`synthseg` — standalone SynthSeg execution#

Schema: SynthSegConfig in src/thesis/core/config/validators.py. Optional top-level block (defaults to None). Configures the synthseg workflow and the embedded SynthSeg stage of tract_synthseg.

Field	Type	Default	Constraints	Description
`t1_image`	`Path \| None`	`None`	—	Explicit T1 input path. When unset, the workflow falls back to `hcp.t1_image`. Processed through `to_path()` (supports `~` and `$VAR`).
`parc`	`bool`	`False`	—	Request cortical parcellation output (`--parc`).
`robust`	`bool`	`False`	—	Enable robust mode (`--robust`).
`fast`	`bool`	`False`	—	Enable fast mode (`--fast`).
`vol`	`bool`	`False`	—	Write volumes CSV (`--vol`).
`qc`	`bool`	`False`	—	Write QC CSV (`--qc`).
`crop`	`int \| None`	`None`	`≥ 1` if set	Optional crop size (`--crop N`).
`cpu`	`bool`	`False`	—	Force CPU execution (`--cpu`).
`threads`	`int`	`1`	`≥ 1`	CPU thread count when `cpu: true`.

Example#

synthseg:
  t1_image: data/raw/114823/T1w/T1w_acpc_dc_restore_1.25.nii.gz
  parc: true
  robust: false
  fast: false
  vol: true
  qc: true
  cpu: false
  threads: 1

Notes#

mri_synthseg must be on PATH (install separately from FreeSurfer/SynthSeg distributions — see docs/guides/installation.md).
The QC output is consumed downstream by qc.synthseg_qc_threshold to flag low-quality segmentations in batch summaries.
Inside the preprocess workflow, SynthSeg has its own config block under preprocess.synthseg — that’s a separate PreprocessSynthSegConfig model.