FA values > 1.0 from MAP extraction

MRI Processing & SCT
1

Hi

I’m running tract-specific DTI analyses across ~100 participants using SCT (PAM50 atlas, MAP extraction method, ~0.9 mm in-plane resolution) with DTI data averaged across C2–C5.

I’m finding FA values exceeding 0.9 — and in some cases exceeding 1.0 — in several tracts, particularly the ventral spinocerebellar tract, ventral reticulospinal tract, and medial reticulospinal tract. Larger, more centrally located tracts (e.g., lateral corticospinal, fasciculus gracilis, dorsal/lateral spinothalamic) show no such issues.

I think this is likely due to the MAP estimation becoming unstable for small/hard difficult to measure tracts? FA values are within the expected bounds when using the weighted average method. The output from the preprocessing and registration steps all look good.

My current thinking is to exclude such tracts entirely from analyses, with the rationale that if the MAP estimate is producing out-of-bounds values in a substantial proportion of participants, the within-bounds estimates for those tracts are also likely unreliable.

I was wondering if anyone else had experienced something similar, and whether there was a recommended set of “reliable” tracts for tract-specific spinal cord DTI studies, or any guidance on which tracts should be avoided?

Thanks

2

I think this is likely due to the MAP estimation becoming unstable for small/hard difficult to measure tracts?

Yes, excellent diagnosis. This is likely the case. Note however that, depending on the algorithm you use to extract DTI metrics, negative FA, or FA above 1 are possible, see these discussions:

My current thinking is to exclude such tracts entirely from analyses, with the rationale that if the MAP estimate is producing out-of-bounds values in a substantial proportion of participants, the within-bounds estimates for those tracts are also likely unreliable.

Good thinking.