Slides from the FMRIPREP & MRIQC Focus group held Jan 13, 2017 at Stanford University. MRIQC provides a series of image processing workflows to extract and compute a series of NR (no-reference), IQMs (image quality metrics) to be used in QAPs (quality assessment protocols) for MRI (magnetic resonance imaging). http://mriqc.readthedocs.org

1. Stanford University
Focus group: MRIQC
Quality Control of structural and functional MRI
Oscar Esteban <oesteban@stanford.edu>
Poldrack Lab, Stanford University
January 13th
, 2017

2. Stanford University
“Have no fear of perfection – you’ll never reach it.”
Salvador Dalí
1/28

3. Stanford University
2/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
But sometimes, we are way too far from perfection
(MRIQC mosaic, courtesy of Joke Durnez)

4. Stanford University
3/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Habitual suspects in structural MRI: motion
http://mriquestions.com/choosing-pefe-direction.html
https://www.cis.rit.edu/htbooks/mri/chap-11/k6-12.htm

5. Stanford University
4/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Habitual suspects in structural MRI: other
Intensity Non-Uniformity (INU) Noise

6. Stanford University
5/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
And expect the unexpected

7. Stanford University
6/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Artifacts in functional MRI
N/2 Ghost Spikes

8. Stanford University
7/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Manual assessment on the ABIDE dataset (N=1102)

9. Stanford University
8/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Close-up

10. Stanford University
9/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Manual assessment on the ABIDE dataset (N=1102)
- Time consuming
- Intra-rater bias
- Inter-rater bias
- Rater 1: 15% reject

11. Stanford University
10/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Objectives of Quality Control
Exclusion criteria – as objective as possible.
Quality Badge – Deciding on using a public dataset (is it appropriate for my
design/study?)
Diagnosing fixable problems with data acquisition process:
Types of sequences
Scanner malfunctions
Head padding
Participant instructions

12. Stanford University
11/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Image Quality Metrics (IQMs)
Physical phantoms (Price et al., 1990)
No-reference Image Quality Metrics (IQMs) (Woodard and Carley-Spencer,
2006)
Aim at artifacts and analyze noise distribution (Mortamet et al., 2009)
Combined general volumetric and artifact-targeted IQMs (Pizarro et al., 2016)

13. Stanford University
11/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Image Quality Metrics (IQMs)
Physical phantoms (Price et al., 1990)
No-reference Image Quality Metrics (IQMs) (Woodard and Carley-Spencer,
2006)
Aim at artifacts and analyze noise distribution (Mortamet et al., 2009)
Combined general volumetric and artifact-targeted IQMs (Pizarro et al., 2016)

14. Stanford University
11/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Image Quality Metrics (IQMs)
Physical phantoms (Price et al., 1990)
No-reference Image Quality Metrics (IQMs) (Woodard and Carley-Spencer,
2006)
Aim at artifacts and analyze noise distribution (Mortamet et al., 2009)
Combined general volumetric and artifact-targeted IQMs (Pizarro et al., 2016)

15. Stanford University
11/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Image Quality Metrics (IQMs)
Physical phantoms (Price et al., 1990)
No-reference Image Quality Metrics (IQMs) (Woodard and Carley-Spencer,
2006)
Aim at artifacts and analyze noise distribution (Mortamet et al., 2009)
Combined general volumetric and artifact-targeted IQMs (Pizarro et al., 2016)

16. Stanford University
12/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
IQMs: Structural MRI
Noise measurement
Signal-to-noise ratio (SNR) - higher is better
Contrast-to-noise ration (CNR) - higher is better
Sharpness (full-width half maximum estimations) - smaller FWHM is
better
Goodness of fit of a noise model into the noise in the background (QI2) -
lower is better (Mortamet et al., 2009)
Coefficient of Joint Variation (CJV) - lower is better
Information theory
Foreground-Background Energy Ratio (FBER) - higher is better
Entropy Focus Criterion (EFC) - lower is better
Artifacts
Segmentation using mathematical morphology (QI1) - lower is better
Measurements on the estimated INU (intensity non-uniformity) - values
around 1.0
Partial Volume Errors (PVE) - lower is better
Other: summary statistics, intracranial volume fractions (ICV)

17. Stanford University
13/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
IQMs: Functional MRI
Noise measurement: SNR, tSNR, temporal standard deviation
Information theory: EFC, FBER
Confounds and artifacts:
Framewise Displacement (FD) - lower is better
(Standardized) DVARS (D referring to temporal derivative of timecourses,
VARS referring to RMS variance over voxels) - lower is better
Ghost-to-Signal ratio (GSR) - lower is better
Global correlation (GCOR) - lower is better
Spikes (high frequency and global intensity)
AFNI’s outlier detection and quality indexes

18. Stanford University
14/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Design of MRIQC
Inputs: BIDS (Gorgolewski et al.,
2016b)
Command-line interface: BIDS-Apps
(Gorgolewski et al., 2016a)
The simplest possible pipeline
The fastest possible pipeline
Robust: works on all data

19. Stanford University
15/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
MRIQC Features
What can be expected from MRIQC:
A table of IQMs per subject
The group visual report
An individual visual report per subject
A first-round exercise for the data
What is not expected from MRIQC:
The triage of participants (WIP)
The derivatives of processing
Non-standard morphologies: developing brains, pathology, etc.

