-volreg_warp_final_interp and -tshift_interp when comparing multi-echo and single-echo analyses


This question may be pedantic or in practice the difference may be negligible. However, it would be nice to hear your opinion to understand a bit more details. Should I set -volreg_warp_final_interp wsinc5 and -tshift_interp -wsinc9 for single echo (e.g., the 2nd echo) analysis if I would like to compare the GLM results computed using multi-echo?

In the afni_proc.py examples, volreg_warp_final_interp and tshift_interp only appear in multi-echo analysis, especially in the TEDANA note section. In APMULTI_Demo1_rest, all multi-echo related do_[1-26]_ap_me_*.tsch scripts have these two options including the optimal combined method. do_20_ap_se.tsch for single echo does not have them.

My understanding is wsinc5 or 9 preserves better the sharpness of the image, would it be harmful to also apply for single echo analysis just to have a closer apple to apple comparison? Probably there are other considerations suggesting to not use these two options for single echo? Thanks a lot.


Hi Zhengchen,

Applying -volreg_warp_final_interp wsinc5 will indeed make the result look sharper, and it will generally lower the estimated blur in the data, and so should keep the results more similar. Use of -tshift_interp -wsinc9 is less common.

The reason that we have historically avoided these interpolants (including Fourier) is that they can cause ringing artifacts. As with so many things, there are tradeoffs...

  • rick

Hi Rick,

Thanks for your explanation. It is a good point regarding the blur estimation. I guess I will consider -tshift_interp as a special option for multi-echo.