ha) · production segmentHA_T1.sh <sid> <SUBJECTS_DIR> Iglesias et al. NeuroImage 2015 (hippocampal subfields)
Saygin et al. NeuroImage 2017 (amygdala)
FreeSurfer brainstem structures (segment_bs) · production segmentBS.sh <sid> <SUBJECTS_DIR> Iglesias et al. NeuroImage 2015 (brainstem)
FreeSurfer thalamic nuclei (segment_thalamic) · production segmentThalamicNuclei.sh <sid> <SUBJECTS_DIR> Iglesias et al. NeuroImage 2018 (thalamic nuclei)
recon-all-clinical (SynthSeg/SynthDist clinical) (recon_all_clinical) · production recon-all-clinical.sh … Billot SynthSeg; FreeSurfer recon-all-clinical docs
Destrieux (a2009s) parcellation (atlas_destrieux) · production recon-all -s <sid> -sd <SUBJECTS_DIR> -a2009s Destrieux et al. NeuroImage 2010
FreeSurfer aparc.a2009s docs
Schaefer surface projection (atlas_schaefer) · ready mri_surf2surf --srcsubject fsaverage --trgsubject <sid> --sval-annot … Schaefer et al. Cerebral Cortex 2018
FreeSurfer mri_surf2surf / label projection docs
FastSurfer srun / brun HPC wrap (srun_fastsurfer) · production srun_fastsurfer.sh … OR brun_fastsurfer.sh … FastSurfer HPC docs (brun_fastsurfer / srun_fastsurfer)
Mauri claustrum (mri_claustrum_seg) (mauri_claustrum) · production mri_claustrum_seg --i <T1w> --o <outdir> --threads N --topo-correct Mauri C, et al. Hum Brain Mapp. 2025;46:e70303. doi:10.1002/hbm.70303
https://github.com/chiara-mauri/claustrum_segmentation
SynthSeg-PC (Billot GMM → U-Net piriform) (synthseg_pc) · research python Documents/piriform/synthseg_pc/predict_ideas_one.py --images … --model … Billot B, et al. Nat Commun. 2023 (SynthSeg / GMM training)
Steinbart D, et al. Hum Brain Mapp. 2023 (PC labels 125/126)
Mauri et al. 2025 (dense FoV / real-T1 epoch pattern — not claustrum weights)
Hammers A, et al. Hum Brain Mapp. 2003
SynthSR (mri_synthsr) (synthsr) · production mri_synthsr --i <in> --o <out> Billot B, et al. SynthSR (FreeSurfer mri_synthsr)
Iglesias / Billot Synth* family docs
HippUnfold (hippunfold) · ready hippunfold <bids> <out> participant … DeKraker J, et al. eLife 2022;11:e77945

Per-engine detail

SynthStrip

  • Replaces (classical): FS watershed / hybrid skull-strip
  • CLI / resource: mri_synthstrip
  • Why chosen: Best Production skull-strip already in FreeSurfer (≥7.3). Contrast- and resolution-robust via synthetic training; removes a common classical failure mode without a separate container. Prefer over HD-BET inside DLSurfer so the strip utility stays license/tooling-aligned with FS8.
  • Cite:
  • Hoopes A, et al. SynthStrip: skull-stripping for any brain image. NeuroImage 2022.
  • https://synthstrip.io
  • Alternatives: hdbet (ready), robex (research), monai_brain_extract (research)

DeepN4

  • Replaces (classical): N3/N4 / NU bias-field correction
  • CLI / resource: deepn4 / dipy_bias_field_correction --method n4
  • Why chosen: Best Ready learned approximation of N4ITK for T1w (public weights; also in DIPY). Optional pre-step when FastSurfer/FS bias handling is insufficient or when a differentiable/portable N4 stand-in is needed outside FS. Not required for core FastSurfer profiles (VINN path has its own biasfield stats).
  • Cite:
  • Kanakaraj S, et al. DeepN4: Learning N4ITK Bias Field Correction for T1-weighted Images. 2024. PMC10680935.
  • https://github.com/MASILab/DeepN4
  • Alternatives: simko_implicit_bias (research), n4itk (classical)

FastSurferVINN (asegdkt)