20. Stanford University
15/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
MRIQC Features
What can be expected from MRIQC:
A table of IQMs per subject
The group visual report
An individual visual report per subject
A first-round exercise for the data
What is not expected from MRIQC:
The triage of participants (WIP)
The derivatives of processing
Non-standard morphologies: developing brains, pathology, etc.

21. Stanford University
16/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Group Reports

22. Stanford University
17/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Anatomical Reports

23. Stanford University
18/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Functional Reports

24. Stanford University
Running MRIQC
19/28

25. Stanford University
20/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Option 3: “bare-metal”
mriqc <bids_dir>/ out/ participant
Requires a functional python environment and installation through Pypi or
setuptools.

26. Stanford University
21/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Option 2: Singularity
poldracklab_mriqc_0.9.0-rc1-2017-01-12-9d72afa28286.img
<bids_dir>/ out/ participant
Image available in sherlock:
/share/PI/russpold/singularity_images/
poldracklab_mriqc_0.9.0-rc1-2017-01-12-9d72afa28286.img

27. Stanford University
22/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Option 1: Docker
docker run -v <bids_dir>:/data -v <scratch_dir>:/scratch
-w /scratch poldracklab/mriqc:latest
/data /scratch/out participant

28. Stanford University
23/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Get involved
Documentation:
http://mriqc.readthedocs.io
Example reports:
http://mriqc.org
Q&A and support:
https://neurostars.org/tags/mriqc
Devs:
https://github.com/poldracklab/mriqc

29. Stanford University
Questions
24/28

30. Stanford University
25/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Questions TBD in the focus group
Are there any additional quality metrics that you would like to be added?
Are there any additional plots that you would like to be added?
Would you like to have diffusion MRI IQMs and reports?
Would you like to participate in manual triage/rating sessions of s/f/d MRI?

31. Stanford University
26/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
Acknowledgments
The PoldrackLab

32. Stanford University
Thanks!
27/28

33. Stanford University
28/28
Introduction MRIQC Visual reports Running MRIQC Questions References References
References I
Gorgolewski, Krzysztof J. et al. (2016a). “BIDS Apps: Improving ease of use, accessibility and reproducibility of neuroimaging data analysis methods”. en. In: bioRxiv, p. 079145. DOI:
10.1101/079145. URL: http://biorxiv.org/content/early/2016/10/05/079145.
Gorgolewski, Krzysztof J. et al. (2016b). “The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments”. In: Scientific Data 3, p. 160044.
ISSN: 2052-4463. DOI: 10.1038/sdata.2016.44. URL: http://www.nature.com/articles/sdata201644.
Mortamet, BÃľnÃľdicte et al. (2009). “Automatic quality assessment in structural brain magnetic resonance imaging”. en. In: Magnetic Resonance in Medicine 62.2, pp. 365–372. ISSN:
1522-2594. DOI: 10.1002/mrm.21992. URL: http://onlinelibrary.wiley.com/doi/10.1002/mrm.21992/abstract.
Pizarro, Ricardo A. et al. (2016). “Automated Quality Assessment of Structural Magnetic Resonance Brain Images Based on a Supervised Machine Learning Algorithm”. English. In:
Frontiers in Neuroinformatics 10. ISSN: 1662-5196. DOI: 10.3389/fninf.2016.00052. URL:
http://journal.frontiersin.org/article/10.3389/fninf.2016.00052/abstract.
Price, Ronald R. et al. (1990). “Quality assurance methods and phantoms for magnetic resonance imaging: Report of AAPM nuclear magnetic resonance Task Group No. 1”. In: Medical
Physics 17.2, pp. 287–295. ISSN: 0094-2405. DOI: 10.1118/1.596566. URL:
http://scitation.aip.org/content/aapm/journal/medphys/17/2/10.1118/1.596566.
Woodard, Jeffrey P. and Monica P. Carley-Spencer (2006). “No-Reference image quality metrics for structural MRI”. en. In: Neuroinformatics 4.3, pp. 243–262. ISSN: 1539-2791, 1559-0089.
DOI: 10.1385/NI:4:3:243. URL: http://link.springer.com/article/10.1385/NI:4:3:243.