fMRIPrep Bootcamp | Day 1

name: title
layout: true
class: center
---
layout: false
count: false

.center[
<a href="https://www.nipreps.org/assets/fmriprep-bootcamp-geneva2024/day1-01-fmriprep-primer/">
 <img src="images/qr-talk-url.svg" alt="workflow" style="width: 20%" />
 
 https://www.nipreps.org/assets/fmriprep-bootcamp-geneva2024/day1-01-fmriprep-primer/
</a>

## Overview of the fMRI neuroimaging pipeline & *fMRIPrep*

Chris Markiewicz <<code>markiewicz@stanford.edu</code>>

Oscar Esteban <<code>phd@oscaresteban.es</code>>

]

???

NOTES TO OURSELVES GO HERE

---
name: newsection
layout: true

.perma-sidebar[

<a rel="license" href="http://creativecommons.org/licenses/by/4.0/"><img alt="Creative Commons License" style="border-width:0; height: 20px; padding-top: 6px;" src="https://i.creativecommons.org/l/by/4.0/88x31.png" /></a>
 Day 1 :: Introduction

]

---

# Outlook

.right-column3.center[
<a href="https://www.nipreps.org/assets/fmriprep-bootcamp-geneva2024/day1-01-fmriprep-primer/">
 <img src="images/qr-talk-url.svg" alt="workflow" style="width: 100%" />
 
 https://www.nipreps.org/assets/fmriprep-bootcamp-geneva2024/day1-01-fmriprep-primer/
</a>
]

.left-column3.larger[

* Overview of the neuroimaging pipeline structure

* Why do data require preprocessing?

* *fMRIPrep*

* Introduction
    * Design and requirements
    * Community
    * Continuous validation

* *NiPreps*

]

???

---

# The neuroimaging workflow

.boxed-content[
<img src="images/neuroimaging-workflow-large.svg" alt="workflow" style="width: 100%" />

.center[
[Esteban et al., (2020)](http://doi.org/10.1038/s41596-020-0327-3);
[Niso et al., (2022)](https://doi.org/10.1016/j.neuroimage.2022.119623)
]
]

???

The selling points are:

- Data collection and organization is mostly data management. It's true sequences improve over time and that demands updates in the data management workflow (e.g., BIDS conversion), but you can get a very stable solution here (esp. if you use things like ReproIn and datalad) and standardization has been implemented (following different approaches and standards at each institution, true).
- Statistical modeling requires some flexibility to come up with new and better models, so you may allow some analytical variability not to styme progress.
- Preprocessing is the step in between, where standardization can (i) give you as an individual researcher a solid ground for this and future studies; (ii) make your study more comparable to others who standardize the same way.

WDYT?

---

# Why do data require preprocessing?

.boxed-content[

.large.center[
 MRI measurements *generally* **cannot** be directly analyzed. 
]

.pull-left[
.center.larger[ **Spatiotemporal location**]

signal drawn from the same location and accurate time at all time points,
sampled consistently with the analysis' design (e.g., surface, volume)
]

.pull-right[
.center.larger[ **Signal *validity***]

extraction of confounds, identification/accounting for artifacts, spatiotemporal filtering for denoising, etc.
]

.boxed-bottom.large.center[
**End goal**: minimize false positives without increasing false negatives
]
]

---

# Why do data require preprocessing?

.boxed-content[

.large[
All analyses require consistency and accuracy in **spatiotemporal location**.
]

.larger[

**within runs** (head-motion, slice-timing),

**within individuals** (coregistration between runs, coregistration with anatomy [T1w images, surfaces, etc.], susceptibility distortion), and

**across subjects** (spatial normalization)

]

---

## Example: susceptibility distortion

---

# Why do data require preprocessing?

.boxed-content[
.large[
All analyses must consider confounders and covariates that undermine the **validity** of the measurements.
]
.larger[
* confounders such as global signals, signal drifts, DVARS, etc.

* spatiotemporal filtering to increase SNR,

* known artifacts such as head-motion parameters and derivations, etc.
]

.boxed-bottom.large[
**Minimal preprocessing**: .gray-text[*fMRIPrep* is *conservative* in that it will not regress out signals or apply intended smoothing kernels]
]
]

---

.boxed-content[
.large.center[
# Processing and reproducibility
]
.larger.center.gray-text[How do they interact? Why *fMRIPrep*?
]]

---

.center[
<img src="images/reproducible-definition-grid.svg" alt="HPC" style="width: 68%" />
.small[
*The Turing Way project* illustration by Scriberia. Used under a CC-BY 4.0 licence. doi:<a href="https://doi.org/10.5281/zenodo.3332807">10.5281/zenodo.3332807</a>.]
]

---

# Reproducibility: increasing measurements' reliability

.large[
 Repeat the measurement

.gray-text[Random errors cancel out]

Standardize the measurement procedure

.gray-text[Reduce methodology variance (potentially at the cost of bias)]
]

.center.large[
Both strategies *reduce the variance* of the measurement.

