OHBM 2022 | 95 | Talk | Fernando M Ramirez | 7T BOLD fMRI of human V1 reveals an orientation-tuned…

Title: 7T BOLD fMRI of human V1 reveals an orientation-tuned signal consistent with a columnar organization.
Session: Talk
Speaker: Fernando M Ramirez
The potential of neuroimaging methods to reveal information encoded in cortical columns remains debated. Decoding of grating orientation from V1 activation patterns has been hypothesized to leverage the biased sampling of cortical columns (Haynes & Rees, 2005; Kamitani & Tong, 2005). However, other work suggests it leverages coarse-scale maps, not columns (Freeman et al., 2011). Such maps are attributed to grating orientation interacting with the annulus or “vignette” of the stimulus; a mechanism modeled by adding at each image location energy from orientation-tuned filters spanning a full rotation, thus leading to an unbiased representation (Carlson, 2014; Roth et al., 2018). Here, we tested a key prediction of this model: the dissimilarity structure of brain activation patterns should be anticorrelated for opposite phases of a suitably chosen radial-frequency modulator used to vignette oriented gratings (Cichy, Ramírez, and Pantazis, 2015). We found that the dissimilarity structure of fMRI patterns as a function of grating angular disparity significantly correlated with the same orientation-tuned representation (p0.05, signtest; p0.05 in 5/6 subjects), regardless of phase of the modulators. We did not find evidence of the interaction of pattern dissimilarity structure with radial-frequency phase predicted by the 'vignetting' model. Interestingly, the detected orientation-tuned signal was also found to generalize across stimulus contrast levels (p0.05, signtest; p0.05, 5/6 subjects). A model considering the biased sampling of orientation-tuned neurons (arranged in columns) (Ramírez et al., 2014) and contrast response functions informed by measurements from macaques matched our observations.