Synthseg

synthseg - Neural Network-Based Brain MRI Segmentation

Part of the micaflow processing pipeline for neuroimaging data.

This module provides an interface to SynthSeg, a deep learning-based tool for automated brain MRI segmentation that works across different MRI contrasts without retraining. SynthSeg segments brain anatomical structures in T1w, T2w, FLAIR, and other MR contrasts, providing both whole-brain segmentation and optional cortical parcellation.

Features:

  • Contrast-agnostic segmentation working across different MRI acquisition types

  • Whole-brain anatomical structure segmentation with 37 labels

  • Optional cortical parcellation (up to 95 additional regions)

  • Multiple execution modes: standard, robust (higher quality), and fast

  • Volumetric analysis with CSV output for region-wise measurements

  • Quality control metrics for assessing segmentation reliability

  • GPU acceleration with optional CPU-only execution

API Usage:

micaflow synthseg

–i <path/to/image.nii.gz> –o <path/to/segmentation.nii.gz> [–parc] [–robust] [–fast] [–vol <path/to/volumes.csv>] [–qc <path/to/qc_scores.csv>] [–threads <num_threads>]

Python Usage:

>>> from micaflow.scripts.synthseg import main
>>> main({
...     'i': 'input_image.nii.gz',
...     'o': 'segmentation.nii.gz',
...     'parc': True,
...     'robust': False,
...     'fast': True,
...     'vol': 'volumes.csv',
...     'threads': 4
... })

Command Line Usage

micaflow synthseg [options]

Source Code

View the source code: GitHub Repository

Description

This script runs the SynthSeg neural network-based tool for brain MRI

segmentation. It provides automated segmentation of anatomical structures even across different contrasts and acquisition types.

Full Help

╔════════════════════════════════════════════════════════════════╗
║                         SYNTHSEG                               ║
╚════════════════════════════════════════════════════════════════╝

This script runs the SynthSeg neural network-based tool for brain MRI
segmentation. It provides automated segmentation of anatomical structures
even across different contrasts and acquisition types.

────────────────────────── USAGE ──────────────────────────
  micaflow synthseg [options]

─────────────────── REQUIRED ARGUMENTS ───────────────────
  --i PATH       : Input image(s) to segment (file or folder)
  --o PATH       : Output segmentation file(s) or folder

─────────────────── OPTIONAL ARGUMENTS ───────────────────
  --parc         : Enable cortical parcellation
  --robust       : Use robust mode (slower but better quality)
  --fast         : Faster processing (less postprocessing)
  --threads N    : Set number of CPU threads (default: 1)
  --cpu          : Force CPU processing (instead of GPU)
  --vol PATH     : Output volumetric CSV file
  --qc PATH      : Output quality control scores CSV file
  --post PATH    : Output posterior probability maps
  --resample PATH: Output resampled images
  --crop N [N ...]: Size of 3D patches to analyze (default: 192)
  --ct           : Clip intensities for CT scans [0,80]
  --v1           : Use SynthSeg 1.0 instead of 2.0

────────────────── EXAMPLE USAGE ────────────────────────

# Basic segmentation
micaflow synthseg \
  --i t1w_scan.nii.gz \
  --o segmentation.nii.gz

# With cortical parcellation
micaflow synthseg \
  --i t1w_scan.nii.gz \
  --o segmentation.nii.gz \
  --parc

# Batch processing with volume calculation
micaflow synthseg \
  --i input_folder/ \
  --o output_folder/ \
  --vol volumes.csv

────────────────────────── NOTES ───────────────────────
• SynthSeg works with any MRI contrast without retraining
• GPU acceleration is used by default for faster processing
• The robust mode provides better quality but is slower
• For batch processing, input and output paths must be folders