Neither ensures the *validity* of the measurement]

---

# Preprocessing & Reproducibility

.boxed-content[

.large[
The reproducibility of preprocessing sets bounds to the reproducibility of downstream analyses.
]

.larger[
* NARPS ([Botvinik-Nezer et al., 2020](https://doi.org/10.1038/s41586-020-2314-9)): single dataset, 70 teams, 9 ex-ante hypotheses.

Striking analytical variability (even with teams using same preprocessing)

* [Li et al., (2024)](https://doi.org/10.1038/s41562-024-01942-4): single test-retest dataset, 5 pipelines

Moderate inter-pipeline agreement, limiting cross-study <s>reproducibility</s> [replicability]
]
]

---

# Preprocessing & Reproducibility

.boxed-content[

.large[
Beyond analytical variability, other sources of variability are on the way:
]
.larger[
* Random seeds

Keep track (*do not fix*) and report.

* [Chatelain et al., (2023)](https://doi.org/10.48550/arXiv.2307.01373): Random rounding of floating-point calculations.

Uncovered substantial changes between patch-releases of *fMRIPrep*
]

]

---

# fMRIPrep: bird's eye picture

.boxed-content.center[
 
 
<img src="https://github.com/oesteban/fmriprep/raw/f4c7a9804be26c912b24ef4dccba54bdd72fa1fd/docs/_static/fmriprep-21.0.0.svg" width="100%" />
]

---

.pull-left.center[
<img src="images/fmriprep-anatomical.png" width="90%" />
]

.pull-right[

# Anatomical processing

.large[
Delivered within *sMRIPrep*
]

.larger[

Denoising

INU correction

Averaging (multi-session)

Brain extraction

Spatial normalization

Brain tissue segmentation

Surface reconstruction
]

]

---

.pull-left.center[
 
 
<img src="images/fmriprep-functional.png" width="100%" />
]

.pull-right[

# Functional processing

.larger[

Reference volume

Slice-timing (optional)

Head-motion (estimation)

Susceptibility distortion (estimation)

Co-registration with anatomical T1w

Resampling into specific spaces

Confound collection
]

]

---

# *TemplateFlow*

.boxed-content.center[
<img src="https://www.templateflow.org/assets/templateflow_fig-birdsview.png" width="70%" /> 
(<a href="https://doi.org/10.1038/s41592-022-01681-2">Ciric et al., 2022</a>)
]

---

# *SDCFlows*

.boxed-content[
.large[
SDC Susceptibility-derived Distortion Correction
]

.larger[
*SDCFlows* "caters" preprocessed fieldmap estimations for *fMRIPrep* to reconstruct the nonlinear displacements field to revert geometrical distortion.
]

.center[<img src="images/sdc-the-problem.svg" width="40%" />]
]

---

# fMRIPrep's design

.boxed-content[

.large[ Robust & adaptive fMRI preprocessing]

.larger[
* Minimalist requirements (T1w + BOLD run)
* Will utilize additional scans (T2w/FLAIR, field maps) when available
* Capable of preprocessing multi-echo BOLD.
]

.large[ Cross-suite tooling]
.larger[
* Select tools from each suite according to strengths
* Use *Nipype* to abstract from tool details and construct modular workflows
]

.large[ Standardization and best-practices — [check (Esteban, 2024)](https://doi.org/10.31219/osf.io/42bsu)]
.larger[
* BIDS (inputs), BIDS Derivatives (outputs), BIDS Apps (architecture).
* Containers, versioning, LTS, open-source, CI/CD, etc.
* Visual reports
]

]

---

# fMRIPrep's constrains and limitations

.boxed-content[

.large[ Agnostic to downstream analysis]

.larger[
* Minimalist processing that refrains from applying analysis-specific steps.
]

.large[ Cross-sectional studies]
.larger[
* *fMRIPrep* assumes individuals' anatomy does not change
* Today, execution can be *hacked* but a *longitudinal mode* does not exist yet.
]

.large[ Driven by the 80/20 principle]
.larger[
* 7T data fell on the 20 side (although that's about to change)
]

]

