Frequently Asked Questions

Below are some questions we have received. If anything is unclear, please feel free to raise an issue, and we will post the answer here.

The reconstructed human surface is part of NSD dataset, and you can access it here (Subject 5). In our paper, we use functional beta (extracted and saved in Data/others/FMRI.zip) and the structural surfaces in each subject’s native space (fsnative). Most of the computations were performed in MATLAB, and the visualization of statistical maps on the cortical surface was carried out using python toolbox Nilearn. Example notebooks can be found [here] and [here]. Another great tool to plot surface stats map is pycortex.

Thank you for your interest! Researchers are welcome to build on findings from our dataset and generate new hypotheses. For questions that can be addressed using the same experimental setup—specifically, experiments where monkeys fixate on images without requiring additional behavioral tasks—it is possible to contact us (Pinglei Bao) to discuss potential collaborations and further validation on the same experimental platform. This approach allows new insights to be tested efficiently while leveraging the existing dataset and infrastructure.

For each recording site, we first registered the structural scan containing the tungsten electrodes to the high-resolution structural scan using AFNI’s align_epi_anat.py. The registered image was then imported into 3D Slicer, where the crosshair was manually aligned to the probe to read the expected electrode coordinates. These coordinates were subsequently transformed to the NMT template space using the registration files obtained from the @animal_warper command, which applies a three-step process to align the high-resolution structural scan to the NMT template. The resulting electrode coordinates on the template are stored in AreaXYZ.xls.
To generate Figure S1, we opened the NMT template in AFNI, navigated to the electrode coordinates, and projected them onto the cortical surface using SUMA. More details on this procedure can be found in the downloaded markdown file though here.

No, the current triple-N dataset is complete and includes 90 sessions (from V1, V2, V4, PIT, CIT to AIT). However, we are carrying out related projects on visual coding using the same experimental protocol and platform, with more specific focuses, such as category-specific coding and the perceptual significance of IT responses. The resulting datasets will be made available in the future as a supplement to the triple-N dataset, analogous to how the NSD-OOD dataset complements NSD-Core.