Dear FSL experts,
I have two groups of participants (20 patients and 20 healthy). The subjects have PET images. These images were generated using five different pipelines.
I studied the difference in PET signal between these two groups using five voxel wise analysis (in every analysis I fed PET images as an output of one pipeline). Randomise was employed in these five analysis (i.e. the same command, design and contrasts were applied in these five voxel-wise analyses). I would like to find out which pipeline will give the best difference between the groups. In this case:
1. Is it enough just to threshold the statistical maps and see which analysis has the lowest P values in the clusters of the significant difference between the groups?
2. Do I need to evaluate the variance between these five voxel-wise analyses. If so, how can I estimate the variance of the voxel wise analyses?
3. Do I need to run sensitivity analysis to compare the results between these voxel-wise analyses? Are there any tools in FSL that can help to do sensitivity analysis for the voxel wise analyses, or just I need to extract the PET signal in specific ROI then run ROC analysis in any statistical software?
Your guidance is highly appreciated!
John
|