[2T6S] Pipeline reproduction (SPM - deriv)

[bclenet]

Softwares

SPM

Input data

derivatives (fMRIprep)

Additional context

Pipeline reproduction is almost done, only one issue remains.

For the hypotheses 5 and 7, correlations values between reproduced maps and results are below zero (~= -0.99). The reason for that is that the team provided statistical maps in which negative values represent negative activation, although the opposite was expected in the NARPS study.

Therefore, results were modified before analysis. We need to do the same for the project. Changes need to be done on the results (narps_open.data.results) instead of in the pipeline code.

---

Here is the description of the 2T6S results in their neurovault collection.

• hypo5_unthresh SPM{T_[53.0]} - contrast 1: Gamble_Loss
• hypo6_unthresh SPM{T_[53.0]} - contrast 2: - (Gamble_Loss)
• hypo7_unthresh SPM{T_[53.0]} - contrast 2: - (Gamble_Loss)
• hypo8_unthresh SPM{T_[53.0]} - contrast 1: Gamble_Loss

Here are the value ranges in the images:

• hypo5_unthresh [-2.92893, 0.97576]
• hypo6_unthresh [-0.70500, 2.22962]
• hypo7_unthresh [-0.97576, 2.92893]
• hypo8_unthresh [-2.22962, 0.70501]

[bclenet]

Update

During the NARPS study, Botvinik-Nezer performed a rectification (see : https://github.com/poldrack/narps/blob/master/ImageAnalyses/original_notebooks/PrepareMaps.ipynb) in which statistical maps for hypotheses 5 and 6 were multiplied by -1 for this reason:

Negative values in the images represent negative activation

After applying this rectification to the 2T6S neurovault correction, and with the current version of the code (PR #75 , commit e56bb63) we get the following correlation coefficients for hypotheses 1 to 9 (with 20 subjects) : [0.63, 0.51, 0.63, 0.51, -0.62, 0.57, 0.62, -0.57, 0.45].

Therefore, the code will now return the following files, so the correlation values are all positive :

• hypothesis 5 :
  • l2_analysis_equalIndifference_nsub_{nb_sub}/_contrast_id_0003/_threshold0/spmT_0001_thr.nii
  • l2_analysis_equalIndifference_nsub_{nb_sub}/_contrast_id_0003/spmT_0001.nii
• hypothesis 8 :
  • l2_analysis_equalRange_nsub_{nb_sub}/_contrast_id_0003/_threshold1/spmT_0002_thr.nii
  • l2_analysis_equalRange_nsub_{nb_sub}/_contrast_id_0003/spmT_0002.nii

The reasons for this issue are something we need to further understand.