  • Replaces (classical): CA aseg / classical volumetric labeling + DK-style cortex labels
  • CLI / resource: run_fastsurfer.sh (asegdkt module)
  • Why chosen: Primary whole-brain morphometry engine for DLSurfer. Resolution-independent (0.7–1.0 mm) DL segmentation into FreeSurfer-compatible DKT/aseg products in minutes on GPU. This is the core reason DLSurfer defaults to FastSurfer rather than classical recon-all for 'core' profiles. Atlas default = DKT; other atlases are orthogonal (see docs/ATLASES.md) — DLSurfer is not DKT-connectome.
  • Cite:
  • Henschel L, et al. FastSurfer. NeuroImage 2020;219:117012.
  • Henschel L, Kuegler D, Reuter M. FastSurferVINN. NeuroImage 2022;251:118933.
  • Alternatives: synthseg (production), quicknat (research), deepscan (research), assemblynet (research), slant (research)

SynthSeg

  • Replaces (classical): CA aseg; clinical / odd-contrast labeling
  • CLI / resource: mri_synthseg
  • Why chosen: Best Production alternative when T1 quality/contrast/resolution is non-standard, or when using profile fs8 / seg-quick-synthseg. Domain-randomized; ships in FreeSurfer. Not the default for research-grade isotropic T1 (VINN preferred).
  • Cite:
  • Billot B, et al. SynthSeg. Medical Image Analysis / FreeSurfer wiki SynthSeg.
  • https://surfer.nmr.mgh.harvard.edu/fswiki/SynthSeg
  • Alternatives: fastsurfer_vinn (production)

FastSurfer-CC

  • Replaces (classical): Classical / limited CC post-processing
  • CLI / resource: run_fastsurfer.sh (cc; disable with --no_cc)
  • Why chosen: Validated CC morphometry module in the default FastSurfer stack; included in 'core' so DLSurfer matches upstream FastSurfer defaults rather than stripping modules.
  • Cite:
  • Pollak C, et al. FastSurfer-CC. Imaging Neuroscience 2026. doi:10.1162/IMAG.a.1221
  • Alternatives: classical_cc (classical), skip_cc (production)

CerebNet

  • Replaces (classical): Limited classical cerebellum sub-segmentation
  • CLI / resource: run_fastsurfer.sh (cereb; --no_cereb to disable)
  • Why chosen: Production cerebellum sub-segmentation in FastSurfer; kept on in 'core' for parity with upstream morphometry defaults.
  • Cite:
  • Faber J, et al. CerebNet. NeuroImage 2022;264:119703.
  • Alternatives: classical_cereb (classical), skip_cereb (production)

HypVINN

  • Replaces (classical): Sparse classical hypothalamic labeling
  • CLI / resource: run_fastsurfer.sh (hypothal; optional --t2)
  • Why chosen: Production hypothalamus + adjacent structure module. Preferred over ScLimbic when the FastSurfer engine is selected (XOR with FS-only ScLimbic path).
  • Cite:
  • Estrada S, et al. FastSurfer-HypVINN. Imaging Neuroscience 2023.
  • Alternatives: sclimbic (production), skip_hypothal (production)

neuroLIT / FastSurfer-LIT

  • Replaces (classical): No native classical lesion fill in FS/FastSurfer
  • CLI / resource: lit-inpainting ; lit-postprocessing ; or run_fastsurfer.sh --lesion_mask
  • Why chosen: Best Production lesion inpainting (DDPM). Standalone lit-inpainting is the DEFAULT lesion path in DLSurfer because native --lesion_mask cannot use FastSurfer --seg_only/--surf_only (breaks GPU/CPU HPC split). lit-postprocessing restores LIT packaging after ordinary recon. Native --lesion_mask remains an explicit opt-in profile.
  • Cite:
  • Pollak C, et al. FastSurfer-LIT. Imaging Neuroscience 2025. doi:10.1162/imag_a_00446
  • https://github.com/Deep-MI/neurolit
  • FastSurfer LIT docs: https://deep-mi.org/FastSurfer/stable/overview/modules/LIT.html
  • Alternatives: neurolit_native_flag (production), vbg (ready), leapp_vbt (ready), enantiomorphic (ready)

FastSurfer recon-surf (classical geometry)

  • Replaces (classical): FS surface stages (tessellate→topology→white/pial→sphere→thickness)
  • CLI / resource: run_fastsurfer.sh (surf) / recon-surf.sh
  • Why chosen: Still the Production surface path for SUBJECTS_DIR / QSIRecon compatibility. Not deep learning (spectral embedding + FreeSurfer binaries). Kept as default surfaces backend until Phase-3 surf-dl backends pass the TBI±LIT gate. This is intentional: DLSurfer does not fake a DL surface drop-in before validation.
  • Cite:
  • Henschel L, et al. FastSurfer. NeuroImage 2020 (recon-surf description).
  • Fischl B. FreeSurfer. NeuroImage 2012 (underlying surface methods).
  • Alternatives: topofit (production), v2c_flow (ready), cortexode (research), pialnn (research), corticalflow (research), surfnn (research), deepcsr (research), fastcsr (research), recon_all_clinical (production)