---

# Standardization in the research workflow

.boxed-content[
 
.center[
<img src="images/neuroimaging-workflow-large.svg" alt="workflow" style="width: 70%" />
]

.larger[
Data collection easy to standardize (BIDS, scanners' software)

Statistical modeling keep analytical flexibility not to stifle the development of new ideas

Preprocessing offers an **opportunity** to reduce analytical flexibility and **enable** cross-study comparisons
]

]

---

# Standardizing preprocessing: visual reports

.boxed-content[
<iframe src="https://fmriprep.s3.amazonaws.com/bootcamp-geneva-2024/sub-15.html" width="100%" height="540px" style="border: 0; margin-top: 15px" />
]
---

# Standardizing preprocessing: community building

.large[fMRI practitioners have massively adopted *fMRIPrep* globally:]

---

# Standardizing preprocessing: community building

.large.center[ which has permitted the creation of a large community.]

---

# Building a community was important to *fMRIPrep*

.boxed-content.large[
.no-bullet[
* Feedback & support:
 * error & bug reports help development,
 * feature requests help driving a roadmap,
 * questions are the entry point to deliver support
 * [ NeuroStars.org](https://neurostars.org)

* Engage contributors (see [our guidelines](https://www.nipreps.org/community/CONTRIBUTING/)):
 * documentation, assistance debugging, code patches, everything counts!

* Reach out, increase user-base.
]
]

---

# Standardizing preprocessing: software versions

.larger[
 It's critical to report **exact versions**

Semantics inform about compatibility: **24.0.1**

Long-term support (LTS) program.

]

---

# Standardizing preprocessing: validation

.boxed-content.center[
<img src="./images/fmriprep-validation-smoothness.png" width="95%" /> 
(<a href="https://doi.org/10.1038/s41592-018-0235-4">Esteban et al., 2019</a>)
]

---
count:false

# Standardizing preprocessing: validation

.boxed-content.center[
<img src="./images/fmriprep-validation-smoothness-2.png" width="46%" /> 
(<a href="https://doi.org/10.1038/s41592-018-0235-4">Esteban et al., 2019</a>)
]

---
count:false

# Standardizing preprocessing: validation

.boxed-content.center[
<img src="./images/fmriprep-validation-workflow.png" width="40%" /> 
(<a href="https://doi.org/10.1038/s41592-018-0235-4">Esteban et al., 2019</a>)
]

---

<img src="../../nipreps-chart.png" width="63%" /> 
NiPreps (<a href="https://doi.org/10.31219/osf.io/ujxp6">Esteban et al., 2020</a>)

---

# *NiPreps* produce *analysis-grade* data

.boxed-content[
 
.larger.center[
"*analysis-grade*" data data **directly consumable by analyses**
]

.pull-left[
 
 
*Analysis-grade* data is an analogy to the concept of "*sushi-grade (or [sashimi-grade](https://en.wikipedia.org/wiki/Sashimi)) fish*" in that both are:

.large[ **minimally preprocessed**,]

and

.large[ **safe to consume** directly.]
]

.pull-right.center[
<img src="../../nipreps-transparent.png" width="100%" />
<a href="https://www.nipreps.org"><img src="images/qr-nipreps-url.svg" width="40%" /> 
www.nipreps.org</a>
]
]

???

---

## Highlights

.boxed-content[
.no-bullet[
* .large[ MRI data cannot directly be analyzed]
 * Spatiotemporal location, and signal 'validity'
 * **Analytical variability**

* .large[ Preprocessing is an area with standardization potential]
 * *fMRIPrep* as an an example of implementation in fMRI
 * *fMRIPrep* design principles
 * **Visual reports**

* .large[ Introduced *NiPreps* as a generalization of *fMRIPrep*]
 * Concept of *analysis-grade data*
 * Tooling landscape
]]

???

---
layout: false
count: false

.center[
<a href="https://www.nipreps.org/assets/fmriprep-bootcamp-geneva2024/day1-01-fmriprep-primer/"><img src="images/qr-talk-url.svg" alt="workflow" style="width: 20%" /></a>

## Thanks

#### Chris Markiewicz <<code>markiewicz@stanford.edu</code>> & Oscar Esteban <<code>phd@oscaresteban.es</code>>

Overview of the fMRI neuroimaging pipeline & *fMRIPrep*

Funding: [SNSF 185872](https://data.snf.ch/grants/grant/185872), [RF1MH121867](https://reporter.nih.gov/project-details/10260312), [CZI EOSS5-000266](https://chanzuckerberg.com/eoss/proposals/nipreps-a-community-framework-for-reproducible-neuroimaging/)

]

???