Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

How Do You Know It’s Real? Controlling False Positives in Whole-Brain Neural Decoding

Zihan Li¹, Timothy T. Rogers¹; ¹University of Wisconsin - Madison

Presenter: Zihan Li

Whole-brain neural decoding can reveal distributed neuro-cognitive representations, but high-dimensional fMRI data make feature selection vulnerable to false positives. We introduce Knockoff-Max (KO-Max), a resampling-based false-positive control framework, and evaluate it with Iterated LASSO (iLASSO), a whole-brain decoding method designed to recover distributed informative voxels. In simulations with controlled ground-truth signals, KO-Max maintained false-positive rates below 0.05 across multiple noise levels. Applied to an existing fMRI dataset decoding visual stimulus category (face, place, or object), KO-Max yielded a sparser but still spatially distributed set of selected voxels after FPR control. These results suggest that distributed neural representations can remain detectable after explicit feature-level false-positive control.

Topic Area: Methods, Tools, Theory & Neural Coding