SynthMorph

  • Replaces (classical): Talairach / CA linear+nonlinear volumetric registration
  • CLI / resource: mri_synthmorph
  • Why chosen: Best Production acquisition-agnostic DL registration in FreeSurfer. Used as a utility profile for multimodal alignment / MNI mapping — not required inside FastSurfer 'core' (VINN path), but exposed so classical CA register is not reintroduced by lab scripts.
  • Cite:
  • Hoffmann M, et al. SynthMorph. IEEE TMI 2022; Imaging Neuroscience 2024.
  • https://synthmorph.io
  • Alternatives: easyreg (production), voxelmorph (research), ants (classical)

EasyReg

  • Replaces (classical): Classical symmetric diffeomorphic volume registration
  • CLI / resource: mri_easyreg
  • Why chosen: Best Production symmetric, invertible, modality-agnostic registration CLI in FreeSurfer 7.4+. Prefer when directionality/symmetry matters; companion to SynthMorph for utility registration tasks.
  • Cite:
  • Iglesias JE, et al. EasyReg. Scientific Reports 2023. doi:10.1038/s41598-023-33781-0
  • https://surfer.nmr.mgh.harvard.edu/fswiki/EasyReg
  • Alternatives: synthmorph (production), voxelmorph (research)

FreeSurfer 8 recon-all

  • Replaces (classical): FS≤7 classical early stages (embeds SynthSeg/Strip/Morph)
  • CLI / resource: recon-all
  • Why chosen: Alternate morphometry engine (XOR with FastSurfer). FS8 already embeds DL for strip/seg/reg while keeping classical surfaces. Offered as profile 'fs8' for sites standardized on FreeSurfer outputs or longitudinal FS workflows.
  • Cite:
  • FreeSurfer 8.0.0 release notes (SynthSeg, SynthStrip, SynthMorph in recon-all).
  • Fischl B. FreeSurfer. NeuroImage 2012.
  • Alternatives: fastsurfer_core (production), recon_all_clinical (production)

TopoFit

  • Replaces (classical): Topology fixer + white-surface placement (partial)
  • CLI / resource: mris_estimate_wm
  • Why chosen: Best Production DL white-surface fit already shipped in FreeSurfer (mris_estimate_wm). First Phase-3 surf-dl building block: lowest integration risk. Does not alone replace pial/thickness/aparc — pair with recon-surf pial or V2C-Flow for full surfaces.
  • Cite:
  • Hoopes A, et al. TopoFit. MIDL 2022.
  • https://github.com/ahoopes/topofit
  • Alternatives: v2c_flow (ready), cortexode (research), recon_surf (classical)

V2C-Flow (Vox2Cortex-Flow)

  • Replaces (classical): White+pial surfaces; often sphere.reg via template IDs
  • CLI / resource: vox2organ / V2C-Flow pretrained
  • Why chosen: Best Ready full explicit CSR: joint WM+pial, <2s, fsaverage-compatible vertex correspondence. Primary candidate to replace recon-surf for speed once TBI±LIT validation passes. Not SUBJECTS_DIR drop-in — adapter must map meshes into FS tree / analysis convention (scanner RAS).
  • Cite:
  • Bongratz F, et al. Vox2Cortex. CVPR 2022.
  • Bongratz F, et al. V2C-Flow. arXiv:2401.12938 (2024).
  • https://github.com/ai-med/Vox2Cortex
  • Alternatives: vox2cortex (ready), cortexode (research), pialnn (research), surfnn (research), topofit (production)

SUGAR

  • Replaces (classical): Spherical registration to fsaverage (sphere.reg)
  • CLI / resource: SUGAR (see upstream README)
  • Why chosen: Best Ready dedicated spherical surface registration (large multi-dataset eval, sub-second). Used when recon-surf or TopoFit meshes exist but classical sphere.reg is the measured bottleneck — or when V2C-Flow correspondences are not used.
  • Cite:
  • Ren J, et al. SUGAR. arXiv:2307.00511 (2023).
  • https://github.com/pBFSLab/SUGAR
  • Alternatives: s3reg (ready), v2c_flow_ids (ready), classical_sphere_reg (classical)

