Process and QC multiple M0 images independently before combination

[tsalo]

Summary

ASL acquisitions may include multiple M0 images, including repeated M0 volumes embedded in the ASL series or separate calibration scans. Early averaging can hide motion, corruption, coverage differences, or signal drift between the M0 observations.

ASLPrep could retain and process individual M0 images before combining them. The HCP-ASL pipeline provides a precedent by applying corrections and registration separately to its two calibration images.

Additional details

Independent M0 processing could be useful when:

• M0 images are acquired before and after the ASL time series.
• Included M0 volumes are separated in time.
• The participant moves between M0 observations.
• One M0 image contains dropout or reconstruction artifacts.
• The imaging slab or distortion differs between M0 observations.
• Signal drift occurs during a long acquisition.
• M0 and ASL images require separate registration.

For each M0 observation, ASLPrep could:

1. Preserve the original volume and acquisition order.
2. Apply applicable distortion and bias corrections.
3. Estimate alignment to the ASL reference.
4. Calculate brain coverage and registration quality.
5. Measure global and tissue-specific intensity.
6. Compare the image with the other M0 observations.
7. Assign a combination weight or exclusion reason.

Possible combination methods include:

--m0-combine {robust-mean,mean,median,first,last}

When more than one M0 is available, a robust mean could be selected automatically, with the option available to reproduce a simple mean or select a specific observation.

Potential outputs include:

• Each corrected and ASL-aligned M0 image.
• The combined M0 image.
• A voxelwise M0 standard-deviation or disagreement image.
• Combination weights.
• A TSV containing motion, intensity, coverage, and exclusion metrics.
• A reportlet comparing all M0 observations.

If M0 observations were acquired before and after the ASL series, a later extension could evaluate whether time-dependent M0 interpolation improves calibration. That behavior could remain separate from the initial robust-combination feature.

M0 images acquired with materially different preparation, gain, scaling, or readout settings may not be directly combinable. Metadata checks could identify such cases and retain the scans separately.

References:

• HCP-ASL calibration-correction implementation
• Kirk et al. (2025), HCP-ASL pipeline
• ASLPrep processing documentation
• BIDS ASL calibration metadata

Next steps

• Inventory current handling of included and separate M0 volumes.
• Preserve individual M0 images through alignment and correction.
• Define M0 quality and compatibility metrics.
• Implement robust combination and user-selectable alternatives.
• Save combination weights and exclusion reasons.
• Add individual-M0 and combined-M0 reportlets.
• Test motion, drift, corruption, and incompatible-metadata cases.