Help interpreting the FBA results

1

Dear experts,

I followed the FBA tutorial and finished with my single shell data (30 b=1000, 3 b0). Then I used TractSeg to get the fixel mask of 72 tracts and extracted the significant fixel numbers of each tract.

To my surprise, compared to healthy controls, I observed decreased FD at some tracts in patients, but FC and FDC didn’t vary. Is this reasonable?

Best wishes,
Yuting

2

It has been a long time but I still can’t figure out a reasonable explanation. Could anyone offer insight into my results? I would greatly appreciate any advice and possible explanations.

3

Hi @hyting,

There’s nothing unreasonable about what you’re reporting: FD could change to a significant extent without any significant changes in the other two measures. Bear in mind when you say ’ FC & FDC didn’t vary’, you should add ‘to a sufficient extent to reach statistical significance’. Absence of significance does not imply absence of effect: it just means whatever effect there might be is too small relative to the variance to be distinguishable from noise.

In general though, you won’t get much response without providing and detail about your cohort or what condition you’re looking at. You’ll need to provide more details about your project before anyone can provide any meaningful advice…

All the best,
Donald

4

Hello @jdtournier,

Thank you for your response and clarification. I apologize for my previous lack of precision in the description. Below, I will provide some supplementary information regarding my study:

  1. There are three groups of participants in my study: healthy control group (20 participants), disease Subgroup 1 (9 patients), and disease Subgroup 2 (13 patients). DTI data (30 b=1000, 3 b=0) were used, and therefore, the SS3T-CSD algorithm was applied to estimate the FODs.
  2. My study aimed to investigate differences in FBA metrics (FD, FC, and FDC) among the three groups of participants. To achieve this, I selected 10 healthy participants, 5 patients from Subgroup 1, and 5 patients from Subgroup 2 to generate a study-specific population template. Differences among the three groups were compared using t-tests respectively (healthy control group vs. disease subgroup 1, healthy control group vs. disease subgroup 2, and disease subgroup 1 vs. disease subgroup 2), with FWE correction applied and significance threshold set at p < 0.01.
  3. The results showed that in some fiber tracts, (1) FD exhibited significant inter-group differences, while FDC and FC did not; or (2) FC showed significant inter-group differences, while FDC and FD did not; or (3) FDC displayed significant inter-group differences, while FD and FC did not.

Based on my study methodology and results, I have some confusion and would appreciate your insights:

  1. Would employing one-way ANOVA instead of multiple t-tests render more reliable results? I attempted to perform one-way ANOVA using fixelcfestats -ftests -fonly, but I don’t know how to conduct post-hoc tests after one-way ANOVA in MRtrix3. Could you please guide me?
  2. Based on your reply, could I consider the three types of results that I mentioned above to be reasonable?

Absence of significance does not imply absence of effect: it just means whatever effect there might be is too small relative to the variance to be distinguishable from noise.

Thank you very much for your response and assistance.

Best regards,
Yuting