DL-CTE / CortexMorph family

  • Replaces (classical): Surface-based cortical thickness (when surfaces not required)
  • CLI / resource: dl-cte / cortexmorph
  • Why chosen: Best Ready fast thickness without full FreeSurfer surfaces (CNN seg + diffeomorphic DiReCT-style registration). Profile 'thickness-fast' only — NOT a substitute for surface morphometry or connectome parcels.
  • Cite:
  • Rebsamen M, et al. DL+DiReCT. Hum Brain Mapp 2020.
  • CortexMorph arXiv:2307.11567 (2023).
  • Wu et al. DL-CTE 2025. https://github.com/wujiong-hub/DL-CTE
  • Alternatives: dl_direct (ready), cortexmorph (ready), surface_thickness (classical)

FreeSurfer longitudinal (-base / -long)

  • Replaces (classical): Cross-sectional-only morphometry
  • CLI / resource: recon-all -base … ; recon-all -long …
  • Why chosen: Best Production alternative for the longitudinal gap. FreeSurfer -base/-long remains the field standard; FastSurfer longitudinal support is limited.
  • Cite:
  • Reuter et al. NeuroImage 2012 (longitudinal FreeSurfer)
  • FreeSurfer Longitudinal Processing docs
  • Alternatives: synthseg_per_tp (production), v2c_long (early)

FreeSurfer hippocampal/amygdala subfields (HA)

  • Replaces (classical): Whole-hippocampus aseg only
  • CLI / resource: segmentHA_T1.sh <sid> <SUBJECTS_DIR>
  • Why chosen: Best Production alternative for the HA gap — segmentHA_T1 / recon-all -hippocampal-subfields-T1 on an existing SUBJECTS_DIR.
  • Cite:
  • Iglesias et al. NeuroImage 2015 (hippocampal subfields)
  • Saygin et al. NeuroImage 2017 (amygdala)
  • Alternatives: hippunfold (ready), hippodeep (research)

FreeSurfer brainstem structures

  • Replaces (classical): Whole brainstem aseg label
  • CLI / resource: segmentBS.sh <sid> <SUBJECTS_DIR>
  • Why chosen: Best Production alternative for brainstem nuclei gap (segmentBS.sh).
  • Cite:
  • Iglesias et al. NeuroImage 2015 (brainstem)
  • Alternatives: none_dl_default (none)

FreeSurfer thalamic nuclei

  • Replaces (classical): Whole thalamus aseg label
  • CLI / resource: segmentThalamicNuclei.sh <sid> <SUBJECTS_DIR>
  • Why chosen: Best Production alternative for thalamic nuclei gap (segmentThalamicNuclei.sh).
  • Cite:
  • Iglesias et al. NeuroImage 2018 (thalamic nuclei)
  • Alternatives: none_dl_default (none)

recon-all-clinical (SynthSeg/SynthDist clinical)

  • Replaces (classical): Standard recon-all on clinical / low-res / odd-contrast
  • CLI / resource: recon-all-clinical.sh …
  • Why chosen: Best Production alternative for infant/clinical gap when standard recon-all or FastSurferVINN is inappropriate (clinical stream in FS7.3+/FS8).
  • Cite:
  • Billot SynthSeg; FreeSurfer recon-all-clinical docs
  • Alternatives: synthseg (production), infant_freesurfer (production)

Destrieux (a2009s) parcellation

  • Replaces (classical): DKT-only cortical labels
  • CLI / resource: recon-all -s <sid> -sd <SUBJECTS_DIR> -a2009s
  • Why chosen: Best alternative for Destrieux gap — classical FreeSurfer -a2009s on existing surfaces (sphere.reg required). Not VINN-native.
  • Cite:
  • Destrieux et al. NeuroImage 2010
  • FreeSurfer aparc.a2009s docs
  • Alternatives: mris_ca_label_manual (classical)

Schaefer surface projection

  • Replaces (classical): Anatomical-only cortical atlases
  • CLI / resource: mri_surf2surf --srcsubject fsaverage --trgsubject <sid> --sval-annot …
  • Why chosen: Best Ready alternative for Schaefer/MMP gap — project fsaverage annot after sphere.reg (classical FS surf2surf). Annot source is external.
  • Cite:
  • Schaefer et al. Cerebral Cortex 2018
  • FreeSurfer mri_surf2surf / label projection docs
  • Alternatives: hcp_mmp (ready), yeo_networks (ready)

FastSurfer srun / brun HPC wrap

  • Replaces (classical): Manual sbatch of run_fastsurfer.sh
  • CLI / resource: srun_fastsurfer.sh … OR brun_fastsurfer.sh …
  • Why chosen: Best Production alternative for the srun gap — use upstream srun_fastsurfer.sh / brun_fastsurfer.sh (enabled via --hpc srun).
  • Cite:
  • FastSurfer HPC docs (brun_fastsurfer / srun_fastsurfer)
  • Alternatives: slurm_template (production)

