# *SDCFlows* manuscript

###### tags: `Collecting-feedback`

## The gist

Demonstrate *SDCFlows* in estimating the fieldmap (and evaluate the correction) in a simple task fMRI dataset:

* GRE (phasediff)
* PEPOLAR (four directions AP/PA/LR/RL available for every session, so we can test combinations thereof)
* SyN (we have T1w and T2w for every session)

Questions:

* Q1: What are the characteristics of each option? More local, more global, etc.
* Comparison of the different options (I would expect GRE to be a smoother estimation compared to PEPOLAR and to SyN)

* Q2: Evaluate the different options in obtaining group activation maps

* Q3: Demonstrate with one simple example generalization to rodents

Impact:

* Understand what fieldmap strategies we should use
* Trade-offs of different strategies
* Value in implementation:
* formalizing all fieldmaps with B-Splines
* fit & transform paradigm
* transform model (value in combining, e.g., with head motion)
* Value in combining strategies (e.g., GRE+PEPOLAR)
* Value in standardizing approach across modalities (e.g., a dMRI or ASL example?) and species

## Data

We would use the HCPh dataset:

* One subject, 36 sessions
* fMRI: multi-echo EPI (with phase and magnitude) of a ~3min task with finger-tapping, gaze movement, and a visual grating (a quality control task).
* fieldmaps (see above) - all sessions
* T1w and T2w (all sessions)

For the rodents, some data Eilidh can provide.

## Experiments

### Q1 - Fieldmap estimation and characterization

**Methods**

- Calculate T1w&T2w templates (outside fMRIPrep)
- Estimate fieldmaps with *SDCFlows*:
- GRE/phasediff
- PEPOLAR: AP/PA
- PEPOLAR: LR/RL
- PEPOLAR: AP/LR
- PEPOLAR: AP/RL
- PEPOLAR: PA/LR
- PEPOLAR: PA/RL
- PEPOLAR: AP/PA/RL/LR
- SyN: Template T1w
- SyN: Template T2w
- GRE + PEPOLAR AP/PA
- GRE + PEPOLAR LR/RL

We will generate two sets of fieldmaps for each above:
- 1-level of coefficients (topup-like)
- 2-levels of coefficients (*SDCFlows* enables this and it is the vision it could improve the estimation)

- Align all coefficients of fieldmaps on the T1w template (fieldmap reference to T1w registration).

**Endpoints**

- Visual inspection (evaluate reports)
- 2D Frequency domain analysis (to see global vs. local solutions)
- Jacobian variability across sessions, *uncertainty* map
- Fieldmap quality metrics?

### Q2 - Activation maps

**Methods**

- fMRIPrep the task data (requires some language to "fast-track" the fieldmaps as precomputed output)
- Generate 1st and 2nd level activation maps

**Endpoints**

- Visual inspection of correction
- BOLD variability maps per block in task
- Group differences between correction strategies, per task's block
- Effect of jacobian modulation?

### Q3 - A rodent example

**Methods**

- Remark how rodents require adaptations

**Endpoint**

- A dataset with visual assessment of before/after?

## Questions/comments

* Comparison vs. TOPUP as a baseline?
*