fw-gear-fast-surfer 1.0.7

Wrapper for Fast Surfer

FastSurfer

Overview

• This GEAR wraps FastSurferCNN part of FastSurfer.

Summary

• FastSurferCNN is an advanced deep learning architecture capable of whole brain segmentation into 95 classes in under 1 minute, mimicking FreeSurfer anatomical segmentation and cortical parcellation (DKTatlas).

Cite

Henschel L, Conjeti S, Estrada S, Diers K, Fischl B, Reuter M, FastSurfer - A fast and accurate deep learning based neuroimaging pipeline, NeuroImage 219 (2020), 117012

License: MIT

Classification

Category: Analysis Gear

Gear Level:

• Project
• Subject
• Session
• Acquisition
• Analysis

---

[[TOC]]

---

Inputs

• 3DT1-Weighted Volume
  • Name: T1w
  • Type: file
  • Optional: false
  • Classification: file
  • Description: Input 3D-T1w nifti in zipped format
  • Notes: {Gear accepts one zipped nifti file}

Config

• {Config-Option}
  • Name: use-gpu
  • Type: boolean
  • Description: Use GPU for computing the FastSurfer segmentation. Gear will default to CPU if no GPU is available.
  • Default: true

Outputs

Files

• {Output-File}

  • Name: aparc+aseg.nii.gz
  • Type: file
  • Optional: false
  • Classification: file
  • Description: Segmentation output with 95 labels
  • Notes: The segmentation is based on the DKT atlas

• {Output-File}

  • Name: deep-seg.log
  • Type: file
  • Optional: false
  • Classification: file
  • Description: Log file containing all processing details

• {Output-File}

  • Name: 001.mgz
  • Type: file
  • Optional: false
  • Classification: file
  • Description: intermediate file generated during inference

• {Output-File}

  • Name: orig.mgz
  • Type: file
  • Optional: false
  • Classification: file
  • Description: intermediate file generated during inference

Usage

Description

• This gears operates by using the Eval script from FastSurferCNN.
• The input is a 3D-T1w MRI exam in zipped nifti format.
• The output is saved to the Analysis section in the Flywheel platform.

File Specifications

This Gear has only 1 required input file. The details are below:

{Input-File}

A 3D-T1w brain MRI exam is required for this gear. Ideal acquistions are 1mm isotropic. The Gear can handle non-isotropic data such as 0.9mmx0.9mmx1.0mm.

Workflow

graph LR;
    A[Input-File]:::input --> C;
    C[Upload] --> D[Parent Container <br> Project, Subject, etc];
    D:::container --> E((Gear));
    E:::gear --> F[Analysis]:::container;
    
    classDef container fill:#57d,color:#fff
    classDef input fill:#7a9,color:#fff
    classDef gear fill:#659,color:#fff

Description of workflow

1. Upload file to container
2. Select file as input to gear
3. Gear places output in Analysis. All output files can be downloaded from Analysis.
4. The output files are available under "Outputs" section in job log. The aparc+aseg.nii.gz can be seen with OHIF viewer.

Use Cases

Use Case 1

*Conditions*:

[x] Collection of 3D-T1w brain MRI exams for which a fast Freesurfer-style segmentation is needed. [x] For each timepoint, the data is available as a zipped nifti file.

Steps

1. Upload the zipped nifti files to your Flywheel instance a. Follow these steps to upload de-identified data to Flywheel

2. Select the subjects' sessions that are to be analyzed

   a.

3. Navigate to "Run Gear", select "Analysis Gear", and select Fast Surfer under the "Image Processing - Segmentation" section. a.

4. Select the zipped nifti file to be segmented.

5. Run gear and review the results under Analysis or in the job log output section.

Logging

• Logging will indicate if the processing was run on a GPU or CPU.
• Three models are run to generate the final output.
• The processing time for each model will be displayed in the log.

FAQ

FAQ.md

Contributing

[For more information about how to get started contributing to that gear, checkout CONTRIBUTING.md.]