Mauri claustrum (mri_claustrum_seg)

  • Replaces (classical): No claustrum in standard aseg / FastSurferVINN
  • CLI / resource: mri_claustrum_seg --i <T1w> --o <outdir> --threads N --topo-correct
  • Lab path: Documents/clau_seg
  • Why chosen: Best Production DL tool for claustrum — FreeSurfer-integrated contrast-agnostic U-Net (Mauri 2025). Lab path Documents/clau_seg; put0 FoV lock for IDEAS. Not a piriform / PC model — do not reuse weights for Steinbart 125/126.
  • Cite:
  • Mauri C, et al. Hum Brain Mapp. 2025;46:e70303. doi:10.1002/hbm.70303
  • https://github.com/chiara-mauri/claustrum_segmentation
  • Alternatives: albishri_cnn (research), casamitjana_bayesian (research)

SynthSeg-PC (Billot GMM → U-Net piriform)

  • Replaces (classical): MAPER / JLF piriform (Steinbart 125/126) when DL gate passes
  • CLI / resource: python Documents/piriform/synthseg_pc/predict_ideas_one.py --images … --model …
  • Lab path: Documents/piriform/synthseg_pc
  • Why chosen: Lab SynthSeg-style PC segmenter (Documents/piriform/synthseg_pc): GMM on-the-fly train → U-Net predict on Hammers dense FoV. NOT FreeSurfer mri_synthseg (no PC class). NOT Mauri claustrum weights. Experimental until Hammers/IDEAS eval clears JLF bar; IDEAS production PC remains JLF until that gate.
  • Cite:
  • Billot B, et al. Nat Commun. 2023 (SynthSeg / GMM training)
  • Steinbart D, et al. Hum Brain Mapp. 2023 (PC labels 125/126)
  • Mauri et al. 2025 (dense FoV / real-T1 epoch pattern — not claustrum weights)
  • Hammers A, et al. Hum Brain Mapp. 2003
  • Alternatives: jlf_pc (production), maper_pc (production), nnunet_pc (research)

SynthSR (mri_synthsr)

  • Replaces (classical): Manual clinical T1 upsampling / contrast salvage
  • CLI / resource: mri_synthsr --i <in> --o <out>
  • Why chosen: Optional T1 preprocess only — 1 mm T1-like from poor clinical scans (Billot). Does NOT segment piriform or claustrum. Feed output into core / structure-* profiles. Matches Documents/piriform/synthseg_pc/maybe_synthsr.sh policy.
  • Cite:
  • Billot B, et al. SynthSR (FreeSurfer mri_synthsr)
  • Iglesias / Billot Synth* family docs
  • Alternatives: skip_synthsr (production), recon_all_clinical (production)

HippUnfold

  • Replaces (classical): FS segmentHA volumetric subfields (surface-unfold alternative)
  • CLI / resource: hippunfold <bids> <out> participant …
  • Lab path: Documents/HippUnfold
  • Why chosen: Best Ready DL alternative to segmentHA for hippocampal sheet morphometry. Optional profile; default HA gap-fill remains classical segmentHA_T1.
  • Cite:
  • DeKraker J, et al. eLife 2022;11:e77945
  • Alternatives: segment_ha (production), hippodeep (research)

Deferred (not defaults)

  • hdbet — Credible Ready stripper, but SynthStrip is already in FS and preferred for one-license tooling.
  • vbg — Ready fill, but Deutscher 2025 shows VBG can worsen FastSurfer on ms-TBI; TBI policy required first.
  • leapp_vbt — Stroke-oriented; lab VBT orientation SIGSEGV on radiological TBI T1s.
  • cortexode — Strong Research CSR; V2C-Flow chosen first as Ready pretrained full WM+pial.
  • pialnn — WM→pial only; consider after TopoFit if V2C-Flow fails TBI gate.
  • s3reg — Ready spherical reg; SUGAR chosen first for multi-dataset eval + speed claims.
  • quicknat — Credible seg; VINN/SynthSeg preferred for FS-shaped trees.
  • deepscan — Strong hybrid-pipeline seg; keep as documented alternative to VINN.

  • beyondfs_zoo — SLANT/AssemblyNet/GIF/FIRST/etc. belong in piriform/beyondfs identity (structures FS does not give) — do not dilute DLSurfer core with a methods zoo.

  • mauri_weights_as_pc — Never cite or run Mauri claustrum weights as Steinbart piriform — wrong structure.